Cloud Availability Management: Know Thy Cloud

On Thursday last week our Twitter handle @Adaptivity barely said a word.  The reason was that our primary link to Twitter, HootSuite, crashed.  It crashed along with more than 100 other sites when Amazon Web Services experienced an outage.  If you had asked me Thursday morning if we had any exposure to Amazon Web Services I probably would have said, “I don’t think so”.  I would have been wrong.

Obviously, losing our ability to access our Twitter client for the better part of a day is not a critical loss.  That’s why we didn’t bother to set up a redundant service or backup our stored tweets or perform any sort of availability management around it.  We simply took our SaaS provider’s word for it that they would be a reliable solution (I should note that with that one exception, they have been).  As a solution architect though, it did make me wonder what the implications are for enterprises that are dealing more and more with cloud services.

The lesson I took away is that, for mission critical applications, you need to know all the building blocks that comprise your cloud service.  Look at the picture above; no matter what level you’re purchasing the services for (IaaS, PaaS or SaaS) somewhere there is software, platforms and infrastructure providing your service.  Only when you understand these building blocks can you get a clear picture of your overall availability.

Specifically you need to know the following.

• Are Your Redundant Applications Redundant? It used to be if I trusted two different merchant banks to process credit cards, I could be relatively certain I’d have virtually no chance of failure (maybe .001 * .001).  Now, I have to be certain that my merchant banks aren’t both running on the same hardware in Amazon’s cloud.  If both companies offer 99.9% availability but have the same 0.01%, I don’t get any added advantage of having two providers.
• Can You Move Your Services? If you’re using cloud services for SaaS, PaaS or IaaS, can you move those services from one service provider to another?  Can you run them concurrently in separate clouds for higher availability?  With PaaS and SaaS in particular, it’s often easy to become locked in to a particular cloud environment, without even realizing your doing it.  Be sure to ask your providers how you could jettison your workloads if necessary.
• Am I Doing All I Can With My Current Provider? Most cloud providers offer disaster recovery and/or high availability solutions for their customers.  Since many moves to the cloud are cost based, it’s not always smart to use these for every application but though they may greatly improve availability.  For example, last week Netflix was able to maintain availability as their IaaS provider, Amazon, crashed because their application is distributed regionally.

The bottom line is that cloud simplifies availability management considerably, but it doesn’t erase the need for it.  There are still availability concerns beyond, “what percent uptime is your cloud?”

Tags: Featured

Category: Cloud Computing