Seeking a Standard Blade I/O

Arthur Cole

How long will it be before we see some commonality among blade networking architectures? Quite a while, judging by the number of proprietary technologies kicking around.


Blade servers have proven themselves in the data center since their introduction at the beginning of the decade. They allow you to pack more processing power per square foot and are more flexible and cost-effective than the mainframes of old. Still, their benefits tend to end at the rack because of their limited I/O capabilities.


A recent report by Gartner summed it up rather succinctly: although blades are the fastest-growing segment of the server industry, the lack of a common interconnect standard makes it difficult to push the technology to its full potential. After all, what good is having rack upon rack of separate servers if their ability to work together is not optimized?


The top blade vendors downplay the seriousness of this state of affairs, saying their own proprietary management systems -- Virtual Connect from HP and Open Fabric Manager from IBM, and now FlexAddress from Dell -- are optimal enough. But this is limited thinking on their part. By backing closed architectures, they not only refuse customers the flexibility they enjoy elsewhere in the data center, but they deny themselves the ability to expand their own customer bases by deterring the expansion of mixed environments.


A side affect of this state of affairs is the large number of third-party vendors teaming up to improve I/O performance both between blades themselves and between blades and storage environments. Earlier this year we saw Verari and XSigo unite their technologies to reduce the amount of storage and network connections needed in the Verari BladeRack environment even while increasing the number of virtual servers it can handle.


And just this week, Adaptec shelled out a cool $41 million for Aristos Logic, a maker of multi-protocol RAID storage ASICs that the company hopes will help it expand its I/O offerings to blades, external storage systems and high-performance desktops. Look for a 6 Gbps serial RAID controller for the company's Unified Serial SATA and SAS line before too long.


It's too bad that as the rest of the enterprise is moving toward unified fabrics and open architectures, the blade environment is stuck in a proprietary mindset. Blades are still a value proposition, but with just a little cooperation between the major vendors, they could be made a lot more valuable.

Add Comment      Leave a comment on this blog post
Aug 28, 2008 11:41 AM Jeff Allen Jeff Allen  says:
Blades are a funny business. In a single box you combine components from companies that are partners as well as competitors. You may be asking a good question, but you are at least a decade ahead of yourself. The blade market is growing like crazy, but it's such a small part of the overall server market today. Standard 19" racks and the common "U" didn't happen 8 years into the switch and router market. Also, who's going to build the "standard" connector? IBM would never use an HP design and vice-versa and neither are going to trust an outside vendor's design as blades are crucial to future success.And what makes you think it's not opimized already? What would make it a "full potential" play? To allow Cisco to build a single switch that works in all blade enclosures? Is that "optimal"? Cisco doesn't build open standard technologies to better their competition - why would HP, Dell and IBM want to do so? You mention Virtual Connect and Open Fabric Manager. Two very different approaches to simplifying datacenter server management. I'd love to see you you guys put those 2 head-to-head in a bake-off and compare the pros and cons of both. Reply

Post a comment





(Maximum characters: 1200). You have 1200 characters left.



Subscribe to our Newsletters

Sign up now and get the best business technology insights direct to your inbox.