5 Comments

Summary:

InterNAP, the data center company which has been around for a while has been hit with a strange sort of a curse – curse of power outages. A few weeks ago, a power outage in LA caused one of their data centers to hit the deck, […]

InterNAP, the data center company which has been around for a while has been hit with a strange sort of a curse – curse of power outages. A few weeks ago, a power outage in LA caused one of their data centers to hit the deck, taking along with it MySpace, the social networking giant. John Cook of SeattlePI points to a story about InterNAP being hit by another power outage, this time in Seattle’s Fisher Plaza facility. One more such outage, and it will seem like a trend. Maybe InterNAP needs some divine intervention to prevent that from happening! On a more serious note, these two incidents go on to show the fragility of the modern web infrastructure!

  1. So, the above market prices I pay for bandwidth and “route optimization” will not do me any good when the lights go out? Seems like for the more than 2X price I pay for bandwidth I should get a better hosting facility! Equinix anyone?

    Share
  2. Apart from Equinix, what are the other options? Are there any enterprise class data centers out there?

    Share
  3. There are plenty it all depends in what you are looking for and where. If you are looking for cheap dedicated servers then look at:

    http://www.softlayer.com
    http://www.voxrox.com
    http://www.steadfast.net

    If you are looking for enterprise class hosting check out:

    http://www.rackspace.com
    http://www.datapipe.com
    http://www.savvis.net

    If you are looking for great colocation:

    http://www.colo4dallas.com
    http://www.choopa.net

    For carrier-neutral dc’s you’ll need to find the ones in your area.

    Share
  4. Fragility? IMHO, if anything, it merely exposes expectations of uptime that are too high.

    Share
  5. If this had happened to a large ISP, it would be concerning. However, InterNAP isn’t really an ISP so much as they are a large bandwidth aggregator.

    At any carrer hotel you go to, you’ll find a number of small bandwidth aggregators. The model is they’ll usually carry one or two “Tier 1″ (a term that is relatively meaningless anymore) like UUNet or Level3 providers, and then some lower-end providers like Wiltel or Cogent. They’ll setup BGP, and resell their mix of bandwidth. This is what Internap does, but in a much larger scale.

    Internap’s model is such that they are in dozens of different carrier-neutral datacenters, reselling all of the “high-end” IP providers that serve that facility. Each datacenter, or in Internap’s lingo, PNAP, is a separate Autonomous System.

    Internap registers a separate ASn (Autonomous Systems Number) from each datacenter. This limits the impact on their global customer base. A power-outage at One Wilshire might affect all of their LA customers, but none of their Ashburn customers.

    The real problem, is the skyrocketing costs of real estate and IP transit at west coast datacenters. For the most part, the power infrastructures of these carrier hotels are over provisioned, and under maintained. The mentality of modern datacenter providers could be likened to that of a used-car salesman, or a slumlord. Very little thought is put into their customer’s needs in two years, instead all focus is set on meeting today’s sales quotas.

    Lesson? No different. Build your infrastructure over multiple datacenters, which have diverse, non-shared infrastructures for power and connectivity.

    Share

Comments have been disabled for this post