As a presales systems engineer working for WHIPTAIL, my job crosses a lot of functional groups within the company. It’s my job to figure out how to best help our customers apply the technology built into our products to solve not only technical, but also business, problems. I also have the luxury of advance view into our product roadmaps and help to shape to best fit our current customers, future customer needs and the industry trends. With a select group of customers we’ve been sharing these roadmaps to not only get insight to their thoughts on where we think it’s best to head, but also figure out what types of advanced problems we can apply our solution sets.
One of the most popular questions I’ve gotten after running through the INVICTA product and roadmap has been: “The performance numbers with even a single ACCELA array are extremely impressive, and with a fully populated INVICTA chassis they’re outside of what I ever thought possible. So… what do you use all that horsepower for?”
That’s a great question. INVICTA at launch supports up to six 2U Silicon Storage Nodes supporting 7GB/s and more than 600,000 IOPS, all at latency numbers measured in microseconds, and not milliseconds.
What DO you use that for?
VDI for many customers has been a great idea in concept, but the cost to build storage to meet the performance numbers demanded by a VDI infrastructure often stalls the project. We’ve been able to meet that need in the past with our 2U stand-alone array. But, as customers start scaling past the thousands and into the tens of thousands of desktops, complexity becomes an issue. Imagine the ability to support the performance needs of more than 12,000 desktops from a single storage system that sits in a 14U footprint – that’s INVICTA.
Elsewhere in the data center, other applications also drive incredible amounts of transactions and IO, and the gaps between the application processing and the storage performance continue to widen. Performance numbers for systems hosting databases are measured in the number of transactions per second (IOPS), and those transactions are very often limited by how long it takes to write that data to a storage medium (latency). While the IO per second numbers we can deliver with INVICTA are impressive, what that translates to for a customer is the sheer number of transactions that database may be able to service in a minute, or an hour, or a day. From a single INVICTA unit, it’s possible to support the transaction load driving hundreds of millions, if not a billion or more, transactions in a single day. Those numbers, with a very flat latency at 200 microseconds, could very well mean the difference between thousands of concurrent customers processing a transaction to closure or losing their patience and moving on to your closest competitor.
Other customers use solid state less for the transactional loads and more for movement of sheer volumes of data. The fact that INVICTA can support multiple GigaBYTE transfer rates means tremendous things in areas, to name just a few, like medicine (where doctors may require high resolution stills and video instantly to save a patient), video processing (to service extremely high performance rendering operations), data and number crunching (for high performance computing clusters), and data streaming (high resolution sensors, audio and video caching). To put it in perspective, a single INVICTA cluster supports the transfer of about two full-length DVDs per second. If you’re familiar with streaming movies from your favorite pay-per-month service, imagine streaming more than 10,000 of them at the same time from a single storage system. INVICTA can support those types of workloads.
All of the use cases I’ve mentioned above are use cases we’ve already helped customers with, or customers have come to us with those scales of problems to look for a solution. All of them are incredibly exciting opportunities to let us take the training wheels off of INVICTA and prove what it’s capable of. But, the most exciting thing for me as a presales engineer will be all of the things we haven’t even thought of yet we’ll now be able to address because the storage performance in the data center is no longer the problem.
Imagine if the storage response time was the furthest thing from your mind in YOUR data center. How would that change things for you? If you’re intrigued, let us know. The possibilities may surprise you.