VAAI Comparison – Block versus NAS

Posted on November 8, 2012 by Cormac — 12 Comments ↓

I get a lot of questions around how the vSphere APIs for Array Integration (VAAI) primitives compare from a protocol perspective. For instance, a common question is to describe the differences between the primitives for NAS storage arrays (NFS protocol) and the primitives for block storage arrays (Fibre Channel, iSCSI, Fibre Channel over Ethernet protocols). It is a valid question because, yes, there are significant differences and the purpose of this blog post is to detail them for you.

Let’s begin with a brief overview of VAAI. Its primary purpose is to offload certain I/O tasks to the storage array, freeing resources on the ESXi host to do other tasks. So what primitives are there? Here is a table listing the Block, NAS & Thin Provisioning primitives:

What are the main differences between the block and NAS primitives?

Continue reading →

Tintri adds new VM-aware features and VAAI Support

Posted on September 21, 2012 by Cormac — 2 Comments ↓

During VMworld 2012 in San Francisco, I had a chance to catch up once again with the team from Tintri. My first introduction to Tintri was at last year’s VMworld, where they received runner-up in the ‘Hardware for Virtualization’ category by TechTarget for best of VMworld 2011. Well this year they went one better, and won the Best of VMworld 2012 Gold award in Hardware for Virtualization. And for good reason. Let’s see the enhancements to the Tintri platform over the last 12 months have brought.

Snapshot & Cloning Capabilities

Tintri were one of the first VM-Aware arrays. All the operations on the array were built around the VM. In fact, this is something that VMware is also embracing with the tech preview of virtual volumes that was given in this year’s VMworld keynote, so Tintri are ahead of the curve here. With a VM-aware array, snapshots and clones are done on a per-VM basis rather than on a per LUN or per volume basis. This approach gives customers greater operational granularity and space savings. The demo given to me by Ed Lee (Tintri Chief Architect) and Kieran Harty (Tintri CEO) at this years VMworld 2012 in San Francisco was very captivating. Ed showed me their ability to create many hundreds of VM snapshots/clones in the space of seconds. All of these snapshots and clones are thinly provisioned and already space efficient, so that any unused space can be reclaimed by the array. The space efficiency aspect is something that VMware are also embracing with the announcement of SE Sparse Disks in vSphere 5.1. Incorporated into the Tintri snapshot and clone mechanism are VSS callbacks for quiescing applications and filesystems within the Guest Operating System so that consistent snapshots can be taken. Tintri also have the ability to use existing VMware customization specifications to customize snapshot/clone copies of the VM from a Guest OS (sysprep) perspective. These features, combined with support for VCAI (View Composer Array Integration) in View 5.1 lend themselves very nicely to VDI, and it comes as no surprise that Tintri have positioned themselves as an ‘array of choice’ for VMware View implementations, amongst other enterprise applications.

VAAI Integration

Tintri’s VM-aware datastore is in fact NFS, so it is nice to see that they have fully supported the VAAI-NAS primitives. VAAI is short for VMware’s vSphere Storage APIs for Array Integration. The first of the primitives implemented is the extended statistics feature. This is where details about the actual space consumed on the ‘storage pool’ on the array can be surfaced up into vSphere. This allows admins to more easily understand the storage consumption rate and plan in advance for additional storage, rather than ‘fire-fighting’ when disk space suddenly runs out. The second primitives related to the implementation of the Reserve Space primitive, which allows you to create thick provisioned VMs on NFS datastores so that you don’t need to worry about your VM running out of space as it writes to new areas of the virtual disk. The third and most impressive feature is the implementation of the File Cloning primitives, which allows you to create clones from the vSphere GUI on Tintri in just a couple of seconds, regardless of the size of the VM that you are cloning! You can learn more about VAAI and the different primitives from a post I did on the vSphere Storage blog here.

Futures

Ed also gave me a sneak-peak at their upcoming replication technology. Basically, again because this is VM-aware storage, replication can be done between arrays on a per VM basis. One of the stand out parts of their replication technology is the ability to make a clone copy of the replica at the destination and verify that it does indeed power-on and the application running in the Guest OS is in a consistent state. Tintri have plans to integrate this replication technology with VMware’s Site recovery Manager product, but no timeframes on that just yet. One of the things which I quite liked was the ability to observe the replication network bandwidth savings due to dedupe and compression within the UI. While many vendors claim dedupe and compression ratios, Tintri put it right out there for you to see for yourself. Very nice.

You can learn more about their new enhancements here. Tintri will also be making their second VMworld EMEA appearance in Barcelona this year. I’d urge you to go along and see them.

Get notification of these blogs postings and more VMware Storage information by following me on Twitter: @VMwareStorage

Pure Storage adds vCenter Web Client Plugin, VAAI, iSCSI & more to new release

Posted on September 20, 2012 by Cormac — 1 Comment ↓

Pure Storage are an all-flash enterprise storage company. I first met these guys at VMworld 2011 and was quite impressed by their product. Like many Flash Array vendors at the time, there wasn’t a great amount of vSphere integration features. However, with this latest release of Purity v2.5, Pure Storage are addressing this and more. I had a chance to meet and discuss these new features with Matt Kixmoeller & Ravi Venkat of Pure Storage recently. Not only are they now VMware-Ready certified, but they’ve got a whole bunch of integration features. Let’s have a look at the features that they have added to their new release.

vCenter Web Client Plugin

For those of you who have not being paying a lot of attention to the 5.1 announcements, the current C# vSphere client is being deprecated. VMware is moving to a new web client. So it is great to see that Pure have already announced their array management plugin for the new web client. There are a lot of cool management and monitoring features in the client for Pure Storage arrays, including the ability to show various levels of space consumption at the VMDK, VMFS and at the array back-end level. It also shows the gains achieved from dedupe & compression, as well as the ability to reclaim stale or stranded space via the UNMAP primitive. I believe Pure Storage are the first array vendor to have a web client management plugin available for vSphere 5.1. Great job guys.

iSCSI Support

When I met with Matt at VMworld 2011, Pure Storage only supported Fibre Channel interconnects. With this new release, Pure Storage arrays will now also support both FC & iSCSI. I think this is a very good move. Many vendors these days support lots of different protocols on the same array, and many customers may feel that an IP based network is more desirable than a dedicated FC infrastructure, especially with the advent of 10Gb (and faster networks).

VAAI Support

This is always welcome. The ability to offload tasks to the storage array rather than consume resources on the ESXi host is such a great idea, and it nice to see Pure Storage support the VAAI block primitives. When we spoke, Pure had VAAI certification for two primitives on their FA-320 arrays. They now support the ATS (Atomic Test & Set) primitive which is the replacement for SCSI reservations as well as the Write Same/Block Zero primitive. There is no XCOPY support yet (for cloning) but one assumes that this is a work in progress. Great to see.

Snapshots

Pure have introduced a new ZeroSnap Snapshots technology in this release too. These snapshots are thin, compressed and deduped. Any of these snapshots can then be converted into a new volume for backup/recovery purposes. Because of their lightweight nature and the underlying technology, Pure are claiming to be able to create 1,000 snapshots in a minute.

Replication

Pure Storage do not have their own replication technology just yet. However, there is full support for VMware’s vSphere Replication product which can replicate Virtual Machines across different storage arrays without the need for a third party replication software.

VMworld EMEA & Europe in general

Pure Storage are going to be at VMworld EMEA 2012, exhibiting their flash array and the new features discussed here. I’d urge anyone interested in flash technology to go and see them. Another cool thing of course is that Pure Storage have announced their intention to expand into EMEA and are actively seeking an EMEA HQ. You can find out more about the new Purity 2.5 features here.

Get notification of these blogs postings and more VMware Storage information by following me on Twitter: @VMwareStorage

Heads Up! ESXi cannot distinguish between EagerZeroedThick & LazyZeroedThick

Posted on September 14, 2012 by Cormac — 3 Comments ↓

VAAI NAS introduced the ability to create LazyZeroedThick & EagerZeroedThick disks on NFS datastores. Without VAAI NAS, one can only create thin VMDKs on NFS datastores. For those of you who are using VAAI NAS plugins, there is an important note in the 5.0U1 release notes that you should be aware of.

ESXi cannot distinguish between thick provision lazy zeroed and thick provision eager zeroed virtual disks on NFS datastores with Hardware Acceleration support
When you use NFS datastores that support Hardware Acceleration, the vSphere Client allows you to create virtual disks in Thick Provision Lazy Zeroed (zeroedthick) or Thick Provision Eager Zeroed (eagerzeroedthick) format. However, when you check the disk type on the Virtual Machine Properties dialog box, the Disk Provisioning section always shows Thick Provision Eager Zeroed as the disk format no matter which format you selected during the disk creation. ESXi does not distinguish between lazy zeroed and eager zeroed virtual disks on NFS datastores.

This can be confusing, especially if you select “Thick Provision Lazy Zeroed” when creating a virtual machine disk during deployment, and later you edit the properties of the VM and it states that the disk is “Thick Provision Eager Zeroed”. This is the crux of the issue; we will always report Eager Zeroed, even if the VMDK is Lazy Zeroed. Some of the guys at EMC have already run into the issue, which I how I got a heads-up on it.

On VMFS, we can query TBZ (To Be Zeroed), which can tell us that there are blocks that are provisioned but not yet initialized. I wrote an KB article on how you can do this via vmkfstools. On NFS, we have no way of retrieving this information.

Get notification of these blogs postings and more VMware Storage information by following me on Twitter: @VMwareStorage

vSphere 5.1 Storage Enhancements – Part 3: vCloud Director

Posted on September 6, 2012 by Cormac — 16 Comments ↓

In this post, I want to highlight a number of storage improvements made in vSphere 5.1 that are going to be leveraged by the next release of vCloud Director.

Scalability

First off, we have the new file sharing scalability enhancements made in VMFS-5, which now allows up to 32 hosts to share a single file. This is covered in detail in part 1 of this vSphere 5.1 storage enhancements series of blog posts, but what this does mean for vCloud Director is that vApps deployed on linked-clones can now have many more hosts sharing the base disk on a VMFS-5.

VAAI NAS Offload

Sphere 5.0 introduced the offloading of linked clones for VMware View to native snapshots on the array via NAS VAAI primitives. You can read more about this here. vSphere 5.1 NAS VAAI enhancements will allow array based snapshots to be used for vCloud Director vApps based on linked clones, in addition to being used for VMware View.

When VMware vCloud Director does a fast provision of a vApp/VM, it will transparently use VAAI NAS to offload the creation of the subsequent linked clones to VAAI supported arrays.

Just like VAAI NAS support for VMware View in vSphere 5.0, this feature will also require a special VAAI NAS plug-in from the storage array vendor.

At the time of writing this article, NetApp already have this feature included in their next VSC release (4.1) which is currently in beta.

If “Fast Provisioning” is used on the Org vDC Storage settings AND the check box “Enable VAAI for fast provisioning” on the overall system Datastore settings is selected, it will trigger the right commands to use a native array-based snapshot for a linked clone instead of a standard redo log based one.

Profile Driven Storage Interoperability with vCloud Director

Storage Profiles are now represented in vCloud Director. Storage Profiles still must be configured from the vSphere layer, but they now surface up into vCloud Director. The storage profiles must first be added to a Provide vDC. For example, you might have Gold, Silver & Bronze storage profiles created. This then allows storage to be allocated and managed on a per ORG vDC. Again, continuing our example, this organization can only use datastores which are tagged as ‘Silver’. This support for Storage Profiles allows a high level of seperation between organizations at the storage level. Below is a snapshot of an ORG vDC with two storage profiles, one for iSCSI storage and one for NFS storage.

If the Storage Profile associated with a vApp is changed (this can be done via the properties of a vApp), the vApp is automatically Storage vMotion’ed to a compliant datastore. It is great to see vCloud Director leveraging this excellent vSphere feature.

Storage DRS Interoperability with vCloud Director

One of the major enhancements in vSphere 5.1 is to provide interoperability between Storage DRS and vCloud Director. This essentially means that vCloud Director 5.1 now recognises datastore cluster objects from Storage DRS. Just like Storage Profiles, the configuration of Storage DRS is done at the vSphere layer, but the resulting datastore clusters and their respective configuration surface up into vCloud Director. In order for this interoperability to work, Storage DRS now understands linked clones (which it didn’t do previously). Going forward, vCloud Director can now use Storage DRS for initial placement, space utilization and I/O load balancing of vApps based on linked clones.

Snapshots

The last feature introduced in vSphere 5.1 & vCloud Director 5.1 is the ability to take Virtual Machine snapshots from within vCloud Director. Previously one had to take these snapshots at the vSphere layer. As per the screen shot on the left, you can now Create, Remove and Revert a snapshot via the vCloud Director UI.

Although this might be considered a minor improvement, it does alleviate some additional administration which was necessary in previous versions of vSphere/vCloud Director.

I guess the next question then is how do you tell if you have a snapshot on the VM?

By default this information is not displayed on the Virtual Machine view. To show this information, select the option to display the Column headings which is on the right of the screen. Place a tick in the Snapshot column. You will now have a column denoting whether or not there is a snapshot for the Virtual Machine as per the diagram below,

It is nice to see these vSphere storage features being leveraged by vCloud Director. It’s especially nice to see some of the interoperability between products and features.

Get notification of these blogs postings and more VMware Storage information by following me on Twitter: @VMwareStorage