In this Episode I’m with Hans O’Sullivan, StorMagic‘s CEO, and we talk about:

– Hyperconverged infrastructures for the distributed enteprise
– How HA is managed in small and remote sites
– Manging multiple (up to thousands) sites
– KVM support

StorMagic will be presenting at next TECH.unplugged event in Amsterdam next 24/9. A one day event focused on cloud computing and IT infrastructure with an innovative formula combines a group of independent, insightful and well-recognized bloggers with disruptive technology vendors and end users who manage rich technology environments.

Here the full transcript of the show

Enrico: Hi everyone and welcome to a new episode of JukuBeats. Today, I’m here with Hans O’Sullivan, CEO of StorMagic. Hi Hans. How are you?

Hans: Hello Enrico. Nice to meet you again.

Enrico: You know how this podcast works. The idea is three questions in five minutes and I go straight to the question. First of all, StorMagic has an interesting VSA, which can be used to build hyper-converged infrastructure. Contrary to what we usually see from other vendors, this one is targeted to distribute an enterprise developments. What does it really mean?

Hans: Yeah. Well, our target customer, the distributor’s enterprise is typically a large enterprise customer with multiple sites. The problems that they actually have is very, very different to other forms of organizations. Our competitors, unfortunately, consider that a distributor’s enterprise is like a series of small to medium sized businesses and that they can solve their problems the exact same way.

However, distributor’s enterprise is very, very different. They typically have a hundred, maybe a thousand, maybe more than a thousand sites. Each one of those sites tends to be the exact same as the others. It’s doing a job that’s identical to the others. They typically will have no IT capability on those sites. They typically will not make many changes to those sites as they go through the years. They don’t want anybody onsite to be able to get on to the system, surf the web, or anything like that. They just wanted to do a particular job.

Very, very important to them is, One: Keep the it very, very simple. It needs to be simple. It needs to do the job and do the job well. Two: Centralized Management. They will not have those IT resources to make changes or to modify things. They need to be able to monitor and to make changes to those sites from a central location, often monitoring thousands of sites.

That is really where we tend to specialize. One, in the centralized management. Two, in keeping it extremely simple at the edge of the enterprise, where all those hundreds of thousands of sites are. That starts with even the implementation. We can provide a virtualized solution, with full high availability in a simple, two-server configuration. We are the only company that can actually do that in a simple two-server configuration.

Enrico: Well, two nodes. Two nodes, it’s really a small infrastructure and it could be really difficult to manage folds, especially split-brain conditions. How do you manage this kind of problems?

Hans: Yeah. It is something we’ve spent a lot of time and effort in, once more, trying to keep it extremely simple. Being able to do it in two nodes is very important because if you have a thousand sites, a third server on each of those sites adds 50 percent of cost. That’s an awful lot of cost, but at the same time, you still need some of the functionality that our competitors insist that you have three servers for. To solve the split-brain problem, the best technical way of doing that is to have a third node, a witness node. A node that actually arbitrates on which one of your two … the two sides of your mirror, should actually become the primary in the case of a failure.

What you’re worried about in split-brain is a situation where both halves are not capable of talking to each other and they don’t know if the other half is dead or alive. Both of them could end up modifying the data. If they’re both modifying the data at the same time, that leads to data corruption. What we do to solve that is we still provide a quorum device, but that quorum devise can be remote. Typically, the monitoring system that’s in the center in headquarters could provide quorum devices for hundreds of thousands or remote sites, in one simple configuration, also with high availability built in to it. We also provide techniques that, should you lose communications to the center, that we have multiple communications paths between the two nodes. Even in that short period of time, where you may lose the quorum, we built in some fail safe structures in there to minimize the chance of you ever setting into a split-brain type situation. In fact, you would have to have four different failures before you could ever get to split-brain.

Enrico: It sounds really interesting. You mentioned that the central administration many times. This is a compartment thing when you have to manage thousands of different sites. Is it a web service? Is it a central web server? How does it work?

Hans: Well, its actually quite important to be able to build in to the management structure that the customer themselves, self wants and are used to. We can monitor over standards like SNMP and the actual management would be done through consoles such as, in the case of VMware, it would be vCenter. In the case of Hyper-V, it would be System Center. Or if the customer wishes, you can also use our web-based service to actually provide that managing capability. The most important thing is that it fits into what he is actually used to and is familiar with.

Enrico: Good. You mentioned also Hyper-V. Usually I ask a free question, but in this case, I’m really curious. You already support Hyper-V, which I think it’s almost identical to VMware from the point of view of the implementation of VSA. Do you also plan to support KVM in the future?

Hans: Yeah. Our intention is to actually be agnostic to all hypervices. If I look at what’s happening with our business at the moment, traditionally, the vast majority of our business has been on vSphere. If I look at what’s in our pipeline, there’s this significant portion in the pipeline that is now Hyper-V. A lot of the market is beginning to move that way. If I look at what our salespeople are telling us, we’re receiving quite a lot of queries about KVM support. However, we’re not actually seeing that translating yet into business. We’re projecting that it will. We do have KVM working today in our lab, but we have not [inaudible 00:06:14] that yet, ready for shipping. We do expect it to become very important in the not too distant future.

Enrico: Okay. Thank you very much.

Hans: Thank you Enrico. It was a pleasure talking to you again.

Enrico: Thank you again. Bye.

Hans: Bye.