23 Oct 2012

To be sure - How to avoid AWS US-East 1 downtime

Another AWS EBS outage yesterday evening (UK time), and as before it was flaky old US-East 1. Only Zone C this time, although I experienced a degree of latency whilst helping a client re-launch his webserver in Zone A (praise be once more for AutoSnappy). I'm continaully surprised by the amount of 'cloud aware' businesses who get caught out, however, bad platform design aside, on checking outage history there's an easy option to limit the damage, even if you're running a single instance (and don't use Scalr or haven't designed for for zone redundancy).

Firstly, as mentioned, if you look at the history of outages you'll see the major (biggest/longest) outages have all been in the US-East 1 Region. It's the oldest region so maybe there are related problems with legacy technology? Who knows, but it would seem the sensible money might be to simply avoid using this region. For me as a predominant UK provider with a UK client base the EU (Ireland) Zone is a better choice anyway. Yes it's had outages too, and costs a few cents more, but for increased reliability and reduced latency (to the UK) along with less history for downtime it would seem a better choice.

All well and good in theory, and fine for new setups, but if you (we, me) have existing US-East instances and farms then how to migrate? Single AMI (region) migration is possible via the AWS CLI tools but has a reported history for re-launch issues, and as yet there is no region migration utility in the AWS admin panel (which is a shame as it would be a killer feature).

Thankfully there are plenty of third party management suites available to help us out, with Ylastic being the most attractive solution in this case with (one click?) region migration included amongst its extensive set of AWS admin utilities.

Fig:1 - Ylastic AMI Region migration.

Region migration does bring with it certain issues, for example an EIPs will need changing. Not a huge admin overhead but if you're running a web and/or email server it's something to be aware of with regard to reverse DNS, PTR and SPF records in order to keep things running clean and your server & systems off any spam lists (AWS block EIPs are renowned for this).

Ylastic prices are most reasonable for the extensive range of tools provided, in fact their entry level $25 p/m has to be worth it for the region migration option alone. You could spend days re-creating and re-populating AMIs in alternate regions and how much do you pay your cloud infrastructure guy? I sincerely hope it's more than $25.

Ylastic pricing becomes even more cost effective for multiple AWS accounts. Their base package will facilitate x5 separate accounts and the $50 p/m next level x20, definitely something to consider if you work in a co-op/co-working environment or can refactor spread costings across multiple client support packages.

Check out the Ylastic feature set. Yes, it's quite long, but that's a good thing.

If you would like help or advice with AWS region migrations, or any other aspects of AWS related technology and/or platform design I (we) would be more than happy to hook up with you to discuss options, check out the main Cirronix site for more details - http://cirronix.com.

No comments:

Post a Comment