Microsoft's Own SQL Server Upgrade
Microsoft sponsored a webinar earlier this week to highlight how its answers.microsoft.com site upgraded from SQL Server 2008 to 2014. The services engineer who led the project, Michael Schaeffer, did a fantastic job outlining the challenges and the options for overcoming those issues on the call. Anyone looking to adopt 2014 and AlwaysOn will get some great lessons learned from Michael.
Among the issues Michael highlighted:
- adopting virtualization meant they couldn’t keep using Windows Failover Cluster
- lack of clustering support meant the AlwaysOn AG Listener had no failover
- AlwaysOn does not support any kind of load balancing
- leveraging the read-only secondaries in AlwaysOn would take significant reprogramming of the application
- trying to do SQL load balancing with TCP load balancers would introduce single points of failure
- needed to extend failover to support zero-downtime maintenance to streamline patching and testing
Working at Microsoft, of course, gives Michael the luxury of tapping into the smartest of the smart to determine the optimal architecture for deploying SQL Server. As he investigated options from the SQL Server product group and other infrastructure teams at Microsoft, he learned that many of his peers had investigated ScaleArc to complement their SQL Server deployment.
The group tested ScaleArc and scoped out a deployment that would span the four global data centers supporting answers.microsoft.com. After successful failover and performance testing, the group went live with ScaleArc earlier this year.
Michael went on the recount the successes of the deployment. The ScaleArc solution:
- Accelerated the project timeline – cutting it from nine months down to four
- Provides auto failover across data centers – and avoids app errors during that failover
- Supports automatic read/write split – so the team did not need to make any application changes
- Provides geo-aware load balancing – using time to first byte as the indicator of the best-performing server to direct reads to
- Enables zero-downtime maintenance – letting Microsoft patch on demand
- Supports 8x traffic capacity – boosting the sites overall performance
- Utilizes all deployed SQL Servers – because all servers handle traffic, instead of having idle secondaries, the architecture went from 24 to 8 servers
Anyone looking at a SQL Server 2014 upgrade, and struggling with issues like read/write split and AlwaysOn load balancing, would benefit from hearing Michael’s discussion of the challenges. Tune in for the 15 minute discussion, and stick around for the live ScaleArc demo too!.
comments powered by Disqus