Blade servers: the fine print matters

commentary The category of blade servers has matured so quickly that the major blade makers -- IBM, HP, Dell, RLX -- are finding little room for differentiation.For the most part, they're all singing the same tune: blades offer more servers per square inch, save power, wipe out cabling nests, make for great server consolidation targets, and are easier to manage because of centralised management tools.
Written by David Berlind, Inactive
commentary The category of blade servers has matured so quickly that the major blade makers -- IBM, HP, Dell, RLX -- are finding little room for differentiation.

For the most part, they're all singing the same tune: blades offer more servers per square inch, save power, wipe out cabling nests, make for great server consolidation targets, and are easier to manage because of centralised management tools. Blade vendors also like to talk about how blades can help facilitate the on-demand dream since they offer re-provisioning flexibility of idyllic proportions.

In fact, so nuanced are the innovations in the blade space that you have to read the fine print to understand where the innovations lie, who's ahead, and who's behind.

I was reminded of this during a recent news briefing with Douglas Erwin, chief executive of RLX , the company that invented the category and is now "the little blade company that could" -- could get bought or squashed, that is. RLX is looking to win the hearts and minds of blade buyers who have IBM, HP and perhaps Dell on their short lists. (Though it has a blade offering, Dell has been relatively quiet on the blade front lately.) Erwin claims that even though IBM and HP dominate the market, RLX's constant pushing of the blade innovation envelope means that you could be selling yourself short by opting for the bigger but slower movers.

Indeed, RLX reminds me of Sequent and Tandem and, to a lesser extent, ALR -- three companies that during the 1990s catered to buyers seeking higher performance and more availability from their servers. At the time, all three companies offered something that we would eventually expect from an average tier one server offering, and all three were eventually swallowed by those tier one competitors. Looking at RLX's self-proclaimed innovations, I can't help but think that the big guns won't be far behind. Everything RLX claims to be doing differently is as much common sense as it was for Sequent, Tandem, or ALR salespeople to ask if you'd like parallel processing or fault tolerance with your server.

For example, RLX claims that its main differentiator is the newly released version 6G of Control Tower, its server management product. RLX's primary point of attack against IBM and HP has been the fact that each company's management solutions are cobbled together from multiple products (IBM's Tivoli/Director, HP's Insight Manager/Altiris' provisioning tools).

While HP and IBM admit to the componentised nature of their total solutions, each downplays any resulting inconveniences to systems administrators. In contrast, Dell spokesperson Wendy Giever says that manageability is currently the chief inhibitor to blade acceptance -- but not in the way you might think. Referring to a standard-sized, rack-mountable server, Giever said: "Many customers are still sticking with their 1U designs, which means they aren't as interested in the density that blades have to offer. But when you probe deeper, you learn that what they don't want is any additional layers of management beyond what they already have in place. Unfortunately, even if you can add blades to your existing server management solution, you still have to manage the enclosures, which is that extra layer that no one wants."

RLX's thinking appears to be in line with Dell's. Erwin described potential customers this way: "They look at the technology. They see blades and they say 'That's cool, let's go check it out.' And they go and they look and they say 'It's not the form factor that's the problem, it's the cost. It's more expensive than I want. I'm going back to the 1U.' Well, we were losing more business to people going back to 1Us."

To address the 1U issue, RLX has introduced a 1U that can be as easily managed with Control Tower as RLX's blades can. RLX will have some convincing to do before companies already using HP, IBM, or Dell's management solutions will consider taking on yet another systems management product. Giever said that regardless of what is making management more complex -- more devices like enclosures to manage, or a different management product altogether -- mass market resistance to blades is one reason why Dell hasn't recently refreshed its lineup. Giever indicated that Dell is regrouping on its blade strategy but offered no specifics or dates.

Calling 6G a Trojan horse that gets RLX a foot in the door at companies that usually buy IBM or HP, Erwin claims the newest enhancements to Control Tower demonstrate how more management software may eventually mean less work. Citing studies that show how the ratio of FTE (full-time equivalent employee) to servers varies from 1:50 to 1:100, Erwin said that RLX has customers where the ratio is more like 1:800. To overcome the problem of Control Tower becoming an extra ingredient in a customer's already complex management stew, Erwin claimed that RLX's support of standards like Pre-Boot Execution Environment (PXE), Intelligent Platform Management Interface (IPMI), and Simple Network Management Protocol (SNMP) -- all of which are supported by IBM and HP's blade offerings -- might enable customers to use Control Tower in lieu of the native management offerings for servers from those companies.

Erwin conceded, however, that if the benchmark for re-provisioning a blade with an operating system and some applications was nothing more than a click here and a drag there, then RLX wouldn't have much to hang its hat on.

RLX is leading the innovation curve, Erwin noted, by intersecting blade management with unattended autonomics. "When IBM [and HP] talk about provisioning, they don't really mean full provisioning. We think our solution -- where we talk about imaging, installing, updating, tasking, repurposing all at the click of the button or a mouse drag -- is provisioning. What we've now got are customers beginning to say, 'Let's take it to the next step and automate things.' We've got rules in our business that say 'If this and that, then do this.' [Until now], we have not had the opportunity for non-manual intervention for hot-swapping or failover. We'd send them an alert just like everybody else and then it's easy to do -- you point and mouse click and you re-provision another blade or a spare. But now, with the 6G announcement, you can put all of that in the background with automation."

While officials from both IBM and HP concede that RLX can innovate, they argue that size does matter. IBM is quick to tout recently released figures from IDC that show, for the second quarter in a row, IBM as the top blade server vendor in terms of units shipped and revenue. In turn, HP cites a Gartner study that says, "HP again secured the top spot in the worldwide blade server market (revenue and units shipped)." Whose figures are right is beside the point, which is that the two companies have the same degree of command over the blade server market that IBM and BEA have over the J2EE-based application server market.

James Mouton, vice president of HP's platform and industry standard servers division, called RLX a "mom and pop shop", Later, via e-mail, another HP spokesperson dug at RLX, saying "RLX is a start-up company with limited resources that cannot deliver full customer solutions comprising servers, storage, software and services. Its entire business is based on its server blade systems. With limited resources, it can't offer a full portfolio of front-end, mid-tier and back-end systems, nor can RLX offer the full solutions and services that HP can."

"We recently hit the 100,000 blade servers sold market and HP is committed and we are investing and you will see more innovation from us," said Mouton. Evidence of that innovation, added Mouton, is showing up on the hardware side where, later this month, HP is rolling out new blades that raise the current per-enclosure system density maximum from eight dual-processor systems to 16 dual-processor systems. Opteron-based blades are on the way from HP as well.

IBM is also working on the hardware, albeit on the storage front rather than the servers themselves. Earlier this year at Networld + Interop, Big Blue rolled out what I call "Brocade on a blade." IBM and Brocade partnered to pack Brocade's 16-port Fibre Channel switching technology on a blade that fits into IBM's BladeCentres (the name for IBM's chassis). Instead of a nest of cables running from each of the blades to a Fibre Channel switch, all of that traffic has been consolidated into a dedicated channel on BladeCentre's backplane. Instead of taking up a slot that could otherwise be used for a server, the Brocade switch gets inserted into a switch slot at the back of the chassis. According to Tim Dougherty, director for IBM's BladeCentre products, "There are four switch slots in the back so you can have any combination of switches: two Ethernet and two fibre, or four Ethernet." In mentioning Ethernet, Dougherty was referring to a similar partnership with Cisco to offer a cable-less (except for the up or downlinks) Gigabit Ethernet switch with its own dedicated channel in the BladeCentre's backplane.

IBM is the first to offer a Fibre Channel switch, Dougherty claimed. Madhu Matta, Brocade's director of product marketing, agreed. "We work with OEMs like IBM, HP, and Dell," said Matta. "Of those, IBM is first to market with a Fibre Channel switch in their blade server offering."

Refuting Erwin's claims that RLX's autonomic responses are unique, Dougherty said "You can have IBM Director take an automatic action. Suppose blade No. 6 fails in chassis D, [Director can automatically] create an image of that blade over in chassis D on blade No. 7." Dougherty also addresses the innovation front: "Say you've got a memory issue on a specific blade. Because we have predictive failure alerts, you will know there's a problem long before failure. If you let it ride and the blade fails, we have something called LightPath diagnostics. We use indicators that tell us which chassis is having a problem, which blade is having a problem, and I can pull that blade out and push a blue button inside, and a red light illuminates next to the failing component. Nobody is doing that."

The bottom line is that the main selling points of blades aren't changing. Innovation is largely incremental and, where it's happening and where vendors will push you on it, be on the lookout for how a little innovation can force a big change to your systems. Opt for RLX's Control Tower for example, and you may have to revisit your entire server management strategy. Go for IBM's BladeCentre because of its ability to host a Cisco or Brocade switch, and you may end up revisiting your network or storage infrastructures.

Keep an eye on RLX, too. Although IBM and HP were quick to put down RLX based on its size, they also acknowledged that the company was first to market with blades and continues to innovate. If for no other reason, watch RLX to help you set the specs for that next (or first) blade purchase.

Finally, the 1U issue raised earlier by Dell's Giever and RLX's Erwin is no red herring. Staying with a 1U strategy is significantly less expensive than going the blade route. Blade vendors will argue that the savings on other fronts will pay for that differential; but you'll have to determine whether or not that's true for you.

Editorial standards