Black-box storage: What’s it doing now?

Recently, Dell announced the newest members of its EqualLogic PeerStorage line of virtualized iSCSI arrays: the PS6000XVS and PS6010XVS. Unlike the other PeerStorage arrays (which are either entirely SATA, SAS, or SSD), the XVS models mix eight SSDs and eight 15,000-RPM SAS disks in the same enclosure, giving you tiered storage in a single box.

Mixing two vastly different types of storage media in the same shelf isn’t that unusual. Here’s what’s out of the ordinary: The XVS arrays automatically load-balance heavily used data blocks onto SSD while leaving less-used blocks on the SAS disk. In theory, this is an excellent idea. It should allow the array to adapt very quickly to changing I/O patterns and provide the best performance possible from its mix of very fast, but small SSDs and comparatively slow, but much larger SAS disks.

[ Looking to reduce your storage requirements? See InfoWorld’s comparative review of four storage appliances that use data deduplication technology. | Also check out Matt Prigge’s High-Availability Virtualization Deep Dive. ]

Take, for instance, the task of providing storage to a heavily used Microsoft SQL database. In most cases, you’d combine high-speed disk for the data volumes and, if needed, SSD space for transaction logs and tempdb. The XVS arrays attempt to make these sorts of architectural decisions unnecessary: The array knows best and can put your data on the storage media it deserves to be on.

Or does it?

I don’t know about you, but having unseen software make decisions for me about what data deserves to be where gives me the willies. Sure, I’m all for anything that can make my job easier and yield better overall performance. But giving up control to a load-balancing algorithm or automation engine requires a level of trust I and many others like myself struggle with. And I don’t think such discomfort is limited to storage admins, server admins, or even technologists in general.

As both an aviation and technology enthusiast, I get a certain amount of glee from reading stories about airline pilots who make the transition from largely mechanical aircraft such as the Boeing 737 to the heavily automated Airbus A319/A320. Veteran Airbus captains often sit quietly in the left seat and wait for the newly minted First Officer to ask, “What’s it doing now?”

You can easily see the parallels between the challenges those pilots face and that of old-school storage administrators who must maintain “intelligent” storage devices, and it doesn’t take a genius to conclude this is where we’re headed. With IT budgets shrinking and our data exploding, we can’t really afford to micromanage every disk volume anymore. If you’re one of the “crusty old captains” mentioned in the blog I linked to above, transitioning from a traditional “these disks in this shelf make up this LUN, which gets assigned to this server” SAN to a virtualized SAN where your LUN could be virtually anywhere constitutes a huge leap of faith — one that many of us still don’t want to make.

The evolution from a world where careful thought and detailed design work was necessary for a successful storage implementation to one where the storage sort of figures it out on its own will be difficult for a lot of people. In addition, it’s likely that some attempts to get this automation to work properly will fail in ways that are hard to diagnose and equally hard to correct. I have yet to get my hands on Dell’s XVS line (keep your eyes peeled for a Test Center review in the near future), but I’m willing to bet it comes in a black box.

This article, “Black-box storage: What’s it doing now?,” originally appeared at InfoWorld.com. Read more of Matt Prigge’s Information Overload blog and follow the latest developments in storage at InfoWorld.com.