Server SANs And Healthy Paranoia update from March 2014

Chuck Hollis, chief strategist for VMware's storage and application services business unit, has blogged that the now generally available VMware Virtual SAN (VSAN) is designed to appeal to a new class of storage buyers. The people who are going to buy server SANs like VSAN aren’t established storage teams, but virtualization administrators. I wonder whether some of these virtualization admins are paranoid enough to be trusted with storage at scale.

Storage vendors from EMC to Tegile have sold storage to -- or in the case of SMBs, through -- experienced storage people. VSAN, and server SANs in general, aren’t designed to make steely-eyed storage guys turn away from dedicated storage for the promised land of software-defined storage. Instead, they’re designed to give the virtualization admin an alternative to dealing with the storage guys altogether.

I fully understand that many corporate storage teams can be inflexible, raising the cost of storage for use cases such as test, development and tier-two applications to unreasonable levels. After all, storage guys are the most conservative people in most datacenters and they do sometimes go too far.

It is, however, important to remember why the storage guy is so paranoid. While server and network admins all have their own “oh crud” moments, those rarely rise to the level where they're worried not only that they'll lose your job, but that they've screwed up badly enough to put the company out of business. Once you’ve reverted to the working version of the router firmware, the router works just as well as it did before you flubbed the firmware upgrade.

But when you accidentally swap out the wrong drive on a disk array, wiping out several terabytes of data, you have to spend many hours restoring data from backups. And even if that goes flawlessly, you’ve lost anything stored since the last good backup. I for one don’t really trust anyone with my storage unless they have a visceral fear of anything that might cause data loss.

[Find out the real reason disk drives haven't gotten faster than 15K RPM in "The Myth Of The Supersonic Disk Drive."]

Since today's hypervisors allow us to move workloads between hosts, or storage systems, with only a minimal impact on those VMs' performance, virtualization admins routinely vMotion all the virtual machines off a host for hardware maintenance without scheduling downtime or following complex change control procedures. In contrast, even though manufacturers promise non-disruptive controller firmware updates, storage admins always schedule downtime and plan the process of reverting back to the previous version before updating their systems

When we converge storage into the virtualization environment, admins are going to have to realize that when they take a host offline, it will reduce both the performance and resilience of their storage as well. I’m afraid we’ll start seeing horror stories of an admin that sets up two-way replication in VSAN, takes a host offline for maintenance, and then loses data when a single disk fails.

If the shift to server SANs starts with test, development and tier-two applications, the virtualization guys will have enough close calls to realize that when it comes to storage, some paranoia is a good thing. When former Intel CEO Andy Grove famously said “only the paranoid survive,” he of course was talking about business, but I think the line applies even better to storage.

[Howard Marks will be talking about flash and software-defined storage at Interop in a half-day workshop "Deploying SSDs In the Data Center" and a pair of sessions "Using Flash On The Server Side""Using Flash on the Server Side" and "Software-Defined Storage: Reality or BS?"]