One of my main apps is hosted on Windows Azure's Websites platform. Given Friday's downtime, I realized that I need something that allows my app to stay online if Azure has an outage again.

My current setup is 1 WAWS instance (w/ autoscaling set to 1..3 instances @ 70% CPU) and SQL Azure as a backend.

I really enjoy having WAWS's automatic git deployment feature and would rather keep it (Web Roles can't autodeploy easily afaik), but what other solutions could I use to reach HA + lower latencies for most of my clients?

I've thought about the following scenarios:

  • Keep it as is
  • Generate additional regions using the same autodeploy scripts [but I'd have to mirror my SQL Azure DBs somehow)
  • Move to Web Roles/VMs w/ traffic manager and self-host my DBs [either keeping SQL Server or moving to MariaDB]
  • Move to Amazon, Rackspace or whatever allows me to keep the uptime/performance I want

What would you suggest?


First: Remember that all cloud providers have outages. Windows Azure did not "go to hell" as you put it. A subset of services, in a specific region, had an outage until it was mitigated. Changing cloud providers will not protect you from outages. Just bing it, and you'll find plenty of outage details across multiple cloud providers.

High Availability of a running service always comes with a cost, and priorities will be app-specific. If it's the web tier, then you may indeed want to consider hosting in multiple geo's. If it's a backend processing tier, sometimes it's "ok" to let a service go offline, as long as requests are queued up. If it's the storage system (preventing queueing of messages), perhaps an alternate queue in a different data center could be available for redundancy purposes.

Unfortunately there's no one single right answer, as it's going to be app-specific.

  • All my services over at east-us went down. Had clients on hold for 2 hours - I'd like to have something in-place to avoid this – hb. Aug 26 '13 at 2:38
  • I completely understand. However: The cloud doesn't give you this automatically - this needs to be built into your app. And... as I stated before, all clouds have periodic outages. Look at today's news on Instagram, Vine, Heroku and other outages related to another cloud provider's data center outage. – David Makogon Aug 26 '13 at 3:30
  • 1
    Yeah, that was lovely. Being an Azure architect, what would your suggestion be? – hb. Aug 26 '13 at 5:11

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Not the answer you're looking for? Browse other questions tagged or ask your own question.