Standard for blade servers and chassis technologies?

One of my company's greatest fears about blade servers is that the chassis technologies might not be able to keep up with the servers themselves. We think this will eventually lead to a situation where we have to decide whether to keep the current chassis (and therefore protect our investment while not getting the latest advantages) or have to upgrade the chassis and therefore negate one of the main advantages proposed by blade servers, which is "modular" growth by having to program downtime for what could be a substantial number of servers in order to perform the upgrade.

The second fear is vendor lock-in, which may or may not be completely allayed by the emergence of an industry-wide standard, given the habit most big players have of flouting them at their convenience.

Are there any answers on the horizon that might help dispel these fears? Is there an up-and-coming standard that everybody is going to follow?

    Requires Free Membership to View

    Login

Backplane speeds are much faster than chip speeds due to wire speed technology (meaning that operations that used to happen in software now occur in the hardware). With the 10GBASE-T standard due to publish in a few months, even 10 gigabit network interface chips are not as fast as the backplane. A good comparison of offerings from each manufacturer should include processing power per chassis. Also, whether or not they support increased speed upgrades. For instance, some may be a field upgrade, others may require the forklift upgrade you discuss. It is very important to broach this subject with each vendor.

Another thing to look at is network capabilities. We have all seen instances where additional network connections provide greater throughput. If you are running data-intensive applications or are providing desktop applications over the blade, you do not want to be limited to only a few network connections without a means to add more to increase throughput. Weighing all features with future considerations in mind will provide the best answer, but with all the various vendors with varying capabilities, this is best approached on a vendor-by-vendor basis. Since you are the end user, they should be happy to do this work for you.

As for vendor lock-in -- I think this is valid for a while. At the BladeSystems Alliance, we have done interoperability booths for some time; however, we don't find a push to have one vendor's blade card fit another vendor's chassis. Interoperability has been more centered around storage and supporting gear. A new standard that may help is IEEE 802.3ap (Ethernet in the backplane). This standard is still in the works, but may provide some interoperability through a less complex backplane. But I think at this point it is too soon to tell what the blade vendors will do with this technology. It will certainly speed things up if we no longer have to convert native Ethernet packets, but beyond that...? Also from an ease of use standpoint, it may be easier to stick with a single vendor. Finger pointing is a reality and no end user likes being stuck in this scenario.

Storage standards have also opened up creating better storage interoperability. IBM did a great thing when they created their open systems storage solutions. Organizations like SNIA also helped. DMTF also has some active support for their SMASH program. See my blog for more info.

This was first published in March 2006

Join the conversationComment

Share
Comments

    Results

    Contribute to the conversation

    All fields are required. Comments will appear at the bottom of the article.
    Back to top