0.014: beware of blogketing
Hu Yoshida has done another bang-up job of explaining the utilization challenge bearing down on customer budgets in his latest 2-part bloguturial (my contraction of "blogger's tutorial"). In the first part, he does a great job of outlining the problem, and in part 2, he credibly describes a vision for a solution. In fact, if you removed all the references to Hitachi products, those two blog entries should probably be mandatory reading for every storage architect in both mid-tier and enterprise IT shops (and probably their suppliers as well).
It could have been that good - if only Hu had separated fact from vision.
Written in the style of a print-media advertorial (a contraction of "editorial advertising," where ad space is purchased to run an editorial-like assessment of a product or service), the blogutorial is emerging as a new, and so-far unchecked marketing tool. Importantly, truth-in-advertising laws require published advertorials to be plainly marked as "paid advertisements" (usually in tiny font in the header where you might not notice it). Apparently this isn't so for blogketing ("blog marketing")- even when done by official corporate spokespersons in the public forum of the Internet. Apparently the blog-world is unfettered by truth-in-advertising rules.
This concerns me, because it allows bloggers to act like the snake-oil vendors of the past, getting away unchecked with unsubstantiated claims and misleading inferences of undeliverable capabilities.
Hu's blogutorial is a prime case in point. The way he litters the product references throughout the two articles would lead the uninformed reader to believe that Hitachi's newly-announced (and yet to ship) Dynamic Provisioning solves all the problems Hu describes. And in fact, this is clearly his intent - the two blogs are nothing more than thinly veiled marketing collateral written by an authoritative industry expert and delivered in the latest marketing medium, the Corporate Blog.
But the complete lack of any disclaimers begs for someone to call "foul."
Enter the storage anarchist...
some of the facts behind hitachi's dynamic provisioning
Announced back in May 2007, Hitachi has since been eager to promote the latest addition to their family of enterprise storage products at every opportunity. Hu and other spokespersons spent a lot of time in the initial days answering questions from the press and analysts about their implementation, and their sales teams have since presented dynamic provisioning to virtually every EMC customer who would listen. I've spent a lot of time following all this, and I've had the opportunity to discuss my "catch-22's" (see here and here) with customers and prospects who were swayed by the Hitachi sales pitch (not surprisingly, nobody has told me that Hitachi has effectively resolved any of these gotchas).
And through all the hype a rather stark reality has emerged, and one that is quite different, if not outright contradictory, to the scenario that Hu paints. Follows are several facts that I've read, heard and/or seen attributed to official Hitachi spokespersons about Hitachi's Dynamic Provisioning:
- It is only available on USP-V (and the rebranded equivalents sold by HP and Sun). If you want it, you have to buy new hardware - for whatever reason, dynamic provisioning is not available as a software upgrade to existing USP (and equivalent) installations. So much for investment protection!
- It is not supported on externally virtualized storage - at least, not yet. So, if you want to thin provision your USP-V, it can only be done using the expensive, tier 1 fibre channel storage installed inside a USP-V, because the USP-V flat-out doesn't support low-cost Fibre Channel or SATA drives within the array, and dynamic provisioning doesn't support virtualized storage outside of the array.
- You can't use USP-V dynamic devices as either a source or a target for a local replica. This one is particularly odd, since Hu's blogutorial makes such a big point about replicas being the most significant overhead in underutilized storage. But several sources have verified that local replication isn't included in the initial release.
- You also can't use USP-V dynamic devices as either a source or a target for a remote replica, whether on another USP-V or on an older USP. So, although your limited to using dynamic devices to your expensive tier 1 storage applications, you can't provide any remote BC/DR services for them. Seems counter-intuitive - if it's tier 1, it's important enough to protect, whether or not it's thinly provisioned!
- Dynamic devices aren't free - pricing scales with the amount of physical storage assigned to the supporting storage pool. And I've been told that the pricing effectively doubles the $/GB for the physical storage assigned to the pool. So unless you more than double your utilization, you won't actually save that much money (on the acquisition side, at least).
- Operational management of dynamic provisioning is virtually non-existent. Customers tell us that they expect robust planning, monitoring, alerting, disaster-prevention and recovery tools when they consider thin provisioning. Several have acknowledged that Hitachi's implementation falls far short of their basic requirements in this area, even before you compare their implementation to other more-established alternatives. At least one customer has even asked us to add support for Hitachi's dynamic provisioning to ControlCenter - says something about their faith in Hitachi to finish the product, I guess.
You can't de-dupe Hitachi's Dynamic Provisioning devices in place, you can't use Shadow Image or Universal Replicator to copy them to remote sites, you can't use them across multiple tiers of storage, and I don't even think they're supported on RAID 6 devices. Most significantly, you can't use Dynamic Provisioning at all unless you agree to be locked into Hitachi storage for the next 3-5 years.
The good news about all this? Well, let's just say that it sure helps to ensure that we get the customer requirements and priorities right !
yes, i know, it's only version 1.0
OK. I'm not bashing the fact that the product isn't complete yet - adding major new functionality is always a challenge, especially if time-to-market is a key part of your strategy. All of us vendors struggle for the right balance of features and value vs. timeliness. I'd like to say that EMC's strategy is to get the product complete before we announce and ship, and I think we did a pretty good job with the Security features in 5772. But we've not always meet the objective as well as we (or our customers) would have liked.
And I'm sure that Hitachi will eventually finish the product, adding local and remote replication, thin replicas and all that other stuff that Hu describes. Maybe they won't ever put low-cost storage devices inside a USP-V, but sooner or later they'll have to bite the bullet and support externally virtualized storage behind dynamic provisioning - even if doing so does in fact increase exponentially the risk of total data loss should any bit of the dynamic storage pool become corrupted. Undoubtedly, there will be a Version 2.
My beef is with the blogutorial itself, and the misleading implications that it presents - without any disclaimer. Sure, it's written by a Hitachi Official Corporate Blogger, and it's posted on the Official Hitachi Data Systems web site, so readers should approach with a certain amount of caution and skepticism. But I think this one crossed over the line of responsibility to represent products in a factual manner.
I simply don't think it was appropriate or even necessary to pollute an otherwise excellent article with untruths. In fact, the whole story would have been even more powerful if it included no reference to Hitachi products at all. Just present the case, outline the opportunity, and describe what the "right" solution should look like. Promote the heck out of the article as evidence that you truly understand the problem, then show prospects how your current and future products will fit into that vision.
And if you REALLY want to score some major points, explain how your implementation is going to address the catch-22's I've been discussing.
Or else, please follow the truth-in-advertising guidelines and label your blogs appropriately:
Yet another paid advertisement from Hu Yoshida
Of course then, you will also have to cut out the baloney, and stick to the facts.