Welcome to Knowledge Base!

Volume Sizing and Count

A common question when first provisioning storage on the FlashArray is what capacity should I be using for each volume? VMware VMFS supports up to a maximum size of 64 TB. The FlashArray supports far larger than that, but for ESXi, volumes should not be made larger than 64 TB due to the filesystem limit of VMFS.

Using a smaller number of large volumes is generally a better idea today. In the past a recommendation to use a larger number of smaller volumes was made for performance limitations that no longer exist.

This limit traditionally was due to two reasons :

• VMFS scalability issues due to locking
• Per-volume queue limitations on the underlying array.

VMware resolved the first issue with the introduction of Atomic Test and Set (ATS), also called Hardware Assisted Locking.

Prior to the introduction of VAAI ATS, VMFS used LUN-level locking via full SCSI-2 reservations to acquire exclusive metadata control for a VMFS volume. In a cluster with multiple nodes, all metadata operations were serialized and hosts had to wait until whichever host, currently holding a lock, released that lock. This behavior not only caused metadata lock queues but also prevented standard I/O to a volume from VMs on other ESXi hosts which were not currently holding the lock.

With VAAI ATS, the lock granularity is reduced to a much smaller level of control (specific metadata segments, not an entire volume) for the VMFS that a given host needs to access. This behavior makes the metadata change process not only very efficient, but more importantly provides a mechanism for parallel metadata access while still maintaining data integrity and availability. ATS allows for ESXi hosts to no longer queue metadata change requests, which consequently speeds up operations that previously had to wait for a lock. Therefore, situations with large amounts of simultaneous virtual machine provisioning operations will see the most benefit.

The standard use cases benefiting the most from ATS include:

• High virtual machine to VMFS density.
• Extremely dynamic environments—numerous provisioning and de-provisioning of VMs (e.g. VDI using non-persistent linked-clones).
• High intensity virtual machine operations such as boot storms, or virtual disk growth.

The introduction of ATS removed scaling limits via the removal of lock contention; thus, moving the bottleneck down to the storage, where many traditional arrays had per-volume I/O queue limits. This limited what a single volume could do from a performance perspective as compared to what the array could do in aggregate. This is not the case with the FlashArray.

A FlashArray volume is not limited by an artificial performance limit or an individual queue. A single FlashArray volume can offer the full performance of an entire FlashArray, so provisioning ten volumes instead of one, is not going to empty the HBAs out any faster. From a FlashArray perspective, there is no immediate performance benefit to using more than one volume for your virtual machines.

The main point is that there is always a bottleneck somewhere, and when you fix that bottleneck, it is transferred somewhere in the storage stack. ESXi was once the bottleneck due to its locking mechanism, then it fixed that with ATS. This, in turn, moved the bottleneck down to the array volume queue depth limit. The FlashArray doesn’t have a volume queue depth limit, so now that bottleneck has been moved back to ESXi and its internal queues.

Altering VMware queue limits is not generally needed with the exception of extraordinarily intense workloads. For high-performance configuration, refer to the section of this document on ESXi queue configuration.

VMFS Version Recommendations

Pure Storage recommends using the latest supported version of VMFS that is permitted by your ESXi host.

For ESXi 5.x through 6.0, use VMFS-5. For ESXi 6.5 and later it is highly recommended to use VMFS-6. It should be noted that VMFS-6 is not the default option for ESXi 6.5, so be careful to choose the correct version when creating new VMFS datastores in ESXi 6.5.

Example of VMFS-6  in ESXi 6.5:

When upgrading to ESXi 6.5, there is no in-place upgrade path from VMFS-5 to VMFS-6. Therefore, it is recommended to create a new volume entirely, format it as VMFS-6, and then Storage vMotion all virtual machines from the old VMFS-5 datastore to the new VMFS-6 datastore. Once the migration is completed you can then delete and remove the VMFS-5 datastore from the ESXi host and FlashArray.

Datastore Performance Management

ESXi and vCenter offer a variety of features to control the performance capabilities of a given datastore. This section will overview FlashArray support and recommendations for these features.

esxcli storage core device list –d <naa.xxxxx>
Device Max Queue Depth: 96
No of outstanding IOs with competing worlds: 64

Queue Depth Limits

ESXi offers the ability to configure queue depth limits for devices on an HBA or iSCSI initiator. This dictates how many I/Os can be outstanding to a given device before I/Os start queuing in the ESXi kernel. If the queue depth limit is set too low, IOPS and throughput can be limited and latency can increase due to queuing. If too high, virtual machine I/O fairness can be affected and high-volume workloads can affect other workloads from other virtual machines or other hosts. The device queue depth limit is set on the initiator and the value (and setting name) varies depending on the model and type:

Changing these settings require a host reboot. For instructions to check and set these values, please refer to this VMware KB article:

Changing the queue depth for QLogic, Emulex, and Brocade HBAs

Disk Schedule Number Requests Outstanding (DSNRO)

T here is a second per-device setting called “Disk Schedule Number Requests Outstanding” often referred to as DSNRO. This is a hypervisor-level queue depth limit that provides a mechanism for managing the queue depth limit for an individual device. This value is a per-device setting that defaults to 32 and can be increased to a value of 256.

It should be noted that this value only comes into play for a volume when that volume is being accessed by two or more virtual machines on that host. If there is more than one virtual machine active on it, the lowest of the two values (DSNRO or the HBA device queue depth limit) is the value that is observed by ESXi as the actual device queue depth limit. So, in other words, if a volume has two VMs on it, and DSNRO is set to 32 and the HBA device queue depth limit is set to 64, the actual queue depth limit for that device is 32. For more information on DSNRO see the VMware KB here:

Setting the Maximum Outstanding Disk Requests for virtual machines

In general, Pure Storage does not recommend changing these values. The majority of workloads are distributed across hosts and/or not intense enough to overwhelm the default queue depths. The FlashArray is fast enough (low enough latency) that the workload has to be quite high in order to overwhelm the queue.

If the default queue depth is consistently overwhelmed, the simplest option is to provision a new datastore and distribute some virtual machines to the new datastore. If a workload from a single virtual machine is too great for the default queue depth, then increasing the queue depth limit is the better option.

If a workload demands queue depths to be increased, Pure Storage recommends making both the HBA device queue depth limit and DSNRO equal.

You can verify the values of both of these for a given device with the command:

esxcli storage core device list –d <naa.xxxxx>
Device Max Queue Depth: 96
No of outstanding IOs with competing worlds: 64

Dynamic Queue Throttling

ESXi supports the ability to dynamically throttle a device queue depth limit when an array volume has been overwhelmed. An array volume is overwhelmed when the array responds to an I/O request with a sense code of QUEUE FULL or BUSY. When a certain number of these are received, ESXi will reduce the queue depth limit for that device and slowly increase it as conditions improve.

This is controlled via two settings:

• Disk.QFullSampleSize —The count of QUEUE FULL or BUSY conditions it takes before ESXi will start throttling. Default is zero (feature disabled)
• Disk.QFullThreshold —The count of good condition responses after a QUEUE FULL or BUSY required before ESXi starts increasing the queue depth limit again

The Pure Storage FlashArray does not advertise a QUEUE FULL condition for a volume. Since every volume can use the full performance and queue of the FlashArray, this limit is unrealistically high and this sense code will likely never be issued. Therefore, there is no reason to set or alter these values for Pure Storage FlashArray volumes because QUEUE FULL should rarely (or never) occur.

Storage I/O Control

VMware vCenter offers a feature called Storage I/O Control (SIOC) that will throttle selected virtual machines when a certain average latency has been reached or when a certain percentage of peak throughput has been hit on a given datastore. ESXi throttles virtual machines by artificially reducing the number of slots that are available to it in the device queue depth limit.

Pure Storage fully supports enabling this technology on datastores residing on the FlashArray. That being said, it may not be particularly useful for a few reasons.

First, the minimum latency that can be configured for SIOC before it will begin throttling a virtual machine is 5 ms.

When a latency threshold is entered, vCenter will aggregate a weighted average of all disk latencies seen by all hosts that see that particular datastore. This number does not include host-side queuing, it is only the time it takes for the I/O to be sent from the SAN to the array and acknowledged back.

Furthermore, SIOC uses a random-read injector to identify the capabilities of a datastore from a performance perspective. At a high-level, it runs a quick series of tests with increasing numbers of outstanding I/Os to identify the throughput maximums via high latency identification. This allows ESXi to determine what the peak throughput is, for when the “Percentage of peak throughput” is chosen.

Knowing these factors, we can make these points about SIOC and the FlashArray:

1. SIOC is not going to be particularly helpful if there is host-side queuing since it does not take host-induced latency into account. This (the ESXi device queue) is generally where most of the latency is introduced in a FlashArray environment.
2. The FlashArray will rarely have sustained latency above 1 ms, thus reducing the likely hood that this threshold will be reached for any meaningful amount of time on a FlashArray volume.
3. A single FlashArray volume does not have a queue limit, so it can handle quite a high number of outstanding I/O and throughput (especially reads), therefore SIOC and its random-read injector cannot identify FlashArray limits in meaningful ways.

In short, SIOC is fully supported by Pure Storage, but Pure Storage makes no specific recommendations for configuration.

Storage DRS

VMware vCenter also offers a feature called Storage Dynamic Resource Scheduler (Storage DRS / SDRS). SDRS moves virtual machines from one datastore to another when a certain average latency threshold has been reach on the datastore or when a certain used capacity has been reached. For this section, let’s focus on the performance-based moves.

Storage DRS, like Storage IO Control, waits for a certain latency threshold to be reached before it acts. And, also like SIOC, the minimum is 5 ms.

While it is too high in general to be useful for FlashArray induced latency, SDRS differs from SIOC in the latency it actually looks at. SDRS uses the “VMObservedLatency” (referred to a GAVG in esxtop) averages from the hosts accessing the datastore. Therefore, this latency includes time spent queuing in the ESXi kernel. So, theoretically, a high-IOPS workload, with a low configured device queue depth limit, an I/O could conceivably spend 5 ms or more queuing in the kernel. In this situation Storage DRS will suggest moving a virtual machine to a datastore which does not have an overwhelmed queue.

That being said, this is still an unlikely scenario because:

1. The FlashArray empties out the queue fast enough that a workload must be quite intense to fill up an ESXi queue so much that is spends 5 ms or more in it. Usually, with a workload like that, the queuing is higher up the stack. (in the virtual machine)
2. Storage DRS samples for 16 hours before it makes a recommendation, so typically you will get one recommendation set per-day for a datastore. So this workload must be consistently and extremely high, for a long time, before SDRS acts.

In short, SDRS is fully supported by Pure Storage, but Pure Storage makes no specific recommendations for performance based move configuration.

Datastore Capacity Management

Managing the capacity usage of your VMFS datastores is an important part of regular care in your virtual infrastructure. There are a variety of mechanisms inside of ESXi and vCenter to monitor capacity. Frequently, the concept of data reduction on the FlashArray is seen as a complicating factor, when in reality it is a simplifying factor, or at worse, a non-issue.

Let’s overview some concepts on how to best manage VMFS datastores from a capacity perspective.

VMFS Usage vs. FlashArray Volume Capacity

VMFS reports how much is currently allocated in the filesystem on that volume. Depending on the type of virtual disk (thin or thick), dictates how much is consumed upon creation of the virtual machine (or virtual disk specifically). Thin disks only allocate what the guest has actually written to, and therefore VMFS only records what the virtual machine has written in its space usage. Thick type virtual disks allocate the full virtual disk immediately, so VMFS records much more space as being used than is actually used by the virtual machines.

This is one of the reasons thin virtual disks are preferred—you get better insight into how much space the guests are actually using.

Regardless of what type you choose, ESXi is going to take the sum total of the allocated space of your virtual disks and compare that to the total capacity of the filesystem of the volume. The used space is the sum of those virtual disks allocations. This number increases as virtual disks grow or new ones are added, and can decrease as old ones are deleted, moved, or even shrunk.

Compare this to what the FlashArray reports for capacity. What the FlashArray reports for volume usage is NOT the amount used for that volume. What the FlashArray reports is the unique footprint of the volume on that array.

In the example below we can see that we are using a 5 TB FlashArray volume and VMFS datastorea. The example confirms that the VMFS datastore reports a total of 720.72 GB of used space on the 5TB filesystem. This tell us that there is a combined total of 720.72 GB of allocated virtual disks on this filesystem:

Now let’s look at the FlashArray volume.

The FlashArray volume shows that 50.33 GB is being used. Does this mean that VMFS is incorrect? No. VMFS is always the source of truth. The “Volumes” metric on the FlashArray simply represents the amount of physical capacity that has been written to the volume after data reduction that no other volume shares.

This metric can change at any time as the data set changes on that volume or any other volume on the FlashArray. If, for instance, some other host writes 2 GB to another volume (let’s call it “volume2”), and that 2 GB happens to be identical to 2 GB of that 50.33 GB GB on “sn1-m20-e05-28-prod-ds”, then  “sn1-m20-e05-28-prod-ds” would no longer have 50.33 GB of unique space. It would drop down to 48.33 GB, even though nothing changed on “sn1-m20-e05-28-prod-ds” itself. Instead, another application just happened to write similar data, making the footprint of “sn1-m20-e05-28-prod-ds” less unique.

For a more detailed conversation around this, refer to this blog post:

http://www.codyhosterman.com/2017/01/vmfs-capacity-monitoring-in-a-data-reducing-world/

Why doesn’t VMFS report the same used capacity as the FlashArray for the underlying volume?

Well, because they mean different things. VMware reports what is allocated on the VMFS and the FlashArray reports what is unique to the underlying volume. The FlashArray value can change constantly. The FlashArray metric is only meant to show how reducible the data on that volume is internal to the volume and against the entire array. Conversely, VMFS capacity usage is based solely on how much capacity is allocated to it by virtual machines. The FlashArray volume space metric, on the other hand, actually relates to what is also being used on other volumes. In other words, VMFS usage is only affected by data on the VMFS volume itself. The FlashArray volume space metric is affected by the data on all of the volumes. So the two values should not be conflated.

For capacity tracking, you should refer to the VMFS usage. How do we best track VMFS usage? What do we do when it is full?

Monitoring and Managing VMFS Capacity Usage

As virtual machines grow and as new ones are added, the VMFS volume they sit on will slowly fill up. How to respond and to manage this is a common question.

In general, using a product like vRealize Operations Manager (vROps) with the FlashArray Management Pack is a great option here. But for the purposes of this document we will focus on what can be done inside of vCenter alone.

You need to decide on a few things:

• At what percentage full of my VMFS volume do I become concerned?
• When that happens what should I do?
• What capacity value should I monitor on the FlashArray?

The first question is the easiest to answer. Choose either a percentage full or at a certain capacity free. Do you want to do something when, for example, a VMFS volume hits 75% full or when there is less than 50 GB free? Choose what makes sense to you.

vCenter alerts are a great way to monitor VMFS capacity automatically. There is a default alert for datastore capacity, but it does not do anything other than tag the datastore object with the alarm state. Pure Storage recommends creating an additional alarm for capacity that executes some type of additional action when the alarm is triggered.

Configuring a script to run, an email to be issued, or a notification trap to be sent greatly diminishes the chance of a datastore running out of space unnoticed.

The next step is to decide what happens when a capacity warning occurs.

There are a few options:

1. Increase the capacity of the volume
2. Move virtual machines off of the volume
3. Add a new volume

Your solution may be one of these options or a mix of all three. Let’s quickly walk through the options.

Option 1: Increase the capacity of the volume

This is the simplest option. If capacity has crossed the threshold you have specified, increase the volume capacity to clear the threshold.

The process is:

1. Increase the FlashArray volume capacity.

2. Rescan the hosts that use the datastore.

3. Increase the VMFS to use the new capacity.

4. Choose “Use ‘Free space xxx GB/TB’ to expand the datastore” .

5. This will clear the alarm and add additional capacity.

Option 2: Move virtual machine off of the volume

Another option is to move one or more virtual machines from a more-full datastore to a less-full datastore. While this can be manually achieved through case-by-case Storage vMotion, Pure Storage recommends leveraging Storage DRS to automate this. Storage DRS provides, in addition to the performance-based moves discussed earlier in this document, the ability to automatically Storage vMotion VMs based on capacity usage of VMFS datastores. If a datastore reaches a certain percent full, SDRS can automatically move, or make recommendations for, virtual machines to be moved to balance out space usage across volumes.

1. SDRS is enabled on a datastore cluster.

2. When a datastore cluster is created you can enable SDRS and choose capacity threshold settings, which can either be a percentage or a capacity amount.

Pure Storage has no specific recommendations for these values and can be decided upon based on your own environment. Pure Storage does have a few recommendations for datastore cluster configuration in general:

• Only include datastores on the same FlashArray in a given datastore cluster. This will allow Storage vMotion to use the VAAI XCOPY offload to accelerate the migration process of virtual machines and greatly reduce the footprint of the migration workload.
• Include datastores with similar configurations in a datastore cluster. For example, if a datastore is replicated on the FlashArray, only include datastores that are replicated in the same FlashArray protection group so that a SDRS migration does not violate required protection for a virtual machine.

Option 3: Create a new VMFS volume

The last option is to create an entirely new VMFS volume. You might decide to do this for a few reasons:

• The current VMFS volumes have maxed out possible capacity (64 TB each).
• The current VMFS volumes have overloaded the queue depth inside of every ESXi server using it. Therefore, they can be grown in capacity, but cannot provide any more performance due to ESXi limits.

In this situation follow the standard VMFS provisioning steps for a new datastore. Once the creation of volumes and hosts/host groups and the volume connection is complete, the volumes will be accessible to the ESXi host(s). Using the vSphere Web Client, initiate a “Rescan Storage…” to make the newly-connected Pure Storage volume(s) fully-visible to the ESXi servers in the cluster. One can then use the “Add Storage” wizard to format the newly added volume.

Shrinking a Volume

For additional details on resignaturing and snapshot management, please refer to the following blog posts:

• Mounting an unresolved VMFS
• Why not force mount?
• Why might a VMFS resignature operation fail?
• How to correlate a VMFS and a FlashArray volume
• How to snapshot a VMFS on the FlashArray
• Restoring a single VM from a FlashArray snapshot

Mounting a Snapshot Volume

The Pure Storage FlashArray provides the ability to take local or remote point-in-time snapshots of volumes which can then be used for backup/restore and/or test/dev. When a snapshot is taken of a volume containing VMFS, there are a few additional steps from both the FlashArray and vSphere sides to be able to access the snapshot point-in-time data.

When a FlashArray snapshot is taken, a new volume is not created—essentially it is a metadata point-in-time reference to data blocks on the array that reflect that moment’s version of the data. This snapshot is immutable and cannot be directly mounted. Instead, the metadata of a snapshot has to be “copied” to an actual volume which then allows the point-in-time, which was preserved by the snapshot metadata, to be presented to a host. This behavior allows the snapshot to be re-used again and again without changing the data in that snapshot. If a snapshot is not needed more than one time an alternative option is to create a direct snap copy from one volume to another—merging the snapshot creation step with the association step.

When a volume hosting a VMFS datastore is copied via array-based snapshots, the copied VMFS datastore is now on a volume that has a different serial number than the original source volume. Therefore, the VMFS will be reported as having an invalid signature since the VMFS datastore signature is a hash partially based on the serial of the hosting device. Consequently, the device will not be automatically mounted upon rescan—instead the new datastore wizard needs to be run to find the device and resignature the VMFS datastore. Pure Storage recommends resignaturing copied volumes rather than mounting them with an existing signatures (referred to as force mounting).

Deleting a Datastore

Prior to the deletion of a volume, ensure that all important data has been moved off or is no longer needed. From the vSphere Web Client (or CLI) delete or unmount the VMFS volume and then detach the underlying device from the appropriate host(s).

After a volume has been detached from the ESXi host(s) it must first be disconnected (from the FlashArray perspective) from the host within the Purity GUI before it can be destroyed (deleted) on the FlashArray.

1. Unmount the VMFS datastore on every host that it is mounted to.

2. Detach the volume that hosted the datastore from every ESXi host that sees the volume.

3. Disconnect the volume from the hosts or host groups on the FlashArray.

4. Destroy the volume on FlashArray.

Host and Host Group Creation

This section describes the recommendations for creating provisioning objects (called hosts and host groups) on the FlashArray. The purpose is to outline the proper configuration for general understanding.

The FlashArray has two object types for volume provisioning, hosts and host groups:

• Host —a host is a collection of initiators (Fibre Channel WWPNs, iSCSI IQNs or NVMe NQNs) that refers to a physical host. A FlashArray host object must have a one to one relationship with an ESXi host. Every active initiator for a given ESXi host should be added to the respective FlashArray host object. If an initiator is not yet zoned (for instance), and not intended to be, it can be omitted from the FlashArray host object. Furthermore, while the FlashArray supports multiple protocols for a single host (a mixture of FC, iSCSI & NVMe), ESXi does not support presenting VMFS storage via more than one protocol. So creating a multi-protocol host object should be avoided on the FlashArray when in use with VMware ESXi.
  
  In the example below, the ESXi host has two online Fibre Channel HBAs with WWPNs of 21:00:00:0e:1e:1e:7b:e0 and 21:00:00:0e:1e:1e:7b:e1

• Host Group —a host group is a collection of host objects. Pure Storage recommends grouping your ESXi hosts into clusters within vCenter—as this provides a variety of benefits like High Availability and Dynamic Resource Scheduling. In order to provide simple provisioning, Pure Storage also recommends creating host groups that correspond to VMware clusters. Therefore, with every VMware cluster that will use FlashArray storage, a respective host group should be created. Every ESXi host that is in the cluster should have a corresponding host (as described above) that is added to a host group. The host group and its respective cluster should have the same number of hosts. It is recommended to not have more or less hosts in the host group as is in the cluster. While it is supported to have an unmatching count, it makes cluster-based provisioning simpler, and a variety of orchestration integrations require these to match. So it is highly recommended to do so.
  

Setting the FlashArray “ESXi” Host Personality

In Purity 5.1 and later, there is a new host personality type for VMware ESXi hosts. Changing a host personality on a host object on the FlashArray causes the array to change some of its behavior for specific host types.

In general, we endeavor inside of Purity to automatically behave in the correct way without specific configuration changes. Due to a variety of host types supported and varying requirements (a good example is SCSI interaction for features like ActiveCluster & ActiveDR) a manual configuration was required.

In Purity 5.1, it is recommended to enable the “ESXi” host personality for all host objects that represent ESXi hosts.

The ESXi personality does the following things as of Purity 5.1.0:

• Makes the FlashArray issue a Permanent Device Loss SCSI sense response to ESXi when a pod goes offline due to a mediator loss. If this is not set, no response is sent and vSphere HA does not detect the failure properly and will not restart VMs running on the failed hosts.
  
• ESXi uses peripheral LUN IDs instead of flat LUN IDs—this changes how ESXi views any LUN ID on the FlashArray above 255. Since ESXi does not properly interpret flat LUN IDs, it sees LUN ID higher than 255 to be 16,383 higher than it should be (256 is seen as 16,639) which is outside of the supported range of ESXi. Setting the ESXi personality on the FlashArray for a given host switches the FlashArray LUN methodology to peripheral, allowing ESXi to see LUN IDs higher than 255.

While this personality change is currently only relevant for specific ActiveCluster environments and/or environments that want to use higher-than-255 LUN IDs, it is still recommended to set this on all ESXi host objects. Moving forward other behavior changes for ESXi might be included and doing it now ensures it is not missed when it might be important for your environment.

The ESXi host personality can be set through the FlashArray GUI, the CLI or REST. To set it via the GUI, click on Storage, then Hosts, then the host you would like to configure:

Next, go to the Details pane and click the vertical ellipsis and choose Set Personality…:

Choose the radio button corresponding to ESXi and then click Save.

Connecting Volumes to Hosts

A FlashArray volume can be connected to either host objects or host groups. If a volume is intended to be shared by the entire cluster, it is recommended to connect the volume to the host group, not the individual hosts. The makes provisioning easier and helps ensure the entire ESXi cluster has access to the volume. Generally, volumes that are intended to host virtual machines, should be connected at the host group level.

Private volumes, like ESXi boot volumes, should not be connected to the host group as they should not be shared. These volumes should be connected to the host object instead.

Pure Storage has no requirement on LUN IDs for VMware ESXi environments, and users should, therefore, rely on the automatic LUN ID selection built into Purity.

VMware Native Multipathing Plugin (NMP) Configuration

VMware offers a Native Multipathing Plugin (NMP) layer in vSphere through Storage Array Type Plugins (SATP) and Path Selection Policies (PSP) as part of the VMware APIs for Pluggable Storage Architecture (PSA). The SATP has all the knowledge of the storage array to aggregate I/Os across multiple channels and has the intelligence to send failover commands when a path has failed. The Path Selection Policy can be either “Fixed”, “Most Recently Used” or “Round Robin”.

Name                 Device  Vendor    Model             Driver  Transport  Options                     Rule Group  Claim Options                        Default PSP  PSP Options     Description
-------------------  ------  --------  ----------------  ------  ---------  --------------------------  ----------  -----------------------------------  -----------  --------------  --------------------------------------------------------------------------
VMW_SATP_ALUA                PURE      FlashArray                                                       system                                           VMW_PSP_RR   iops=1

esxcli storage nmp satp rule add -s "VMW_SATP_ALUA" -V "PURE" -M "FlashArray" -P "VMW_PSP_RR" -O "iops=1" -e "FlashArray SATP Rule"

Connect-VIServer -Server <vCenter> -Credential (Get-Credential)
Get-VMhost | Get-EsxCli –V2 | % {$_.storage.nmp.satp.rule.add.Invoke(@{description='Pure Storage FlashArray SATP';model='FlashArray';vendor='PURE';satp='VMW_SATP_ALUA';psp='VMW_PSP_RR'; pspoption='iops=1'})}

Name                 Device  Vendor    Model             Driver  Transport  Options                     Rule Group  Claim Options                        Default PSP  PSP Options     Description
VMW_SATP_ALUA                PURE      FlashArray                                                       system                                           VMW_PSP_RR   policy=latency

esxcli storage nmp satp rule add -s "VMW_SATP_ALUA" -V "PURE" -M "FlashArray" -P "VMW_PSP_RR" -O "policy=latency" -e "FlashArray SATP Rule"

Connect-VIServer -Server <vCenter> -Credential (Get-Credential)
Get-VMhost | Get-EsxCli –V2 | % {$_.storage.nmp.satp.rule.add.Invoke(@{description='Pure Storage FlashArray SATP';model='FlashArray';vendor='PURE';satp='VMW_SATP_ALUA';psp='VMW_PSP_RR'; pspoption='policy=latency'})}

esxcli storage nmp psp roundrobin deviceconfig set --type=latency --device=naa.624a93708a75393becad4e43000540e8

esxcli storage nmp device list -d naa.<device NAA>

purehost monitor --balance --interval <how long to sample> --repeat <how many iterations>

Round Robin Path Selection Policy

To best leverage the active-active nature of the front end of the FlashArray, Pure Storage requires that you configure FlashArray volumes to use the Round Robin Path Selection Policy. The Round Robin PSP rotates between all discovered paths for a given volume which allows ESXi (and therefore the virtual machines running on the volume) to maximize the possible performance by using all available resources (HBAs, target ports, etc.).

Name                 Device  Vendor    Model             Driver  Transport  Options                     Rule Group  Claim Options                        Default PSP  PSP Options     Description
-------------------  ------  --------  ----------------  ------  ---------  --------------------------  ----------  -----------------------------------  -----------  --------------  --------------------------------------------------------------------------
VMW_SATP_ALUA                PURE      FlashArray                                                       system                                           VMW_PSP_RR   iops=1

esxcli storage nmp satp rule add -s "VMW_SATP_ALUA" -V "PURE" -M "FlashArray" -P "VMW_PSP_RR" -O "iops=1" -e "FlashArray SATP Rule"

Connect-VIServer -Server <vCenter> -Credential (Get-Credential)
Get-VMhost | Get-EsxCli –V2 | % {$_.storage.nmp.satp.rule.add.Invoke(@{description='Pure Storage FlashArray SATP';model='FlashArray';vendor='PURE';satp='VMW_SATP_ALUA';psp='VMW_PSP_RR'; pspoption='iops=1'})}

The I/O Operations Limit

The Round Robin Path Selection Policy allows for additional tuning of its path-switching behavior in the form of a setting called the I/O Operations Limit. The I/O Operations Limit (sometimes called the “IOPS” value) dictates how often ESXi switches logical paths for a given device. By default, when Round Robin is enabled on a device, ESXi will switch to a new logical path every 1,000 I/Os. In other words, ESXi will choose a logical path, and start issuing all I/Os for that device down that path. Once it has issued 1,000 I/Os for that device, down that path, it will switch to a new logical path and so on.

Pure Storage recommends tuning this value down to the minimum of 1. This will cause ESXi to change logical paths after every single I/O, instead of 1,000.

This recommendation is made for a few reasons:

1. Performance. Often the reason cited to change this value is performance. While this is true in certain cases, the performance impact of changing this value is not usually profound (generally in the single digits of a percentage performance increase). While changing this value from 1,000 to 1 can improve performance, it generally will not solve a major performance problem. Regardless, changing this value can improve performance in some use cases, especially with iSCSI.
2. Path Failover Time. It has been noted in testing that ESXi will fail logical paths much more quickly when this value is set to a the minimum of 1. During a physical failure of the storage environment (loss of a HBA, switch, cable, port, controller) ESXi, after a certain period of time, will fail any logical path that relies on that failed physical hardware and will discontinue attempting to use it for a given volume. This failure does not always happen immediately. When the I/O Operations Limit is set to the default of 1,000 path failover time can sometimes be in the 10s of seconds which can lead to noticeable disruption in performance during this failure. When this value is set to the minimum of 1, path failover generally decreases to sub-ten seconds. This greatly reduces the impact of a physical failure in the storage environment and provides greater performance resiliency and reliability.
3. FlashArray Controller I/O Balance. When Purity is upgraded on a FlashArray, the following process is observed (at a high level): upgrade Purity on one controller, reboot it, wait for it to come back up, upgrade Purity on the other controller, reboot it and you’re done. Due to the reboots, twice during the process half of the FlashArray front-end ports go away. Because of this, we want to ensure that all hosts are actively using both controllers prior to upgrade. One method that is used to confirm this is to check the I/O balance from each host across both controllers. When volumes are configured to use Most Recently Used, an imbalance of 100% is usually observed (ESXi tends to select paths that lead to the same front end port for all devices). This then means additional troubleshooting to make sure that host can survive a controller reboot. When Round Robin is enabled with the default I/O Operations Limit, port imbalance is improved to about 20-30% difference. When the I/O Operations Limit is set to 1, this imbalance is less than 1%. This gives Pure Storage and the end user confidence that all hosts are properly using all available front-end ports.

For these three above reasons, Pure Storage highly recommends altering the I/O Operations Limit to 1. For additional information you can read the VMware KB regarding setting the IOPs Limit.

ESXi Express Patch 5 or 6.5 Update 1 and later

Starting with ESXi 6.0 Express Patch 5 (build 5572656) and later (Release notes) and ESXi 6.5 Update 1 (build 5969303) and later (release notes), Round Robin and an I/O Operations limit is the default configuration for all Pure Storage FlashArray devices (iSCSI and Fibre Channel) and no configuration is required.

A new default SATP rule, provided by VMware by default was specifically built for the FlashArray to Pure Storage’s best practices. Inside of ESXi you will see a new system rule:

Name                 Device  Vendor    Model             Driver  Transport  Options                     Rule Group  Claim Options                        Default PSP  PSP Options     Description
-------------------  ------  --------  ----------------  ------  ---------  --------------------------  ----------  -----------------------------------  -----------  --------------  --------------------------------------------------------------------------
VMW_SATP_ALUA                PURE      FlashArray                                                       system                                           VMW_PSP_RR   iops=1

For information, refer to this blog post:

https://www.codyhosterman.com/2017/0...e-now-default/

Configuring Round Robin and the I/O Operations Limit

If you are running earlier than ESXi 6.0 Express Patch 5 or 6.5 Update 1, there are a variety of ways to configure Round Robin and the I/O Operations Limit. This can be set on a per-device basis and as every new volume is added, these options can be set against that volume. This is not a particularly good option as one must do this for every new volume, which can make it easy to forget, and must do it on every host for every volume. This makes the chance of exposure to mistakes quite large.

The recommended option for configuring Round Robin and the correct I/O Operations Limit is to create a rule that will cause any new FlashArray device that is added in the future to that host to automatically get the Round Robin PSP and an I/O Operation Limit value of 1.

The following command creates a rule that achieves both of these for only Pure Storage FlashArray devices:

esxcli storage nmp satp rule add -s "VMW_SATP_ALUA" -V "PURE" -M "FlashArray" -P "VMW_PSP_RR" -O "iops=1" -e "FlashArray SATP Rule"

This must be repeated for each ESXi host.

This can also be accomplished through PowerCLI. Once connected to a vCenter Server this script will iterate through all of the hosts in that particular vCenter and create a default rule to set Round Robin for all Pure Storage FlashArray devices with an I/O Operation Limit set to 1.

Connect-VIServer -Server <vCenter> -Credential (Get-Credential)
Get-VMhost | Get-EsxCli –V2 | % {$_.storage.nmp.satp.rule.add.Invoke(@{description='Pure Storage FlashArray SATP';model='FlashArray';vendor='PURE';satp='VMW_SATP_ALUA';psp='VMW_PSP_RR'; pspoption='iops=1'})}

Furthermore, this can be configured using vSphere Host Profiles:

It is important to note that existing, previously presented devices will need to be manually set to Round Robin and an I/O Operation Limit of 1. Optionally, the ESXi host can be rebooted so that it can inherit the multipathing configuration set forth by the new rule.

For setting a new I/O Operation Limit on an existing device, see Appendix I: Per-Device NMP Configuration.

Enhanced Round Robin Load Balancing (Latency Based PSP)

With the release of vSphere 6.7 U1, there is now a sub-policy option for Round Robin that actively monitors individual path performance. This new sub-policy is called "Enhanced Round Robin Load Balancing" (also known as Latency Based Path Selection Policy (PSP)). Before this policy became available the ESXi host would utilize all active paths by sending I/O requests down each path in a "fire and forget" type of fashion, sending 1 I/O down each path before moving to the next. Often times this resulted in performance penalties when individual paths became degraded and weren't functioning as optimally as other available paths. This performance penalty was invoked because the ESXi host would continue using the non-optimal path due to limited insight into the overall path health. This now changes with the Latency Based PSP by monitoring each path for latency, along with outstanding I/Os, allowing the  ESXi host to make smarter decisions on which paths to use and which to exclude in a more dynamic manner.

Name                 Device  Vendor    Model             Driver  Transport  Options                     Rule Group  Claim Options                        Default PSP  PSP Options     Description
VMW_SATP_ALUA                PURE      FlashArray                                                       system                                           VMW_PSP_RR   policy=latency

esxcli storage nmp satp rule add -s "VMW_SATP_ALUA" -V "PURE" -M "FlashArray" -P "VMW_PSP_RR" -O "policy=latency" -e "FlashArray SATP Rule"

Connect-VIServer -Server <vCenter> -Credential (Get-Credential)
Get-VMhost | Get-EsxCli –V2 | % {$_.storage.nmp.satp.rule.add.Invoke(@{description='Pure Storage FlashArray SATP';model='FlashArray';vendor='PURE';satp='VMW_SATP_ALUA';psp='VMW_PSP_RR'; pspoption='policy=latency'})}

esxcli storage nmp psp roundrobin deviceconfig set --type=latency --device=naa.624a93708a75393becad4e43000540e8

How it Works

Like all other Native Multipathing Plugin (NMP) policies this sub-policy is set on a per LUN or per datastore basis. Once enabled the NMP begins by assessing the first 16 user I/O requests per path and calculates their average latency. Once all of the paths have been successfully analyzed the NMP will then calculate the average latency of each path and use this information to determine which paths are healthy (optimal) and which are unhealthy (non-optimal). If a path falls outside of the average latency it is deemed non-optimal and will not be used until latency has reached an optimal response time once more.

After the initial assessment, the ESXi host then repeats the same process outlined above every 3 minutes. It will test every active path, including any non-optimal paths, to confirm if the latency has improved, worsened, or remained the same.  Once again those results will be analyzed and used to determine which paths should continue sending I/O requests and which should be paused to see if they report better health in the next 3 minutes. Throughout this process the NMP is also taking into account any outstanding I/Os for each path to make more informed decisions.

Configuring Round Robin and the Latency Based Sub-Policy

If you are using ESXi 7.0 or later then no changes are required to enable this new sub-policy as it is the new recommendation moving forward. In an effort to make things easier for end-users a new SATP rule has been added that will automatically apply this rule to any Pure Storage LUNs presented to the ESXi host:

Name                 Device  Vendor    Model             Driver  Transport  Options                     Rule Group  Claim Options                        Default PSP  PSP Options     Description
VMW_SATP_ALUA                PURE      FlashArray                                                       system                                           VMW_PSP_RR   policy=latency

If your environment is using ESXi 6.7U1 or later and you wish to utilize this feature, which Pure Storage supports, then the best way is to create a SATP rule on each ESXi host, which can be done as follows:

esxcli storage nmp satp rule add -s "VMW_SATP_ALUA" -V "PURE" -M "FlashArray" -P "VMW_PSP_RR" -O "policy=latency" -e "FlashArray SATP Rule"

Alternatively, this can be done using PowerShell:

Connect-VIServer -Server <vCenter> -Credential (Get-Credential)
Get-VMhost | Get-EsxCli –V2 | % {$_.storage.nmp.satp.rule.add.Invoke(@{description='Pure Storage FlashArray SATP';model='FlashArray';vendor='PURE';satp='VMW_SATP_ALUA';psp='VMW_PSP_RR'; pspoption='policy=latency'})}

Lastly, if you would like to change an individual LUN  (or set of LUNs) you can run the following command to change the PSP to latency (where device is specific to your env):

esxcli storage nmp psp roundrobin deviceconfig set --type=latency --device=naa.624a93708a75393becad4e43000540e8

Tuning

By default the RR latency policy is configured to send 16 user I/O requests down each path and evaluate each path every three minutes (180000ms). Based on extensive testing, Pure Storage's recommendation is to leave these options configured to their defaults and no changes are required.

Verifying Connectivity

It is important to verify proper connectivity prior to implementing production workloads on a host or volume.

This consists of a few steps:

1. Verifying proper multipathing settings in ESXi.
2. Verifying the proper numbers of paths.
3. Verifying I/O balance and redundancy on the FlashArray.

The Path Selection Policy and number of paths can be verified easily inside of the vSphere Web Client.

This will report the path selection policy and the number of logical paths. The number of logical paths will depend on the number of HBAs, zoning and the number of ports cabled on the FlashArray.

The I/O Operations Limit cannot be checked from the vSphere Web Client—it can only be verified or altered via command line utilities. The following command can check a particular device for the PSP and I/O Operations Limit:

esxcli storage nmp device list -d naa.<device NAA>

Please remember that each of these settings is a per-host setting, so while a volume might be configured properly on one host, it may not be correct on another.

Additionally, it is also possible to check multipathing from the FlashArray.

A CLI command exists to monitor I/O balance coming into the array:

purehost monitor --balance --interval <how long to sample> --repeat <how many iterations>

The command will report a few things:

1. The host name.
2. The individual initiators from the host. If they are logged into more than one FlashArray port, it will be reported more than once. If an initiator is not logged in at all, it will not appear.
3. The port that the initiator is logged into.
4. The number of I/Os that came into that port from that initiator over the time period sampled.
5. The relative percentage of I/Os for that initiator as compared to the maximum.

The balance command will count the I/Os that came down from a particular initiator during the sampled time period, and it will do that for all initiator/target relationships for that host. Whichever relationship/path has the most I/Os will be designated as 100%. The rest of the paths will be then denoted as a percentage of that number. So if a host has two paths, and the first path has 1,000 I/Os and the second path has 800, the first path will be 100% and the second will be 80%.

A well balanced host should be within a few percentage points of each path. Anything more than 15% or so might be worthy of investigation. Refer to this post for more information.

The GUI will also report on host connectivity in general, based on initiator logins.

This report should be listed as redundant for all hosts, meaning that it is connected to each controller. If this reports something else, investigate zoning and/or host configuration to correct this.

For a detailed explanation of the various reported states, please refer to the FlashArray User Guide which can be found directly in your GUI:

Disk.DiskMaxIOSize

The ESXi host setting, Disk.DiskMaxIOSize, controls the largest I/O size that ESXi will allow to be sent from ESXi to an underlying storage device. By default this is 32 MB. If an I/O is larger than the Disk.DiskMaxIOSize value, ESXi will split the I/O requests into segments under the configured limit.

If you are running an older release of ESXi (versions listed below) this setting needs to be modified if and only if you are on an old version and have an environment running the following scenarios:

1. If a virtual machine is using EFI (Extensible Firmware Interface) instead of BIOS and is using VMware Hardware Version 12 or earlier.
2. If your environment utilizes vSphere Replication.
3. If your environment contains VMs which house applications that are sending READ or WRITE requests larger than 4 MB.
4. The environment is using Fibre Channel with one of the above scenarios (this issue is not present with iSCSI).

VMware has resolved this issue in two places--fixing it in ESXi itself (ESXi now reads the maximum supported SCSI from the array and will only send I/Os of that size or smaller and split anything larger) and within VMware HW.

This is resolved in the following ESXi releases:

• ESXi 6.0, Patch Release ESXi600-201909001
• ESXi 6.5, Patch Release ESXi650-201811002
• ESXi 6.7 Update 1 Release
• ESXi 7.0 all releases

If you are not running one of these newer releases, it is necessary to reduce the ESXi parameter Disk.DiskMaxIOSize from the default of 32 MB (32,768 KB) down to 4 MB (4,096 KB) or less.

If this is not configured for ESXi hosts running EFI-enabled VMs, the virtual machine will fail to properly boot. If it is not changed on hosts running VMs being replicated by vSphere Replication, replication will fail. If it is not changed for VMs whose applications are sending requests larger than 4MB, the larger I/O requests will fail which results in the application failing as well.

This should be set on every ESXi host in the cluster that VMs may have access to, in order to ensure vMotion is successful from one ESXi host to another. If none of the above circumstances apply to your environment then this value can remain at the default. There is no known performance impact by changing this value.

For more detail on this change, please refer to the VMware KB article here:

https://kb.vmware.com/s/article/2137402

VAAI Configuration

The VMware API for Array Integration (VAAI) primitives offer a way to offload and accelerate certain operations in a VMware environment.

Pure Storage requires that all VAAI features be enabled on every ESXi host that is using FlashArray storage. Disabling VAAI features can greatly reduce the efficiency and performance of FlashArray storage in ESXi environments.

All VAAI features are enabled by default (set to 1) in ESXi 5.x and later, so no action is typically required. Though these settings can be verified via the vSphere Web Client or CLI tools.

1. WRITE SAME—DataMover.HardwareAcceleratedInit
2. XCOPY—DataMover.HardwareAcceleratedMove
3. ATOMIC TEST & SET— VMFSHardwareAcceleratedLocking

In order to provide a more efficient heart-beating mechanism for datastores VMware introduced a new host-wide setting called /VMFS3/UseATSForHBOnVMFS5. In VMware’s own words:

“A change in the VMFS heartbeat update method was introduced in ESXi 5.5 Update 2, to help optimize the VMFS heartbeat process. Whereas the legacy method involves plain SCSI reads and writes with the VMware ESXi kernel handling validation, the new method offloads the validation step to the storage system.“

Pure Storage recommends keeping this value on whenever possible. That being said, it is a host wide setting, and it can possibly affect storage arrays from other vendors negatively.

Read the VMware KB article here:

ESXi host loses connectivity to a VMFS3 and VMFS5 datastore

Pure Storage is NOT susceptible to this issue, but in the case of the presence of an affected array from another vendor, it might be necessary to turn this off. In this case, Pure Storage supports disabling this value and reverting to traditional heart-beating mechanisms.

For additional information please refer to VMware Storage APIs for Array Integration with the Pure Storage FlashArray document.

iSCSI Configuration

Just like any other array that supports iSCSI, Pure Storage recommends the following changes to an iSCSI-based vSphere environment for the best performance.

For a detailed walkthrough of setting up iSCSI on VMware ESXi and on the FlashArray please refer to the following VMware white paper. This is required reading for any VMware/iSCSI user:

https://core.vmware.com/resource/best-practices-running-vmware-vsphere-iscsi

vmkping -d -s 8972 <ip address of Pure Storage iSCSI port>

Name                  Current     Default     Min  Max       Settable  Inherit
--------------------  ----------  ----------  ---  --------  --------  -------
NoopOutInterval       15          15          1    60            true    false
NoopOutTimeout        10          10          10   30            true     true
RecoveryTimeout       10          10          1    120           true     true

Set Login Timeout to a Larger Value

For example, to set the Login Timeout value to 30 seconds, use commands similar to the following:

1. Log in to the vSphere Web Client and select the host under Hosts and Clusters.
2. Navigate to the Manage tab.
3. Select the Storage option.
4. Under Storage Adapters, select the iSCSI vmhba to be modified.
5. Select Advanced and change the Login Timeout parameter. This can be done on the iSCSI adapter itself or on a specific target.

The default Login Timeout value is 5 seconds and the maximum value is 60 seconds.

Disable DelayedAck

DelayedAck is an advanced iSCSI option that allows or disallows an iSCSI initiator to delay acknowledgment of received data packets.

Disabling DelayedAck:

1. Log in to the vSphere Web Client and select the host under Hosts and Clusters.
2. Navigate to the Configure tab.
3. Select the Storage option.
4. Under Storage Adapters, select the iSCSI vmhba to be modified.

Navigate to Advanced Options and modify the DelayedAck setting by using the option that best matches your requirements, as follows:

Option 1: Modify the DelayedAck setting on a particular discovery address (recommended) as follows:

1. Select Targets.
2. On a discovery address, select the Dynamic Discovery tab.
3. Select the iSCSI server.
4. Click Advanced.
5. Change DelayedAck to false.

Option 2: Modify the DelayedAck setting on a specific target as follows:

1. Select Targets.
2. Select the Static Discovery tab.
3. Select the iSCSI server and click Advanced.
4. Change DelayedAck to false.

Option 3: Modify the DelayedAck setting globally for the iSCSI adapter as follows:

1. Select the Advanced Options tab and click Advanced.
2. Change DelayedAck to false.

DelayedAck is highly recommended to be disabled, but is not absolutely required by Pure Storage. In highly-congested networks, if packets are lost, or simply take too long to be acknowledged, due to that congestion, performance can drop. If DelayedAck is enabled, where not every packet is acknowledged at once (instead one acknowledgment is sent per so many packets) far more re-transmission can occur, further exacerbating congestion. This can lead to continually decreasing performance until congestion clears. Since DelayedAck can contribute to this it is recommended to disable it in order to greatly reduce the effect of congested networks and packet retransmission.

Enabling jumbo frames can further harm this since packets that are retransmitted are far larger. If jumbo frames are enabled, it is absolutely recommended to disable DelayedAck.

See the following VMware KB for more information:

http://kb.vmware.com/selfservice/microsites/search.do?language=en_US&cmd=displayKC&externalId=1002598

iSCSI Port Binding

For software iSCSI initiators, without additional configuration the default behavior for iSCSI pathing is for ESXi to leverage its routing tables to identify a path to its configured iSCSI targets. Without solid understanding of network configuration and routing behaviors, this can lead to unpredictable pathing and/or path unavailability in a hardware failure. To configure predictable and reliable path selection and failover it is necessary to configure iSCSI port binding (iSCSI multipathing).

Configuration and detailed discussion are out of the scope of this document, but it is recommended to read through the following VMware document that describes this and other concepts in-depth:

http://www.vmware.com/files/pdf/techpaper/vmware-multipathing-configuration-software-iSCSI-port-binding.pdf

Note that ESXi 6.5 has expanded support for port binding and features such as iSCSI routing (though the use of iSCSI routing is not usually recommended) and multiple subnets. Refer to ESXi 6.5 release notes for more information.

Jumbo Frames

In some iSCSI environments it is required to enable jumbo frames to adhere to the network configuration between the host and the FlashArray. Enabling jumbo frames is a cross-environment change so careful coordination is required to ensure proper configuration. It is important to work with your networking team and Pure Storage representatives when enabling jumbo frames. Please note that this is not a requirement for iSCSI use on the Pure Storage FlashArray—in general, Pure Storage recommends leaving MTU at the default setting.

That being said, altering the MTU is a fully supported and is up to the discretion of the user.

1. Configure jumbo frames on the FlashArray iSCSI ports.

Configure jumbo frames on the physical network switch/infrastructure for each port using the relevant switch CLI or GU I.

2. Configure jumbo frames on the physical network switch/infrastructure for each port using the relevant switch CLI or GUI.
   1. Browse to a host in the vSphere Web Client navigator.
   2. Click the Configure tab and select Networking > Virtual Switches.
   3. Select the switch from the vSwitch list.
   4. Click the name of the VMkernel network adapter.
   5. Click the pencil icon to edit.
   6. Click NIC settings and set the MTU to your desired value.
   7. Click OK.
   8. Click the pencil icon to edit on the top to edit the vSwitch itself.
   9. Set the MTU to your desired value.
   10. Click OK.

Once jumbo frames are configured, verify end-to-end jumbo frame compatibility. To verify, try to ping an address on the storage network with vmkping.

vmkping -d -s 8972 <ip address of Pure Storage iSCSI port>

If the ping operations does not return successfully, then jumbo frames is not properly configured in ESXi, the networking devices, and/or the FlashArray port.

Challenge-Handshake Authentication Protocol (CHAP)

iSCSI CHAP is supported on the FlashArray for unidirectional or bidirectional authentication. Enabling CHAP is optional and up to the discretion of the user. Please refer to the following post for a detailed walkthrough:

http://www.codyhosterman.com/2015/03/configuring-iscsi-chap-in-vmware-with-the-flasharray/

iSCSI Failover Times

A common question encountered here at Pure Storage is why extended pauses in I/O are noted during specific operations or tests when utilizing the iSCSI protocol. Often times the underlying reasons for these pauses in I/O are a result of a network cable being disconnected, a misbehaving switch port, or a failover of the backend storage array; though this list is certainly not exhaustive.

When the default configuration for iSCSI is in use with VMware ESXi the delay for these events will generally be 25-35 seconds. While the majority of environments are able to successfully recover from these events unscathed this is not true for all environments. On a handful of occasions, there have been environments that contain applications that need faster recovery times. Without these faster recovery times, I/O failures have been noted and manual recovery efforts were required to bring the environment back online.

While Pure Storage's official best practice is to utilize default iSCSI configuration for failover times we also understand that not all environments are created equal. As such we do support modifying the necessary iSCSI advanced parameters to decrease failover times for sensitive applications.

Recovery times are controlled by the following 3 iSCSI advanced parameters:

Name                  Current     Default     Min  Max       Settable  Inherit
--------------------  ----------  ----------  ---  --------  --------  -------
NoopOutInterval       15          15          1    60            true    false
NoopOutTimeout        10          10          10   30            true     true
RecoveryTimeout       10          10          1    120           true     true

To better understand how these parameters are used in iSCSI recovery efforts it is recommended you read the following blog posts for deeper insight:

iSCSI: A 25-second pause in I/O during a single link loss? What gives?

iSCSI Advanced Settings

Once a thorough review of these iSCSI options have been investigated, additional testing within your own environment is strongly recommended to ensure no additional issues are introduced as a result of these changes.

Network Time Protocol (NTP)

No matter how perfect an environment is configured there will always come a time where troubleshooting an issue will be required. This is inevitable when dealing with large and complex environments. One way to help alleviate some of the stress that comes with troubleshooting is ensuring that the Network Time Protocol (NTP) is enabled on all components in the environment. NTP will ensure that the timestamps for servers, arrays, switches, etc are all aligned and in-sync. It is for this reason that Pure Storage recommends as a best practice that NTP be enabled and configured on all components.

Please refer to VMware KB Configuring Network Time Protocol (NTP) on an ESXi host using the vSphere Client for steps on how to configure NTP on your ESXi hosts.

Remote Syslog Server

Another helpful tool in the toolbox of troubleshooting is having a remote syslog server configured. There may be times where an investigation is required in the environment but when attempting to review the logs it is discovered that they are no longer available. Often times this is a result of the increased logging that happened during the time of the issue. The increased logging leads to thresholds for file size and counts being exceeded and thus the older logs are automatically deleted as a result.

Pure Storage recommends the use of the VMware vRealize Log Insight OVA. This provides for a quick and easy integration for the ESXi hosts and vCenter. Additionally, the Pure Storage Content Pack can be used with vRealize Log Insight which provides a single logging destination for both the vSphere and Pure Storage environments.

Configuring vCenter Server and ESXi with Log Insight

As explained above, configuring vCenter Server and ESXi is a relatively quick and simple process.

• Login to VMware vRealize Log Insight.
• Click on Administration .
• Under Integration click on vSphere .
• Click + Add vCenter Server .
• Fill in the applicable vCenter Server information and Test Connection.
• Ensure the following boxes are checked:
  • Collect vCenter Server events, tasks, and alarms
  • Configure ESXi hosts to send logs to Log Insight
• Click Save to commit all of the requested changes.

Additional Remote syslog Options

It is understood that not every customer or environment will have vRealize Log Insight installed or available. If your environment takes advantage of a different solution then please refer to the third party documentation on how the best way to integrate it with your vSphere environment. You can also refer to VMware's Knowledge Base article Configuring syslog on ESXi for additional options and configuration information.

Virtual Disk Choice

Storage provisioning in virtual infrastructure involves multiple steps of crucial decisions. VMware vSphere offers three virtual disks formats: thin, zeroedthick and eagerzeroedthick.

To quickly review the types:

1. Thin—thin virtual disks only allocate what is used by the guest. Upon creation, thin virtual disks only consume one block of space. As the guest writes data, new blocks are allocated on VMFS, then zereod out, then the data is committed to storage. Therefore there is some additional latency for new write
2. Zeroedthick (lazy)— zeroed thick virtual disks allocate all of the space on the VMFS upon creation. As soon as the guest writes to a specific block for the first time in the virtual disk, the block is first zeroed, then the data is committed. Therefore there is some additional latency for new writes. Though less than thin (since it only has to zero—not also allocate), there is a negligible performance impact between zeroedthick (lazy) and thin.
3. Eagerzeroedthick—eagerzeroedthick virtual disks allocate all of their provisioned size upon creation and also zero out the entire capacity upon creation. This type of disk cannot be used until the zeroing is complete. Eagerzeroedthick has zero first-write latency penalty because allocation and zeroing is done in advance, and not on-demand.

Prior to WRITE SAME support, the performance differences between these virtual disk allocation mechanisms were distinct. This was due to the fact that before an unallocated block could be written to, zeroes would have to be written first causing an allocate-on-first-write penalty (increased latency). Therefore, for every new block written, there were actually two writes; the zeroes then the actual data. For thin and zeroedthick virtual disks, this zeroing was on-demand so the penalty was seen by applications. For eagerzeroedthick, it was noticed during deployment because the entire virtual disk had to be zeroed prior to use. This zeroing caused unnecessary I/O on the SAN fabric, subtracting available bandwidth from “real” I/O.

To resolve this issue, VMware introduced WRITE SAME support. WRITE SAME is a SCSI command that tells a target device (or array) to write a pattern (in this case, zeros) to a target location. ESXi utilizes this command to avoid having to actually send a payload of zeros but instead simply communicates to any array that it needs to write zeros to a certain location on a certain device. This not only reduces traffic on the SAN fabric, but also speeds up the overall process since the zeros do not have to traverse the data path.

This process is optimized even further on the Pure Storage FlashArray. Since the array does not store space-wasting patterns like contiguous zeros on the array, the zeros are discarded and any subsequent reads will result in the array returning zeros to the host. This additional array-side optimization further reduces the time and penalty caused by pre-zeroing of newly-allocated blocks.

With this knowledge, choosing a virtual disk is a factor of a few different variables that need to be evaluated. In general, Pure Storage makes the following recommendations:

• Lead with thin virtual disks. They offer the greatest flexibility and functionality and the performance difference is only at issue with the most sensitive of applications.
• For highly-sensitive applications with high performance requirements, eagerzeroedthick is the best choice. It is always the best-performing virtual disk type.
• In no situation does Pure Storage recommend the use of zeroedthick (thick provision lazy zeroed) virtual disks. There is very little advantage to this format over the others and can also lead to stranded space as described in this post.

With that being said, for more details on how these recommendations were decided upon, refer to the following considerations. Note that at the end of each consideration is a recommendation but that recommendation is valid only when only that specific consideration is important. When choosing a virtual disk type, take into account your virtual machine business requirements and utilize these requirements to motivate your design decisions. Based on those decisions, choose the virtual disk type that is best suitable for your virtual machine.

• Performance —with the introduction of WRITE SAME (more information on WRITE SAME can be found in the section Block Zero or WRITE SAME) support, the performance difference between the different types of virtual disks is dramatically reduced—almost eliminated. In lab experiments, a difference can be observed during writes to unallocated portions of a thin or zeroedthick virtual disk. This difference is negligible but of course still non-zero. Therefore, performance is no longer an overridingly important factor in the type of virtual disk to use as the disparity is diminished, but for the most latency-sensitive of applications eagerzeroedthick will always be slightly better than the others. Recommendation: eagerzeroedthick.
• Protection against space exhaustion —each virtual disk type, based on its architecture, has varying degrees of protection against space exhaustion. Thin virtual disks do not reserve space on the VMFS datastore upon creation and instead grow in 1 MB blocks as needed. Therefore, if unmonitored, as one or more thin virtual disks grow on the datastore, they could exhaust the capacity of the VMFS. Even if the underlying array has plenty of additional capacity to provide. If careful monitoring is in place that provides the ability to make proactive resolution of capacity exhaustion (moving the virtual machines around or grow the VMFS) thin virtual disks are a perfectly acceptable choice. Storage DRS is an excellent solution for space exhaustion prevention. While careful monitoring can protect against this possibility, it can still be of a concern and should be contemplated upon initial provisioning. Zeroedthick and eagerzeroedthick virtual disks are not susceptible to VMFS logical capacity exhaustion because the space is reserved on the VMFS upon creation. Recommendation: eagerzeroedthick.
• Virtual disk density —it should be noted that while all virtual disk types take up the same amount of physical space on the FlashArray due to data reduction, they have different requirements on the VMFS layer. Thin virtual disks can be oversubscribed (more capacity provisioned than the VMFS reports as being available) allowing for far more virtual disks to fit on a given volume than either of the thick formats. This provides a greater virtual machine to VMFS datastore density and reduces the number or size of volumes that are required to store them. This, in effect, reduces the management overhead of provisioning and managing additional volumes in a VMware environment. Recommendation: thin.
• Time to create —the virtual disk types also vary in how long it takes to initially create them. Since thin and zeroedthick virtual disks do not zero space until they are actually written to by a guest they are both created in trivial amounts of time—usually a second or two. Eagerzeroedthick disks, on the other hand, are pre-zeroed at creation and consequently take additional time to create. If the time-to-first-IO is paramount for whatever reason, thin or zeroedthick is best. Recommendation: thin.
• Space efficiency —the aforementioned bullet on “virtual disk density” describes efficiency on the VMFS layer. Efficiency on the underlying array should also be considered. In vSphere 6.0, thin virtual disks support guest-OS initiated UNMAP to a virtual disk, through the VMFS and down to the physical storage. Therefore, thin virtual disks can be more space efficient as time wears on and data is written and deleted. For more information on this functionality in vSphere 6.0, refer to the section, In-Guest UNMAP in ESXi 6.x, that can be found later in this paper. Recommendation: thin.
• Storage usage trending —A useful metric to know and track is how much capacity is actually being used by a virtual machine guest. If you know how much space is being used by the guests, and furthermore, at what rate that is growing, you can more appropriately size and project storage allocations. Since thick type virtual disks reserve all of the space on the VMFS whether or not the guest has used it, it is difficult to know, without guest tools, how much the guest has actually written. Often it is not known until the application has used its available space and the administrator requests more. This leads to abrupt and unplanned capacity increases. Thin virtual disks only reserve what the guest has written, therefore will grow as the guest adds more data. This growth can be monitored and trended. This will allow VMware administrators to plan and predict future storage needs. Recommendation: thin.

No virtual disk option quite fits all possible use-cases perfectly, so choosing an allocation method should generally be decided upon on a case-by-case basis. VMs that are intended for short term use, without extraordinarily high performance requirements, fit nicely with thin virtual disks. For VMs that have higher performance needs eagerzeroedthick is a good choice.

Virtual Hardware Configuration

Pure Storage makes the following recommendations for configuring a virtual machine in vSphere:

Virtual SCSI Adapter —the best performing and most efficient virtual SCSI adapter is the VMware Paravirtual SCSI Adapter. This adapter has the best CPU efficiency at high workloads and provides the highest queue depths for a virtual machine—starting at an adapter queue depth of 256 and a virtual disk queue depth 64 (twice what the LSI Logic can provide by default). The queue limits of PVSCSI can be further tuned, please refer to the Guest-level Settings section for more information. The virtual NVMe adapter is supported by both Pure and VMware, but at this time there is no significant benefit to its use over PVSCSI. In the future, that will likely change, but as of ESX 7.0 U1 the recommendation (not requirement though) remains PVSCSI.

Virtual Hardware —it is recommended to use the latest virtual hardware version that the hosting ESXi hosts supports.

VMware tools —in general, it is advisable to install the latest supported version of VMware tools in all virtual machines.

CPU and Memory - provision vCPUs and memory as per the application requirements.

VM encryption —vSphere 6.5 introduced virtual machine encryption which encrypts the VM’s virtual disk from a VMFS perspective. Pure Storage generally recommends not using this and instead relying on FlashArray-level Data-At-Rest-Encryption. Though, if it is necessary to leverage VM Encryption, doing so is fully supported by Pure Storage—but it should be noted that data reduction will disappear for that virtual machine as host level encryption renders post-encryption deduplication and compression impossible.

IOPS Limits —if you want to limit a virtual machine or a particular amount of IOPS, you can use the built-in ESXi IOPS limits. ESXi allows you to specify a number of IOPS a given virtual machine can issue for a given virtual disk. Once the virtual machine exceeds that number, any additional I/Os will be queued. In ESXi 6.0 and earlier this can be applied via the “Edit Settings” option of a virtual machine.

In ESXi 6.5 and later, this can also be configured via a VM Storage Policy:

Template Configuration

In general, template configuration is no different than virtual machine configuration. Standard recommendations apply. That being said, since templates are by definition frequently copied, Pure Storage recommends putting copies of the templates on FlashArrays that are frequent targets of virtual machines deployed from a template. If the template and target datastore are on the same FlashArray, the copy process can take advantage of VAAI XCOPY, which greatly accelerates the copy process while reducing the workload impact of the copy operation.

Prior to Full Copy (XCOPY) API support, when virtual machines needed to be copied or moved from one location to another, such as with Storage vMotion or a virtual machine cloning operation, ESXi would issue many SCSI read/write commands between the source and target storage location (the same or different device). This resulted in a very intense and often lengthy additional workload to this set of devices. This SCSI I/O consequently stole available bandwidth from more “important” I/O such as the I/O issued from virtualized applications. Therefore, copy or move operations often had to be scheduled to occur only during non-peak hours in order to limit interference with normal production storage performance.  This restriction effectively decreased the ability of administrators to use the virtualized infrastructure in the dynamic and flexible nature that was intended.

The introduction of XCOPY support for virtual machine movement allows for this workload to be offloaded from the virtualization stack to almost entirely onto the storage array. The ESXi kernel is no longer directly in the data copy path and the storage array instead does all the work. XCOPY functions by having the ESXi host identify a region of a VMFS that needs to be copied. ESXi describes this space into a series of XCOPY SCSI commands and sends them to the array. The array then translates these block descriptors and copies/moves the data from the described source locations to the described target locations. This architecture therefore does not require the moved data to be sent back and forth between the host and array—the SAN fabric does not play a role in traversing the data. This vastly reduces the time to move data. XCOPY benefits are leveraged during the following operations[1]:

• Virtual machine cloning
• Storage vMotion
• Deploying virtual machines from template

During these offloaded operations, the throughput required on the data path is greatly reduced as well as the ESXi hardware resources (HBAs, CPUs etc.) initiating the request. This frees up resources for more important virtual machine operations by letting the ESXi resources do what they do best: run virtual machines, and lets the storage do what it does best: manage the storage.

On the Pure Storage FlashArray, XCOPY sessions are exceptionally quick and efficient. Due to FlashReduce technology (features like deduplication, pattern removal and compression) similar data is never stored on the FlashArray more than once. Therefore, during a host-initiated copy operation such as with XCOPY, the FlashArray does not need to copy the data—this would be wasteful. Instead, Purity simply accepts and acknowledges the XCOPY requests and creates new (or in the case of Storage vMotion, redirects existing) metadata pointers. By not actually having to copy/move data, the offload duration is greatly reduced. In effect, the XCOPY process is a 100% inline deduplicated operation. A non-VAAI copy process for a virtual machine containing 50 GB of data can take on the order of multiple minutes or more depending on the workload on the SAN. When XCOPY is enabled this time drops to a matter of a few seconds.

XCOPY on the Pure Storage FlashArray works directly out of the box without any configuration required. Nevertheless, there is one simple configuration change on the ESXi hosts that will increase the speed of XCOPY operations. ESXi offers an advanced setting called the MaxHWTransferSize that controls the maximum amount of data space that a single XCOPY SCSI command can describe. The default value for this setting is 4 MB. This means that any given XCOPY SCSI command sent from that ESXi host cannot exceed 4 MB of described data. Pure Storage recommends leaving this at the default value, but does support increasing the value if another vendor requires it to be. There is no XCOPY performance impact of increasing this value. Decreasing the value from 4 MB can slow down XCOPY sessions somewhat and should not be done without guidance from VMware or Pure Storage support. For this reason, Pure Storage recommends setting the transfer size to the maximum value of 16 MB.

If an application does need to push a high amount of IOPS to a single virtual disk these limits must be increased. See VMware KB here for information on how to configure Paravirtual SCSI adapter queue limits. The process slightly differs between Linux and Windows.

Refer to this blog post for more information:

http://www.codyhosterman.com/2017/02/understanding-vmware-esxi-queuing-and-the-flasharray/

A few general recommendations:

1. Only increase these limits when needed
2. If you change this limit it is required to change queue depth limits in ESXi as well, otherwise changing these values will have no tangible affect
3. A good rule of knowing if you need to change these values is if you are not getting the IOPS you expect and the latency is high in the guest, but not reported as high in ESXi or on the FlashArray volume

Guest-level Settings

In general, standard operating system configuration best practices apply and Pure Storage does not make any overriding recommendations. So, please refer to VMware and/or OS vendor documentation for particulars of configuring a guest operating system for best operation in VMware virtualized environment.

That being said, Pure Storage does recommend two non-default options for file system configuration in a guest on a virtual disk residing on a FlashArray volume. Both configurations provide automatic space reclamation support. While it is highly recommended to follow these recommendations, it is not absolutely required.

In short:

• For Linux guests in vSphere 6.5 or later using thin virtual disks, mount filesystems with the discard option
• For Windows 2012 R2 or later guests in vSphere 6.0 or later using thin virtual disks, use a NTFS allocation unit size of 64K

High-IOPS Virtual Machines

As mentioned earlier, the Paravirtual SCSI adapter should be leveraged for the best default performance. For virtual machines that host applications that need to push a large amount of IOPS (50,000+) to a single virtual disk, some non-default configurations are required. The PVSCSI adapter allows the default adapter queue depth limit and the per-device queue depth limit to be increased from the default of 256 and 64 (respectively) to 1024 and 256.

In general, this change is not needed and therefore not recommended for most workloads. Only increase these values if you know a virtual machine needs or will need this additional queue depth. Opening this queue for a virtual machine that does not (or should not) need it, can expose noisy neighbor performance issues. If a virtual machine has a process that unexpectedly becomes intense it can unfairly steal queue slots from other virtual machines sharing the underlying datastore on that host. This can then cause the performance of other virtual machines to suffer.

Horizon Connection Server Tuning

1. Use SE sparse Virtual disks format–VMware Horizon 5.2 and above supports a vmdk disk format called Space Efficient (SE) sparse virtual disks which was introduced in vSphere 5.1. The advantages of SE sparse virtual disks can be summarized as follows:
   • Benefits of growing and shrinking dynamically, this prevents VMDK bloat as desktops rewrite data and delete data.
   • Available for Horizon View Composer based linked clone desktops (Not for persistent desktops) only
   • VM hardware version 9 or later
   • No need to do a refresh/recompose operation to reclaim space
   • No need to set blackout periods, as we handle UNMAPs efficiently
2. We recommend using this disk format for deploying linked-clone and instant-clone desktops on Pure Storage due to the space efficiencies and preventing VMDK bloat.
3. Disable View Storage Accelerator (linked-clones only, VSA must be enabled to use instant-clones)
   • The View storage accelerator, VSA, is a feature in VMware View 5.1 onwards based on VMware vSphere content based read caching (CBRC). There are several advantages of enabling VSA including containing boot storms by utilizing the host side caching of commonly used blocks. It even helps in steady state performance of desktops that use the same applications. As Pure Storage FlashArray gives you lots of IOPS at very low latency, we don’t need the extra layer of caching at the host level. The biggest disadvantage is the time it takes to recompose and refresh desktops, as every time you change the image file it has to rebuild the disk digest file. Also it consumes host side memory for caching and consume host CPU for building digest files. For shorter desktop recompose times, we recommend turning off VSA.
4. Tune maximum concurrent vCenter operations—the default concurrent vCenter operations on the vCenter servers are defined in the View configuration’s advanced vCenter settings. These values are quite conservative and can be increased to higher values. Pure Storage FlashArray can withstand more operations including:
   • Max Concurrent vCenter provisioning operation (recommended value >= 50)
   • Max Concurrent Power operations (recommended value >= 50)
   • Max concurrent View composer operations (recommended value >= 50)

The higher values will drastically cut down the amount of time needed to accomplish typical View Administrative tasks such as recomposing or creating a new pool.

Some caveats include:

1. These settings are global and will affect all pools. Pools on other slower disk arrays will suffer if you set these values higher, so enabling these will have adverse effects.
2. The vCenter configuration, especially number of vCPUs, amount of memory, and the backing storage has implications from these settings. In order to attain the best possible performance levels, it is important to note the vCenter configurations and size them according to VMware’s sizing guidelines and increase them as needed if you notice a resource has become saturated.