All Out of Bubblegum

In the classic John Carpenter sci-fi film, They Live, the hero, played by wrestler Roddy Piper, comes across a pair of sunglasses that enable him to see the truth: There are space aliens among us, and they are using subliminal messaging and propaganda to turn earthlings into happy, idiotic consumers. Their ultimate goal is to bilk us of our natural and budgetary resources.

At a pivotal moment, Piper walks into a bank with a shotgun on his shoulder and announces to the mix of aliens and humans in attendance, "I have come here to chew bubblegum and kick posterior]. And I'm all out of bubblegum."

This scene from the 1980s feature is very timely, as I had been struggling with how to respond to several announcements that had been made recently by Hewlett-Packard and EMC. I wanted to frame my response within some sort of reasonable context, when the movie made me realize the root of the difficulty I was having. Put simply, I'm all out of bubblegum.

Let's start with HP, which introduced a few weeks ago a server cluster to front-end a storage area network (SAN), and touting it as a network attached storage (NAS)/SAN gateway that will take market share away from the dominant NAS vendor: Network Appliance. Apparently, the HP guys are getting pretty desperate.

But a bundle of Compaq Proliant servers, SuSe Linux and Polyserve Cluster File System does not a NAS head make. At 9,000 operations per second, the thing can't possibly keep up with the latest storage frames in the market.How can HP expect the product to compete with NetApp? Heck, even the NAS head from OnStor, that little-engine-that-could, does 32,000 operations per second. Plus it supports both Microsoft CIFS/SMB and NFS, while HP's box supports only NFS, I'm told.

What smaller company or branch office (HP's stated market for its product) isn't using Microsoft apps to the max? Isn't the lack of support for CIFS going to create an issue there?

Moreover, the product delivers 9K ops/sec using 635 watts of power. Competing products (again, see OnStor) deliver more throughput using the wattage required to power a light bulb. With energy prices messing with everyone's head, isn't this another downside in the value proposition for HP's clustered gateway?

Finally, there is the potential gotcha of cache coherency as the cluster scales. I won't comment too much on this one, since insiders tell me that the product doesn't scale much beyond four nodes. But from where I'm sitting, it looks like just another file server, and a limited one at that, masquerading as a big NAS head.

Hey, HP: Why not leverage your relationship with Hitachi and start developing a next-gen NAS using the 1-terabyte perpendicular disk-drive that Hitachi has been bragging about? Just slap on a little RAID 1/0 and some journaling code on a single drive, then sell that as a plug-and-play NAS for the SME market. Now that would be something to make the NetApp boys quake in their boots! But with respect to this cluster product, however, I'm all out of bubblegum.Next, we turn to EMC, which provided a full week of idiocy, first with its new virtualization platform, called Invista (or should it be In-vesta), then telling anyone who would listen at their customer soire in New Orleans how they had this information lifecycle management (ILM) thing licked.

The Invista virtualization product has been preceded by months of hype, of course. When word first leaked about the product, in response to IBM's announcement of SAN Volume Controller (SVC), EMC framed it not as an alternative, but as The Only True Virtualization Solution in the industry. The vendor posited that EMC logical unit numbers (LUNs) were different from other LUNs and therefore could not be virtualized by a virtualization product from a non-EMC vendor.

Anyone who knows SCSI knows that a LUN is a LUN. EMC was just playing the FUD card to confuse consumers about the capabilities and limitations of IBM's SVC, as they had with FalconStor, DataCore and StoreAge virtualization products a few years earlier. The objective was clearly to discourage customers from buying technology that EMC feared might gut the value proposition of the microcode value-add on the controllers of their overpriced arrays.

Truth be told, EMC could have waited for Cisco Systems to port the IBM SVC to work with its MDS Fibre Channel switches. That implementation was so lame, and created such inefficiency in Fibre Channel fabrics, that it almost single-handedly destroyed IBM's case for its own, server-based, implementation of SVC.

But, of course, EMC didn't know that. So they started with this "Our LUN is better than your LUN" business and assured everyone that they would fix virtualization once and for all with their forthcoming Storage Virtualization Router, which is now known as Invista and scheduled for release this month. I thought I was all out of chewing gum when I read their press release, but things got worse.At the New Orleans customer event, in typical, heavy-handed Hopkintonian fashion, senior droids took the stage one after another and outlined their solution for global data domination. They claimed that they had bought or built all of the components needed to discipline our data and that ILM, for all intents and purposes, was about to become a real product.

EMC's recipe for data lifecycle control? Just buy a few of its overpriced storage frames, add a heaping portion of Invista virtualization, wrap the whole thing in Documentum, and voila: instant ILM.

Dubious is too understated a response. This proposition is silly on just too many levels to do justice to in this limited space. Let me summarize it this way. First, it will take a long time to actually integrate the stuff that EMC has bought over the last 18 to 24 months at more than just the brochure level. Nobody believes the assertion, "We've listed all of the products on the same brochure, hence they are integrated."

Second, if folks wanted to manage their files (the type of unruly data showing the largest growth rate) with a workflow management system, they would have already done it. Heck, Microsoft builds the capability right into Office, and it is the first thing that everyone turns off when they load the product.

Third, it is worth noting that Invista, excludes copy-on-write (COW) functionality. Was EMC worried that COW would usurp the microcode "value-add" in its array controllers and dismantle the entire value proposition of that other software cash cow, SRDF? Or did it believe nobody would notice this obvious functional omission?Finally, EMC pushes VMWare as part of the ILM/utility storage solution. Don't get me wrong: VMWare is good stuff, but neither Microsoft nor its customer base (about 87 percent of the server market) have seemed terribly interested in propping up their Windows servers on a third-party virtualization layer.

All in all, EMC's ILM is starting to sound a lot like EMC WideSky: Full of sound and fury, appealing only to those that have outsourced their thinking processes to Hopkinton.

I haven't. And I'm all out of bubblegum.

What's got you popping your gum? Drop me a line