I attended Storage Field Day 5 (SFD) a couple of weeks ago; SFD is a part of the Tech Field Day. Before the actual SFD event, we attended a Software-Defined Data Center (SDDC) Symposium. SDDC is a new, sexy buzzword in tech.
One evening I was talking to the guys from X-IO, and we found ourselves asking if hardware-defined storage could still be sexy. Let that sink in for a second. Do the words stability, consistency, and reliability sound exciting to you? Do you really want an exciting data center teetering on the bleeding edge of technology, or would you prefer to have a solid storage system for your applications (software-defined software?) to take advantage of?
I think this is where X-IO comes in. Admittedly, the X-IO folks didn't come in and wow us with their automation tools or with their software-defined story; they talked about the SDDC and about open APIs through which automation can take place. In fact, X-IO embraces cloud and cloud providers, because you absolutely need a solid storage environment in those arenas. However, the X-IO team impressed us by passing around their hardware.
Each node is a 3U box that can scale up to 40 nodes. Everything has dual redundancy with active/active pathing. Each node has what they refer to as super capacitors, which allows up to 8-9 minutes of no data loss on its own. These capacitors are also hot-swappable. This is all pretty cool stuff, but it doesn't necessarily differentiate them from other storage vendors offering scale-out solutions. The way they deal with disks within their systems is where I was really impressed.
David Gustavsson, Engineering VP at X-IO, gave a great technology deep dive on the company's hardware. Check out this video for the full demo.
In many current storage systems, when drives fail they are replaced with brand new drives — even if the entire drive isn't dead. These are generally replaced by a "hot spare," and then a new drive can be purchased and rebuilt at a later date. At X-IO, though, they have a different philosophy: "Not when in doubt, throw it out...but when in doubt check it out." X-IO doesn't use hot spares; they allow for shrinking a drive. If part of the drive doesn't work, X-IO just stops using that part of the drive. So, instead of losing an entire 900 GB drive, you might just use the 820 GB that are still good. Spare capacity is built in so you're also not losing space. All of this results in faster rebuilds. This also changes the failure domain from 40 disks to a possible 240 surfaces, as they would refer to it. There is a possibility of a 20% failure before any replacement actions need to happen. If it ever gets to that point (which is very rare), X-IO does offer a replacement process, which they will perform for the customer (a five-year warranty is offered).
Another really cool aspect of how they handle disks a little differently than other vendors is how they place them within the hardware node itself. Because of X-IO's close relationship with Seagate, the company is able to architect the optimal design for the disk enclosures. Many storage systems use replaceable drives that slide into some sort of enclosure or shelf. X-IO uses what's called a DataPac (shown below).
X-IO has done extensive testing of DataPacs so that vibration and harmonics will cause the least possible harm to your disks. The company has also done extensive thermo testing for heat and cold.
I am not currently using X-IO, so this article is purely opinion based on what I saw in a presentation. However, if you have questions about what I covered, feel free to post them in the discussion, and I'll do my best to answer.
Lauren Malhoit has been in the IT field for over 10 years and has acquired several data center certifications. She's currently a Technology Evangelist for Cisco focusing on ACI and Nexus 9000. She has been writing for a few years for TechRepublic, Tech Pro Research, and VirtualizationAdmin.com. As a Cisco Champion, EMC Elect, VMware vExpert, and PernixPro, Lauren stays involved in the IT community. Lauren has been a delegate for Tech Field Day and has also authored a book called VMware vCenter Operations Manager Essentials.