All your eggs in one availability zone? Tsk, Tsk…

So this morning the big news is that AWS is having issues affecting customers in US-EAST-1.  So far I’ve seen 4sq, reddit, godaddy, quora and many others on the “is down” list.  What always surprises me when this happens is that people point fingers at AWS, and I always shake my head.

If your business relies on a website to be up, why do you allow a failure in a single availability zone to shut down your business?  There are so many tools out there at this point to simplify deployment, scaling and resiliency across multiple availability zones or even across multiple cloud providers – frankly, you have no excuse.  Quora I can maybe excuse at this point … still fairly new, still working on features and functionality (and user retention, but that’s a discussion for another post), but reddit or 4sq?  Really?

Diversify yourself across multiple availability zones, and even better, across multiple providers.  You’ll sleep better at night and will reduce the chance of showing up on the “is down” list with angry users to answer to.

————–

Follow Scott Sanchez on twitter: http://twitter.com/scottsanchez

Notice: This article was originally posted at http://www.CloudNod.com by Scott Sanchez and is his personal opinion.  Copyright 2011 Scott Sanchez, All Rights Reserved.

 

3 Responses to “All your eggs in one availability zone? Tsk, Tsk…”

  1. [...] I blogged yesterday, many of the higher profile sites that were down yesterday simply chose to ignore the options [...]

  2. Jigar Doshi says:

    Elaborate on the tools please :)

    The main issue with these social networks being on AWS is that they end up being dependent on some master db which for performance reasons may be in 1 AZ. The websites are completely memcache driven but the DB is still important and I do not know if they can replicate that at the performance levels they need.

  3. Of course, well-run, big sites already do something along these lines. The problem is, getting these things right is expensive (in engineering costs), and requires hard-to-find talent. When you’re Netflix you can’t afford not to do this, but it’s hardly surprising that so many internet darlings have single points of failure.

    The one that surprised me most was Heroku. A PaaS provider really should have the know-how and desire to be ready for an event like this.

Leave a Reply



Get Adobe Flash playerPlugin by wpburn.com wordpress themes