Last Friday, we experienced a wicked rake of thunderstorms moving from the Gulf to the Atlantic that nearly took my roof off. A lightning strike took out internet access, cable TV access and wired voice until Monday. Power was intermittent during the same period. This all happened as I enter my fourth week of battling my nicotine addiction, cold turkey.
So, I am in a pretty crabby mood, even if things seem to have been mostly repaired and are stabilizing. Anyway, I have been asked to speak at an event occurring in St. Petersburg, FL in September covering the topic of data protection — a topic with which I think I have some familiarity. The fellow asked for a video clip of me presenting on some subject to post on the event site. Given that I have a face made for radio, I tried to discourage him, but he insisted. So, I went on a hunt for video clips of past talks I have given…to my chagrin.
Why does no one tell me when my collar is turned up, or my tie is hanging sideways, or a big piece of hair is jutting out from the side of my head? I found lots of video of me presenting, but none that I really liked. Again, this reinforces my bad mood.
What I did find was a post I had missed in the Virsto blog, posted on June 27, and penned by CEO Mark Davis. It was an appropriately critical piece to a critical piece I had written in response to an email from Virsto’s PR agency — quoting Mr. Davis in a kind of snarky dismissal of DataCore Software’s SANSymphony-V Version 9 release. Overall, the piece was fair and balanced, but there were a couple of things that don’t quite ring true to me.
Mr. Davis, you began with a confession that a quote attributed to you by your PR flaks was not something you would have said.
You proceed to note how predictable I have become in my postings and writings on virtualization generally and storage virtualization in particular. True enough, and entirely intentional. I repeat myself because the other guys (vendors and their PR flaks) are telling lies repeatedly in order to make them sound true and to confuse or obfuscate the history to which I often refer. (Remember: Gartner now vociferously denies that it ever said that “1 in 10 tapes fail on restore” — despite the fact that their finding has been quoted hundreds of times since that day in 1999 when they said it.) But I digress.
You conceded some points to me, which was nice of you and quite unnecessary. But got my ire up by asking if we could leave out the IBM mainframe from the history you wished to recount detailing the origins and development of storage virtualization. To that, I would need to answer, “No.”
Between AMDAHL and IBM mainframes, we did in fact have workload virtualization licked nearly thirty years before the anklebiters started doing virtualization on x86 tinkertoys. So, let’s leave that in, for now.
I do agree that the term storage hypervisor may have come into the vocabulary from many sources, so I won’t belabor whether it came from DataCore or from the random “analyst who made it up on the spot” in 2007. Frankly, Scarlett, I don’t give…
You want to differentiate between storage hypervisor and storage virtualization. Works for me. Feel free to do that if it floats your boat. Some people also want to distinguish storage hypervisors from file systems. An interesting conversation for another time. From my perspective, storage is already virtualized 10 ways from Sunday before Virsto or DataCore touches it. That’s what Dante referred to as the circles of hell.
What I do think we need to discuss is VMware’s plan to build a Storage Hypervisor into its proprietary kluge…er…stack of microkernels for server virtualization. I would be delighted to learn that you are correct about this, and that I am wrong. You said that I was advancing FUD and that you knew better VMware’s plans as a signatory to its supersecret NDA or something. Cool.
Only, it doesn’t seem to fit the facts. For example, I listened to VMware Engineers on stage at the last VMworld state their clear intention to create their own “storage hypervisor.” I do not regard an accurate reportage of VMware’s stated direction as FUD — though you accuse me of this. Nor do I regard the idea that such a proprietary stack centric hypervisor will create problems in contemporary storage infrastructure. After all, we already have evidence of what will happen with VMware’s introduction of nine proprietary SCSI commands into the communications between their hypervisor and connected storage gear — something they call VAAI. VAAI commands, last I checked, were not approved by the ANSI T-10 Committee which is the sole arbitrator of changes to the SCSI language. VAAI primitives are merely self-serving workarounds for the terrible storage IO bottlenecks that VMware creates.
Again, you say you are “under NDA” with VMware and cannot write what you know about their storage hypervisor play. Why not? Why does VMware want to conceal what it is developing to address its storage problems rather than taking the more sensible path of assuring its users and prospective users that its storage microkernel will work and play well in a heterogeneous server and storage environment? What if I have VMware, Hyper-V, Citrix, Oracle and non-virtualized servers and workloads in a single data center? Wouldn’t I be better served by a storage virtualization/hypervisor solution that was server virtualization/hypervisor stack agnostic?
That’s the point of my paean to DataCore — its server hypervisor vendor, as well as storage vendor, agnosticism. You would receive the same kudos, cheerfully, were you not so closely aligned with the fortunes of a specific hypervisor stack.
I am sorry if you feel that this is unfair, but I am somehow not losing any sleep over it. DataCore would not be the recipient of my praises anymore than Virsto if George/Ziya/Bettye/et al decided only to virtualize EMC hardware, or focused their efforts at supporting only Microsoft Hyper-V.
I hope you understand that I am not singling out Virsto. Your business model is to hitch your wagon to a certain vendor’s kit, and I wish you every success with that. If and when VMware takes over the IT universe, you will make a bundle from your choice. I kind of doubt that this will happen anytime soon.
I recommend to my readers that they adopt standards-based technology first, and in cases where standards are unavailable or not yet widely supported, to go with core technologies that are (1) at the very least agnostic to other components of an ecosystem and that (2) avail themselves of common management. Such as strategy, which avoids the friction of lock-ins and manually intensive management, has worked for me for nearly 30 years.
I wish you every success and hope you will send me an email when you are off NDA with VMware and can share more about their storage hypervisor strategy.