|
||
|
|
|
              |
|
 |
 |
||||
 |
|
|||
By William Van Winkle |
||||
SOMEWHERE BETWEEN “SHOULD have” and “did” you’ll find a lot of flawed preconceptions. One of my jobs here at RAM is to try and spot industry opportunities in their early stages and present them to you readers before it’s too late. This month’s cover story on video surveillance is a good example. Obvious stuff, right? But are you doing it? Two years from now, will you be able to say “did” instead of “should have”? I think another hot play right now is in blades. It’s common sense, yes? We all know that the inevitable trend with growing businesses is that they need more computing resources packed into less space. A company that starts off needing two servers may well need six or more down the road. You could have six servers in tower cases taking up floor space, or you could put six servers in a standard 42U rack and leave room for more expansion without sacrificing square feet. For a growing company, that seems pretty obvious to me. Rack servers today typically come in 1U, 2U, and/or 4U form factors. In particular, we’ve given Intel’s 2U “McKay Creek” server, officially known as the SSR212MC2, a lot of recent (and deserved) attention, holding it up as a model of compact efficiency. So is it right to presume that these svelte servers are the de facto choice for compute density and scalability going forward in growing companies? Not according to Chuck Orcutt, blade believer and server business development manager with distributor Seneca Data. Four years ago, Orcutt watched the birth of blade servers in the market and realized that blades were bound to be the future. The growing separation of storage from compute resources made something like blades the most sensible approach. Last year, a paper by Server Technology, Inc. dittoed Orcutt’s thoughts, claiming that blades already accounted for 7% of the server market and expected that “they will be the fastest growing server form factor through 2009.” In February of this year, a survey by TheInfoPro found that 66% of respondents stated that blades would comprise more than half of the new server units they would purchase through 2009. At their most basic, blades are ultra-thin servers, more svelte even than a 1U system, mounted vertically into a blade enclosure. A chassis such as Intel’s SBCE sports 14 hot-plug blade bays in a 7U-high form factor. Not only does this double the number of servers possible in 7U of rack space, the enclosure allows for the consolidation of several potentially redundant components, especially power supplies. This also means having fewer moving parts susceptible to failure; a typical server rack will have at least a couple of case fans per server while a blade enclosure simplifies this down to a pair of large blowers. Early on, enterprises were the chief adopters of blade servers, but Orcutt knew the same as everybody else that what starts in the enterprise usually filters down to SMBs. So he began to watch, learn, and dabble. Today, blades are one of Seneca Data’s top growth priorities. As an example of a blade server sale, Orcutt points to a recent job that entailed building a co-located Linux server array tied to a direct-attach storage box. The storage box housed tens of terabytes of customer data being used for online bill payment, so a combination of scalability, absolute reliability, and heavy duty compute resources to accommodate massive access and transfer bandwidth were imperative. Seneca Data went with an Intel SBCE chassis stocked with three SBXD132 blades, leaving 11 slots free for future expansion. Each SBXD132 received two quad-core Xeon 5345 CPUs and 4GB of memory, plus the configuration used Intel’s 4 Gb Fibre Channel adapter. “This solution gave them a highperforming system to be able to do all the computations they needed,” says Orcutt. “The blade center gave them the reliability. Then the thing we did above and beyond was the validation of the Linux platform, the pre-loading, and the pre-testing. We actually had the blade center on the line with the ability of some of the engineers on the Linux distribution side able to come into our system, make sure all the drivers were working, and all those things. So when it shipped out, it was ready to run.” According to Orcutt, it’s that out-of-the-box experience that separates dedicated blade providers from newbies, and it’s one of the reasons a reseller would want to partner with Seneca Data for blade sales and deployments. Blades are not your typical whitebox, if only because they use 208V power rather than the usual 110V. The distributor offers its blade resellers an Intel-authorized, one-day training seminar to get them fluent with the hardware. Even if you’re a reseller who prides himself on handling a job from start to finish, having a field-experienced partner at your back, at least for the first few blade sales, makes a lot of sense. That all said, blades aren’t always the best solution. In general, anything less than a five-server implementation needs a compelling reason to use the blade model; otherwise blades aren’t cost-effective. In the case of Seneca Data’s co-location sale, the blade box never reached break-even from a hardware perspective, but the secondary benefits to the client, including reliability, the Ethernet switch, and the ability to do a KVM remotely, made the premium more palatable. After going through this excursion with Orcutt, I was all jazzed about blade servers and was preparing to write a story on Intel’s SBCE chassis and its associated blades...when I got inside word from Intel that the product is getting axed. Excuse me—axed?! Could it be that my newly formed preconceptions about blades were entirely wrong? Why would Intel kill off a product that seemed to fit so well with a rising trend? Yeah, Dell OEMs the product under its own name, but pulling a channel product to appease a tier-one hardly seems Intel’s style. Everybody I asked either knew nothing or said nothing, until finally I heard one whisper that set me on a new path: “Fox Cove”. As of this writing, Googling Fox Cove yields one hit tied to this summer’s Intel Channel Conference: “With the upcoming Multi-flex Server and New Intel 4-way Server System S7000FC4UR (Fox Cove) planned to launch Q3’07, the blade and MP system offering is rounded out.” So Fox Cove would appear to be the long-awaited transition for the Xeon 7000 platform to Core microarchitecture, meaning we’ll be able to fit 16 cores on a 4P motherboard. But what’s this Multiflex blade stuff? A little more digging led to the code name “Clear Bay,” which was shown at an analyst meeting in early May. Clear Bay is a modular server platform harnessing six Xeon CPUs, networking, storage, and hot swapping. When I asked my anonymous whisperer if Clear Bay would be like an SBCE blade arrangement with a bunch of hot-swap drives arrayed on top, he said, “Something like that.” Clear Bay will be aimed at SMBs and will feature the ability to hot swap CPUs. This is a nifty feature, but I can’t remember the last time I found myself thinking, “Man, I really need to swap that processor without rebooting!” More preconceptions, apparently. The few details we have about Clear Bay makes good sense. Intel is looking for more channel ties, abandoning the entire blade market would be dumb, and it’s obvious that the tier-ones have most of the enterprise blade business wrapped up tight. Let’s take this one more step. If SMB servers can go the blade route, why not desktops? I never would have had this thought if not for a recent discussion with Tech Data about market trends, and one of the hot topics flying around meeting tables now is “PC blades.” IBM and HP are both in this business, but if you want something more channel-friendly you need to go to the company that debuted PC blades back in the early part of this decade, ClearCube, which has nothing to do with Clear Bay. ClearCube uses a rackmount chassis stocked with blades based on Pentium 4, dual-core Xeon (NetBurst), or Core 2 Duo processors. Each blade also features memory, graphics (integrated or Quadro NVS 285), 10/100 or Gigabit LAN, and everything else needed to constitute a bona fide PC. Out by the user where a PC would normally sit there’s a User Port, sort of a hand-sized cross between a KVM and a converter box. The original User Port design is the C/Port, which uses analog technology to transmit the video over CAT5. The C/Port uses one pair of the four sets of CAT5 wires to send the keyboard, mouse, and USB redirection down the wire up to a distance of 200 meters. The newer I/Port converts this analog data into IP packets and, depending on the model, can support up to four monitors. “Largely, PC blade people are solving environmental problems,” says Tom Josefy, ClearCube’s director of product management. “Like in a hospital operating room, you want to have PC capabilities in there but not something that’s running a fan and kicking out pathogens. You know what a PC looks like when you open it up. Having a C/Port, which has no moving parts and is something you can seal in a plastic bag, is a very good solution. With the I/Port, people can get the same kind of solution but without the distance limit.” Through cooperation with VMware, ClearCube is developing more options to take advantage of virtualization in multi-core blades, sometimes resulting in per-seat costs under $1,000. Given that most knowledge workers only use a fraction of their PC’s capacity, a virturalized blade makes good sense. There’s 99.9% uptime, better environmental and power considerations, no desktop noise, and, thanks to ClearCube’s very robust Sentry management software, the ability to monitor and direct just about any facet of the blade deployment. Most people still blow off the idea of a PC blade. But my wife has a $1,500 laptop on the kitchen counter that gets used solely for Office 2003 and Web surfing. If I could shave 30% off the price, save counter space, and be able to monitor anything my kids were doing online, I’d say that makes sense. Blades in the home? Why not? Given that ClearCube is now revamping its design to accommodate off-the-shelf motherboards, we may soon see SOHO blade systems under $800 per station. By the same token, how many offices suffer from component envy? One guy gets a faster CPU or a cooler case and soon everybody has to get an upgrade. Blades with User Ports eliminate this problem because nobody has a clue what their back-end resources really are. So yes, I think blades may be next big thing at a time when everyone is clueless on how to staunch the bleeding of desktop PC sales. Not today, not tomorrow, but maybe by the end of the decade, particularly if companies like ASUS and Supermicro decide to jump in. It’s too easy to discount the possibility for greatness here because it’s so different from the norm. We presume that blades are inherent to enterprises, and they don’t have to be. They only have to make sense, and you need to go out on a bit of a limb to give them a chance. The conditions are right. The analysts are pointing the way. You only need to do instead of say you should have done. |
||||
 Back to top |
||||
Copyright © 2007 RAM Magazine. All rights reserved.
Do not duplicate or redistribute in any form. |
||||
