X-IO, born in 2007 and based in Colorado Springs (CO), is the result of the fusion of XioTech and the Advanced Storage Architecture division of Seagate (yes, the Hard Disk vendor). It’s not a proper startup actually (XioTech was born in 1995) but it does have most of the characteristics, of which the most important seems to be innovation.

In a world of Software-defined everything they have chosen just the opposite direction with a product that relies on a strong hardware architecture and few (if any) software frills… but it doesn’t mean that it isn’t a good solution for your software-defined strategy.

Just a storage brick

X-IO calls it ISE (Intelligent Storage Element) and that’s it. We are talking about a 3U chassis with two IO controllers and a SAS backplane capable of supporting up to 40 disks. Disks can be SSD or SAS (2.5”/10K RPM) and are organized in two sealed packs (yes, sealed! more on this later).

SW_CAD_ISE_LabeledOne hybrid Brick ISE (10 SSD+30 SAS) is capable of 60.000 IOPS (real-world OLTP workloads), which is good, but the best comes more from the predictability of the performance than from the peak. In fact, X-IO implements a fast automated tiering mechanism which is capable of discovering hot data regions every 5 seconds and promotes them on Flash for faster access. The dimension of each data region is not that small actually (244MB+metadata) but David Gustavsson (VP of engineering at X-IO) told us that this is the best compromise they have found to use ISE resources at best (and it seems to be true, looking at the measured performances shown in the slides). The following video explains it very well:

ISE offers few configuration options (for example only RAID10 and 5 are supported) and most of the tunings are automatic or non existent. The Array was thought up from the ground up to give the best compromise in terms of efficiency in relation to ease of use.

The maintenance of a brick

One of the best features of ISE comes from the self healing mechanism implemented by X-IO. As mentioned above, each ISE has two sealed DataPacs (20 disks each): the end user can’t access the disks…and there are no spare disks either.

The system reserves 20% of the total disk space for failed media recovery. In practice, thanks to the relationship with Seagate, X-IO has access to disk firmwares and can tweak them to do the job that is usually done in a lab to repair/recover a disk. That’s brilliant!
The process is automatic and doesn’t affect performance: in case of a disk failure, that part goes offline and data are re-protected immediately (it doesn’t take long because of the wide striping: all the disks concurrently contribute to all operations). Immediately after that, ISE starts to look at the damage and makes the necessary operations to bring the disk back to life.
HD blueThere are many possible failures in a disk (heads, platters, blocks, and so on) and ISE can isolate and remanufacture the drive. Also when the drive can’t be recovered at its full capacity/performance it can be partially recovered by offlining part of it.

All this internal cleverness makes the product really resilient and reliable, which is why X-IO sells it with a 5 year warranty. Yes, 5 years… and without manual intervention! (in some cases X-IO can swap the whole brick if the number of failures exceed the maximum, but it’s a rare occurrence)

Robust bricks are good for solid infrastructures

Trade features for hardware reliability and durability are not the main trend, especially now that most of the startups use off the shelf hardware to build their feature-rich products.
X-IO chose a different path since its birth, and even though at the time it was too early, now this approach could be much more interesting.
In fact, coupling X-IO ISEs with modern software solutions (Like VSAs or virtualization products) could make a lot of sense from the TCO standpoint.

I think that the bigger the infrastructure, the more the advantage of ISE will be visible, especially because XIO provides management REST APIs (soon to be available with Openstack compatibility via Cinder).

Why it is important

I like the elegance of X-IO ISE and, even though in the past I was dubious about the lack of software features, now this product makes much more sense because software features are migrating to an upper layer of the stack.
Performance predictability, resiliency and durability are all important features and they all lead to an overall better efficiency of the infrastructure.
In the right scenario this kind of system can effectively contribute to a better TCO, especially in big infrastructures.

I strongly recommend you watch the videos recorded during Storage Field Day 5. They are very educational and give a clear idea of the features of this product.

Disclaimer: I was invited to this meeting by Gestalt IT and they paid for travel and accommodation, I have not been compensated for my time and am not obliged to blog. Furthermore, the content is not reviewed, approved or published by any other person than the Juku team.