The ZFS file system and volume manager is quickly becoming a preferred operating system (OS) among providers in the competitive arena of virtual storage. Lending to its popularity, ZFS is flexible, functional and consistently high performing but can occasionally cause minor troubles for inexperienced users. In this fourth and final installment of my interview series with Rob Commins, VP of Marketing at Tegile Systems, we discuss how Tegile’s use of ZFS provides the fully integrated and supported storage platform that organizations want while keeping application availability and performance levels high.
Jerome: Tegile uses ZFS on its operating system. How often does that come into play as a factor when organizations are making a storage buying decision?
Rob: Customers are actually big fans of our use of ZFS. The primary reason they like Tegile’s use of ZFS is that there are hundreds of thousands of instances of ZFS in production in the world. When a new storage vendor comes to market with its own proprietary file system, it may have tens, hundreds or eventually thousands of instances out there in production. But that’s a far cry from hundreds of thousands.
Tegile feels that ZFS has been shook out well and been validated by the market. Tegile finds that the people that have the biggest problem with ZFS are the vendors that do not use ZFS.
From a control standpoint, it is very simple for Tegile to get its hands on a certain branch of code that it likes. Then Tegile optimizes certain portions of ZFS that it thought could be improved while removing certain parts and then completely writing its own sections of code.
For example, ZFS out of the box has deduplication. But the way it is written actually slows down the performance of the array and applications start getting into those long response times. If one reads the user guides out of the Oracle ZFS products, it practically has a big red X in the deduplication sections: “Thou shalt not use deduplication in high performance low latency environments.”
Tegile Systems has actually cut that code off inside of ZFS, wrote its own metadata accelerated code, which increases performance when one uses deduplication. Tegile has patents awarded to it on that process. Tegile has been able to establish and maintain its differentiation in its IP. It is “standing on the shoulders of giants” on the baseline of functionality that ZFS gives us.
Jerome: You provide ZFS with Tegile’s deduplication code in it for the customer and they are able to implement this aggregated code on the Tegile arrays themselves?
Rob: No. Tegile compiles all of the codes into its own operating system upon which it runs its arrays. There are companies out there that have a free download of their ZFS-based operating system which allows customers to roll their own type of array.
Tegile Systems has found that most customers who do so end up being unsuccessful. Storage is really easy until something breaks and most of these customers who have tried to play the game have gotten themselves into trouble. They personally do not have the time or expertise to triage and fix their storage problems on their own.
The customer ends up throwing it away and calling a place like Tegile that has a fully integrated and supported system. The customer may still be a ZFS fan but they simply come to Tegile for support and assistance. Similar to how most car-owners do not change their own oil anymore, most users find that managing storage is not always their highest value-add activity.
Jerome: Anything else you would like to add, Rob?
Rob: I would like to touch on unified storage. Tegile offers block, iSCSI, fiber channel, as well as NFS and CIFS out of the same array.
A few days ago Tegile was at VMware’s headquarters. VMware’s virtual desktop people are making a big push they call “desktop and data” in which VMware will run the desktop operating system and apps over a block interface. Then it will run file services to those users over a file interface. Having a unified storage system that can address both of those use cases out of the same array is very, very attractive.
Jerome: Has Tegile Systems seen a performance hit as a result of using a unified storage platform?
Rob: A negligible hit on performance. Tegile benchmarks a system by running a file interface. Then Tegile runs a block-based benchmark. At the end, Tegile runs them concurrently. On the back end, organizations can saturate their disk hit a limit. But if they leave enough room in the back end to run both those benchmarks concurrently, Tegile has not seen a material performance penalty.
In Part I of this interview series, we talked about how sub-millisecond response times are the new gold standards in storage system performance.
In Part II of this interview series, we examined what the tipping point is when going from hybrid to flash.
In Part III of this interview series, we examined the key technologies that differentiate Tegile from its competitors.