2 Comments

Summary:

The popular Heroku platform as a service had a pretty rough night last night. Taken down by Thursday night’s Amazon Web Services disruption, Heroku took a tumble. But, shouldn’t a tech-savvy company that follows Amazon’s best practices have avoided this mess?

4482434896_78bda4b52d_z

Heroku, the popular platform as a service, had a pretty rough time last night. Impacted by the Amazon Web Services disruption at Amazon’s U.S. East data center, Heroku took a tumble as outlined on the company’s status page and was offline for a few hours.

Beneath all the trash talk about the risks of using the public cloud, the sites affected by the AWS outage were those that deployed heavily on one Amazon region (U.S. East in Ashburn, Va.)  That over-reliance is a practice that Amazon expressly advises against, as several GigaOM readers commented on our previous story. Smart companies hedge their bets by deploying across AWS regions just as smart companies plan for redundancy when they deploy in their own data centers.

Presumably, that’s what a company as tech-savvy as Heroku would do and to be fair, we don’t know for sure that it did not. But we do know the platform went down for a signficant amount of time and it’ s not like this hasn’t happened before. To its credit, Heroku was transparent about its issues, posting updates on its status web site and the Heroku Status Twitter feed. I’ve put in calls to Heroku for comment and will update this post when they respond.

Cloud ripple effect

Last night’s widely publicized outage will likely have repercussions  beyond Amazon and Heroku. Just as major businesses are getting over their cloud phobias, events like this send them back into the bunker. And it’s silly when you think about it. If companies were totally honest about what goes on in their internal data centers, the whole cloud comparison gets a lot more attractive. Company data centers experience power outages and other issues all the time, but  they’re not likely to acknowledge that fact.

Said one GigaOM guest commenter:

I remember supporting offices that could have their in-house business critical [database] down for days on end ‘because shush, the experts are working’ but they’d crawl up my [butt] if hosted mail went away for 2 hours on a Sunday. This isn’t much different. Manage expectations, people.

Michael Skok, general partner at North Bridge Venture Partners, a VC that invests in a lot of cloud infrastructure companies, agreed that people need to get a grip.

“It’s always a big deal when one of these outages happen because we’re early in the market and people are looking for reasons not to jump. But if you could look at what goes on in Fortune 1000 data centers all the time you’d be scared. The difference is Amazon outages are highly public, and interconnected,” he said.

One legitimate concern for IT buyers, is the lack of interoperability between the various cloud platforms that would help mitigate risk. “Right now, it’s not easy to have work interoperate between cloud providers, so most companies optimize for Amazon or Rackspace or whatever,” Skok said. “Once multi cloud deployments get easier, companies [moving to cloud]  will have an easier time going forward.”

In the meantime, the AWS-Heroku meltdown will likely be used by some companies as an excuse to defer corporate cloud deployments further.

Feature photo courtesy of Flickr user loop_oh

  1. Elaad Teuerstein Sunday, June 17, 2012

    “Once multi cloud deployments get easier, companies [moving to cloud] will have an easier time going forward.”

    Multi cloud deployment is easy. Check out http://www.cloudifysource.com.
    Use the same automation “recipe” to deploy applications on different clouds.
    Optionally use the multi-site data replication module to sync the app’s data across the different sites.

    Share
  2. I agree with Elaad – I just don’t think there have been tools to date that properly deal with multi-cloud, Cloudify does look promising.

    Also, per Nati Shalom – a lot of Disaster Recovery tools were built for a pre-cloud era, and didn’t provide the proper means to deal with cloud outages.

    Check out this webinar today that discusses multi-site replication, especially from a DRP perspective.
    http://bit.ly/IM0w9F

    Share

Comments have been disabled for this post