I just got back from LinuxCon, KVM Forum, and oVirt Workshop in Edinburgh, Scotland. I thoroughly enjoyed the conference and the city. It’s always great to see old friends and make new ones. One of the questions that came up frequently, whether asked directly to me or indirectly (overheard), was what is NetApp doing here? The other one was, “why is NetApp getting involved with OpenStack”? The second one seemed to have quite a bit of resentment in addition to the genuine curiosity. Back to that later.
As to why NetApp is involved in Linux, I’ve touched on that before in several ways, most recently here. In short, Open Source and Open Standard are a big deal to us because they’re a big deal to our customers.
Let’s move onto OpenStack. This was definitely a different feeling while I was LinuxCon. I honestly felt some resentment from some folks. As in, “why do THEY have to be here, this is OUR thing”. I did not ever hear those exact words, but there were many references to “expensive enterprise storage” and other comments that were completely dismissive of enterprise storage. And even some looks from some that reminded me when I went to what turned out to be a “locals only” beach in Aruba…
Most of the comments lumped all enterprise storage together, disregarding any intrinsic value that we may be able to provide. Notice that I said “may provide”. I don’t profess that enterprise storage is the right solution for everything. But I don’t automatically dismiss it either. “Enterprise” storage and free Open Source storage are different animals. “Better” is completely dependent on use case, so it’s reckless to dismiss either based solely on perceived cost or prejudice. After all, expense, performance, and reliability are all relative.. Look at it this way:
- If cost was the only thing important to folks, we’d all be driving cars made by Kia, and Starbucks would be long out of business.
- If performance was the only important thing, we’d all be driving exotic sports cars
- If reliability was the only consideration, we’d all be driving a Toyota RAV-4
Here’s something intrinsic to all storage solutions: you can’t scale out your environment any faster than you can scale out your storage. If it’s painful to scale your storage, it will be painful to scale your environment. I don’t think we’ll get much argument from the good folks involved with Gluster or Ceph. On the other hand, I think one of the primary things that they would push back on me is the idea of “software defined storage” as the way of the future.
Guess what.. I’m not arguing. The truth is this, regarding NetApp: the special sauce is not the hardware, it’s the software. The hardware simply enables the software. The stringent specs we have for the hardware ensure predictable performance and the ability to size properly for workloads. It’s more difficult to do that same “right sizing” with software storage on commodity hardware.
The NetApp special sauce includes things like published API’s, volume and file cloning, volume snapshots, file & block storage from the same controller, backup frameworks, that result in true scale out scale out and world record performance. My point is this: if your only requirement from storage is object and/or block storage with multi-site replication, then by all means stick with Ceph, Gluster, or any of the other cool software storage solutions out there.
But if you need high performance, predictable performance, block/file/object storage, cloning, thin provisioning, multi-site replication, or any combination therein, you need to look at enterprise storage. You’d be surprised at how close the cost is when you compare Ceph or Gluster on the latest hardware vs something like NetApp. Again, it’s not all about pure cost, so maybe the extra features are worth it to you once you see the full comparison.
I think you’ll find we’re still very relevant in the enterprise OpenStack space. The OpenStack deployments may be on free or paid distributions of Linux, but enterprise folks get antsy when you start messing with storage. Most folks don’t realize just how important good storage is. As an extreme example, there are enterprise storage players out there that will guarantee zero data loss. If that is priority 1 or 2, then cost might drop out of the top ten, but most certainly out of the top 5. Most of the time though, enterprise customers aren’t looking for “cheap” or “free”, they’re looking for features and value. I promise I’m not blowing smoke up anyone’s trousers here.
So how is NetApp involved with OpenStack? When the OpenStack Foundation opened it’s doors, NetApp joined as a Gold Member immediately. We’ve been actively contributing code since then, and some of that code has come in the form of drivers for NetApp storage. We’ve had drivers upstream since “Folsum”, and continue to code and create OpenStack-based solutions. Most recently, we completed certification of those NetApp drivers for use with Red Hat Enterprise Linux OpenStack Platform. And there are currently some larger reference architectures and solutions being built and tested around RHEL-OSP and NetApp.
This still begs one more question.. Why am I involved in OpenStack. That’s actually the easist to answer – because it interests me and because over 70% of OpenStack deployments are on KVM, my favorite hypervisor.
Hope this helps,