In a recent session of Spiceworks Unplugged, a roomful of IT buyers from a wide range of organizations were asked what type of information they find most helpful when evaluating a product. Product reviews from an unbiased source ranked high on their list. This makes sense; reviews from non-partisan sources tend to convey credibility and authority without the fluff or hype. But too often product reviews come across as dry and are mired in minor details or laundry lists of features instead of providing a meaningful account from an end-user perspective. What many people really want to know is: How will the product actually work in a real IT environment, as seen by someone who has been in the trenches, not a hired mouth?
So we were intrigued when Trevor Potts, a self-proclaimed “full-time nerd” with major tech credentials and a reporter for The Register who has a penchant for insightful and fun write-ups, signed up to perform an independent test of our storage platform. Trevor had complete access to our lab in Mountain View and complete freedom to test whatever he deemed important.
You can read Trevor’s full write up on the test results here. But here are some highlights.
Starting off with a quirky title “You got your Tintri hat, beermat and cat: But what does it all mean? Hype begone: Is this storage management for dummies or what?,” the article definitely piqued my interest. In describing the special way Tintri uses flash, Trevor said “in a typical storage array involving flash, you store your data on a bunch of spinning rust (hopefully with some fault tolerance) and you either copy or move the most frequently accessed (hot) blocks to a flash cache. Tintri's take on flash storage is just that little bit different. They make flash the primary storage and move cold blocks off to the spinning rust.”
You may be wondering, “what’s the big deal about making flash the primary storage (a.k.a Flash First)?" According to Trevor, “The difference is that in Tintri's world new data is ’fast by default.’ This means cloning or snapshotting a virtual machine (VM) takes 3 or 4 seconds, not minutes. The performance of the storage is pretty wild. I was able to clone 12 copies of a 903GB VM in less than a minute; doing so barely seemed to affect the performance of the array or associated VMs at all.” How is this done? 99% of IO is served by flash on a Tintri system. Data on flash is compressed and de-duplicated so that the effective size of flash storage is increased by a factor of 4 to 5. This means a Tintri system can support 3X-5X the number of VMs compared to a general purpose system (with the same amount of flash), while maintaining consistent quality of service.
"Performance trouble shooting? Pop into the UI, select the VM and check the history; the culprit will be plainly visible."
Ever had problems when your application is running slow and you don’t know if the bottleneck is in storage, networking, or servers? Describing the Tintri performance trouble shooting UI as "enterprise storage for dummies class awesome”, Trevor zoomed in on how Tintri “gives you a clear understanding of what's going on with your storage, it actually helps demystify some of the more gnarled issues surrounding bottlenecks in a virtualized environment. Tintri…gives you the ability to find out what subsystem to blame when things start getting slow. You can see IOPS, throughput, latency, flash hit % and more from the level of individual VMs.”
Providing the visibility and management of individual VMs is of course what we have always believed to be invaluable insights for system and storage admins to manage virtualized environments. Before virtualization, when each storage LUN or volume was dedicated to a single application or physical host, an admin was able to perform data management or set policies on dedicated LUNS associated with an individual machine. In a virtualized world, each LUN can store tens or hundreds of VMs. The old ways simply don’t work anymore, hence the importance of storage that’s capable of doing data management, data protection, and troubleshooting at the VM level.
"It just works."
Trevor ends his review with a quote from a Tintri customer, "As far as performance goes, it's incredible. It just works."
The customer, Washington State University, had a very specific pain point to overcome; a looming VDI deployment of 200-250 VMs ranging from XP to Windows 7. When describing what made them take the jump, the answer was: it was the cost. "NetApp could deliver the same level of IOPS but the cost point was much greater. Tintri was also a much smaller physical footprint, lower power consumption, etc. When you first get it you spend a lot of time looking at the UI, the load, etc. Now I'm just confident I can throw stuff at it and there won't be a problem."
In the end, the great thing about this review from Trevor is that it was completely independent and not a paid promotion. Give us a call or request a demo and we’d be happy to let you play with a Tintri system, too. Who knows, you might like it so much you’ll decide to write your own product review!