NetApp and VMware View (VDI) Best Practices for Solution Architecture, Deployment and Management – Part 7

Posted by Chris Gebhardt – Reference Architect, Server and Desktop Virtualization

Intelligent Caching

I hope you have enjoyed reading the previous 6 parts of this blog post.  Today’s blog will focus on NetApp’s Intelligent Caching which is built into Data ONTAP and enabled by the use of block sharing technologies such as FlexClone and Deduplication.

Deduplicated & Non-duplicated data

A couple years ago, I had the privilege of listening to Mike Riley speak on NetApp Deduplication just as NetApp was about to release it.  He talked about the space savings that Dedupe was going to bring to our customers but then he offered a new perspective.  He started to describe our current technologies in terms of “non-duplication.”  He gave an interesting explanation:   Customers understood the benefits of deduplication.  It’s intuitive for most customers.  If you’re familiar with dedupe, well then non-dupe would be just as easy to talk about.  Basically, what Mike meant was that we already had many technologies that focused on not creating redundant copies in the first place that provided us with significant space savings over our competitors.   He began discussing that from the time NetApp started developing technologies such as Snapshot™, RAID-DP™ & FlexClone®, we were taking steps to reduce the customer’s storage footprint.  We primarily viewed these technologies as data protection or enhanced development capabilities.  They are but Mike was viewing them through the prism of Storage Efficiency  (and here) as well.  With Snapshots & RAID-DP it was apparent that customers could save money in not having entire volumes dedicated to snapshots or have to use expensive RAID 10 to get enterprise class data protection.  What I didn’t realize at the time was how important FlexClone and Deduplication would become in that customers could leverage these technologies to accelerate the performance of their VMware environments.

So what is Intelligent Caching?

Intelligent Caching is a technology that allows customers to make read cache more efficient as compared to caching in traditional storage arrays.  NetApp does this by caching only a single copy of blocks that are deduplicated or non-deduplicated (FlexClone).  Cache efficiencies can be as great as 255:1 within a single NetApp volume.  Using Deduplication with File and Volume FlexClone decreases the amount of data on disk and also reduces the amount of Cache required as well.

An Example of Disk and Cache Efficiency

So what does that mean?  For example, a company wanted to role out 1000 Virtual Desktops and their OS disk was 20GB in size and had 250 VMs in a 4 NFS datastores or 10 VMs on 100 LUNS.  Calculating pure capacity on a traditional storage array they would need at least 20TB of storage just for the OS disk. This calculation is excluding all user data, transient data (for example vswap & pagefiles).  The same is true for cache.  If the working set is 200MB per VM, the amount of cache required to serve this data from cache is 250GB.

Using NetApp Dedupe and FlexCloning one would need about 50-100GB of disk space as the OS is 25GB and there is a metadata overhead.  The same would apply to cache.  Since the working set of a XP virtual machine is somewhere in the range of 200MB, with a NetApp solution, it would require roughly 2GB of cache to serve the working set for all 1000 VMs.  This allows customers to save costs even further in their VDI deployments when combining Space Efficient Storage with Intelligent Caching.

Details of Intelligent Caching

Data ONTAP will only store a single block in cache for up to 255 physical blocks. Data ONTAP intelligent caching is available in all versions of Data ONTAP 7.3.1 or higher. This means that intelligent caching can be used in every FAS, V-Series, & IBM N Series that support Data ONTAP 7.3.1 and block sharing technologies.

How Data ONTAP Intelligent Caching Functions

When a data block is requested, ONTAP reads the block into main memory (also known as the WAFL® buffer cache). If that data block is a deduplicated block, in that it has multiple files referencing the same physical block, each subsequent read of that same physical block will come from cache as long as it has not been evicted from cache. Heavily referenced blocks that are frequently read will reside in cache longer then blocks that have fewer references or less frequent access. The effect this has is that since main memory can be accessed much faster then disk, latency is decreased, disk utilization is decreased, and network throughput is increased thus improving overall performance and end user experience without any negative tradeoffs.

Extending NetApp's Intelligent Caching with PAM

NetApp has the ability to extend ONTAP’s native Intelligent Caching with the use of the Performance Acceleration Module.  As long as that block is not evicted from both caches, all subsequent reads will be performed from main memory or PAM, thus improving performance by not having to go to disk.  Again the more heavily the data is deduplicated and more frequently accessed the longer it will stay in cache. Intelligent caching combined with NetApp disk deduplication provides cost savings on many levels without any negative tradeoffs.

So what should I do?

Since read IO can be greatly reduced by using intelligent caching NetApp recommends the use of Data ONTAP 7.3.1 or later.  This version supports intelligent caching across our entire truly Unified Storage product line.  For Mid & High Tier systems the use of Intelligent Caching with both Data ONTAP and extension of cache using the PAM card provides extended capabilities for larger environments.

The net result of intelligent caching is two fold, customers will buy less storage to serve IO’s from intelligent cache and they will go faster by utilizing deduplication and NetApp’s Intelligent Caching in VMware View environments. The end result will be enhanced user experience.

We’ll have more detailed information on this subject in the version 3.0 of TR3705.

In the part 8 of this blog series, I will describe how the NetApp RCU 2.0 can be used to deploy a space efficient VMware View environment.

Comments

[Joannah]

I recently came across your blog and have been reading along. I thought I would leave my first comment. I don't know what to say except that I have enjoyed reading. Nice blog. I will keep visiting this blog very often.

Joannah

http://2gbmemory.net