Crash

Amazon’s S3 utility storage service reportedly suffered a serious outage this past week, one that comes on the heels of a recent Google Calendar crash. I’m a big believer in the utility model, and I’ve been surprised over the past year by how rapidly it’s coming to be embraced, even by larger firms. Nevertheless, utility computing still needs to overcome many hurdles before it really goes mainstream in the business world. Some of those hurdles are technical, but the biggest ones are matters of perception. Corporate box-huggers will need to have a strong sense of confidence in the reliability of computing services before they’ll loosen their grip on their in-house gear and software.

The onus is squarely on the vendors here. As Dan Farber writes, the “utility computing model has [to] prove that it can be more reliable, responsive, secure, lean and cost effective to replace what came before.” That’s true. Outages and other glitches are going to happen, but their frequency needs to be reduced steadily, and when they do happen, vendors need to be clear about what happened and why – and what’s being done to avoid similar problems in the future. The fact is, even if a utility service achieves better reliability overall than the average in-house operation, any downtime will affect a lot of companies and hence will become news – news that damages all utility computing suppliers. Every vendor therefore has a stake in every other vendor’s reliability and in every other vendor’s response to problems. On the computing grid, no company is an island.

On a more dire note, Robert X. Cringely is taking a page from Bob Metcalfe’s book and predicting that 2007 will become “the year the net crashed.” He foresees a meltdown as “video overwhelms the net and we all learn that the broadband ISPs have been selling us something they can’t really deliver.” Metcalfe ended up eating his words, and Cringely probably will as well. Then again, there’s probably more than one reason why Google has been investing billions in building its own shadow Internet.

2 Comments

Filed under Uncategorized

2 Responses to Crash

  1. Even older utilities crash. California’s brownouts are the obvious example.

    While not a great believer in overall move to IT as utility, there are many functions that can be moved to external data centers.

    As UPS and generators can be used to mitigate power outages, similar mechanisms need to be deployed for utility data centers.

  2. The major difference here is that Amazon’s S3 is a single power generator (albeit a highly distributed one within Amazon) whereas as Grid balances demand and supply with multiple power generators.

    A Grid needs transferability between multiple power generators and the existance of multiple power generators. As a customer I’m not concerned where the power comes from, just that is reliable and there.

    So in the case of Amazon’s S3 service – it would only take for example :-

    • the open sourcing of the service
    • provision of the service at many providers e.g. Google, IBM and Microsoft
    • creation of an overall S3 Grid

    and then you would have a utility storage service, which a customer could choose (depending on their requirements) to have data stored by multiple providers.

    Under such a system, you create a competitive utility market between providers and a means of achieving a level of risk mitigation beyond the capability of any one company.

    Sounds far fetched? This is what happened with the formation of the national grid in the UK.