The Next Generation Data Center is a concept that companies like SolidFire use to describe an ideal of storage performance, scale, efficiency and management in multi-tenant cloud environments – public and private clouds. But more than an ideal, companies in this space, like Ebay and PayPal, are actually using solutions like SolidFire’s all-flash arrays to help deliver on that Next Generation Data Center promise. Storage Switzerland asked Jay Prassl, VP of Marketing at SolidFire to talk about some of their customers, both cloud providers and enterprises, that are deploying SolidFire arrays and how they’re being used.
For a detailed discussion on overcoming the storage challenges in the Next Generation Data Center please watch the Storage Swiss webinar “Storage Requirements for the Next Generation Data Center”, available On-Demand.
First Question, Charlie: “Who are some of the customers who are using SolidFire already?”
Jay: A couple folks we mentioned earlier, Ebay and PayPal, are using SolidFire technology. We mentioned Colt, which is another pan-european cloud provider, out of the UK. We also have SunGuard. I’ll give you an interesting data point: SunGuard is deploying SolidFire across six or seven data centers that they have both in the U.S. and overseas. And we see other enterprises (unfortunately I can’t name all of them) that are really coming to the table, looking at the features and functionality that we talked about earlier today. They all need flash and all want that performance. They’re kind of finding flash as the starting line. But as they look to scaling and into the other features we’ve been talking about, really we’re finding them settling on SolidFire as a result.
Charlie: “What is the largest cluster deployed by a customer today?”
Jay: Yeah, really good question. The largest cluster in production today is a little over 25 nodes. And that’s in a single environment, so that’s one storage cluster. We are seeing customers like Ebay running over 110 SolidFire nodes. But in this case they’re doing so across many different clusters that they’ve established. They’re finding themselves taking nodes from one cluster and adding them to other clusters as they need to from time to time.
Charlie: “What is your mix of cloud service providers to enterprise customers?”
Jay: This is a question that actually comes up quite a bit. Today we are in sort of a 70:30 mix, [70%] cloud providers to [30%] enterprises. We expect over the year this year to find that at the 50:50 mark, if not flowing in the other direction, as more and more enterprises get comfortable with the conversation we’re having today.
Charlie: “What do you view as the difference between ‘guaranteed’ and ‘predictable’ performance?”
Jay: Often times I think, “predictable” and “QOS” can be used together, and it’s very broad. Rate limiting is predictable to some extent, but when you really dig into rate limiting, it’s like a governor on your car. It is designed more to protect the car (or the storage system) than it is to actually deliver you a predictable experience. Right?- because there’s no minimum construct in rate limiting. When you think of quality of service from a ‘guaranteed’ perspective, you have to consider more components.
The example I like to give to people (to the question of guaranteeing storage performance) is about being able to deliver that performance regardless of what’s happening in the system – like disk failures, nodes being added or taken away, etc – but also regardless of what other applications are doing on the system. To approach it on both levels like that is a difficult problem to solve, but [it’s] also where guaranteed performance is rooted. Because even in a tiered storage system for instance, you may move data from one tier to the next, [but] what kind of performance do you receive while that movement takes place? And how do you eliminate Noisy Neighbors in that environment? The answer is that it’s very difficult, if not impossible, to do. That’s why guaranteed quality of service is something we both use very specifically in our language, but also back it up when we prove and kind of put the solution on the ground for folks.
Charlie: “You made a comment about how flash is like a mother’s fine china, can you repeat that?”
Jay: (laughs) Being a marketing guy, it’s always useful to have metaphors at the ready. Often I think enterprises are thinking of flash still today as something that is only used for a certain, specific type of application that requires rip-roaring performance. So I use the metaphor ‘like your mom’s fine china’ you only use it for special occasions, or very specific applications in this case. When we look at flash today from a different perspective, flash can actually be used for almost all applications within an enterprise data center today, in particular, I think very powerfully from a SolidFire perspective. Because [as an example] I can take George’s WordPress blog and make sure he gets 100 IOPs and I can take Eric’s database and make sure it gets 5,000 IOPs. I can do it all within a single infrastructure and I’m able to be very specific in what each application receives. You can think of it as using your mom’s fine china for everything at this point.
Charlie: “If SolidFire arrays are co-located with other storage arrays in the same data center, how do they handle integrating management?”
Jay: We certainly see SolidFire being deployed alongside a variety of storage systems out there. I think this is one of SolidFire’s more powerful integrations with key management frameworks like Openstack, VMware, Cloudstack, and others. We’re finding customers start to want to manage many of those storage infrastructures behind those frameworks itself. Because we’re integrated with those systems, we can now expose the unique quality of service function, the unique provisioning functions of the SolidFire array, within those frameworks is how to think about that.
Now, also I want to make sure I’m covering another aspect of this question, which is where you might’ve been going. “Does SolidFire have a management framework that would manage, let’s say an EMC system or something else?” It’s important to know the answer – we do not. We obviously manage our own system very well, it [SolidFire’s management framework] really works with this management or orchestration layer and sits above the SolidFire layer, along with an EMC system, along with your compute and networking side as well and actually orchestrates all of the components together. So hopefully that answers the question well enough for you, if not go ahead and fire another one in here and we’ll make sure we take care of it.
Charlie: “Are your arrays block [storage] only? And if yes, are File [storage] and Object [storage] in the cards?”
Jay: So file and object… It’s not [in] our product today, not something we’re actually delivering. [But] what is important (and why this is a tricky question, although it may not seem that way) is that object storage is something we interface with very nicely. We released a new capability called Integrated Backup and Restore. Many of our cloud providers and some of our enterprises actually have object storage environments. (You can think of Amazon S3 as a good [example] of what object [storage] would be.) You can take backups from a SolidFire system and actually push them [to] and pull them from object storage natively [with] our product.
So that is a very unique type integration. To my knowledge there is no other storage product in the world that is capable of that. So, when you think of a Next Generation Data Center using maybe your own private cloud, as well as having connectivity to an Amazon object storage or a SWIFT based object storage environment, you can push and pull those backups to and from a SolidFire system natively. And the key requirement here is simply support of either AWS or SWIFT API’s that we’re well integrated with.
Large cloud-based businesses are the companies that come to mind when one thinks about the ‘state of the art’ in data center infrastructure design. But it’s clear from this discussion that there’s a lot that enterprises can gain from these Next Generation Data Centers – and indeed are, based on the trends that SolidFire is seeing with their customers. Features like guaranteed quality of service and integrated backup and restore (to object storage systems) are making solutions like those from SolidFire the right choice for more than just the most extreme workloads that many associate with all-flash arrays. For more information, please watch the Storage Swiss webinar “Storage Requirements for the Next Generation Data Center”, available on-demand.
SolidFire is a client of Storage Switzerland