Using AlwaysOn... Except Not Really - Observations from Ignite and PASS
We’re just back from exhibiting at both PASS and Ignite in the past few weeks. We talked to – albeit briefly – about 2100 people across both shows. We also had some fun with a ScaleArc t-shirt contest – when we spotted folks in our “Don’t get caught with your apps down” shirt, we handed out gift cards! You can see one lucky winner here.
Our conversations revealed a few interesting trends about SQL Server users today:
- adoption of AlwaysOn has grown a lot – a far greater portion of the folks we talked to this year are running SQL Server 2012, 2014, or 2016. And a lot of them have implemented AlwaysOn as part of those deployments. We got a bunch of questions about how we complement SQL Server 2016, too, since we announced support for the updated release. So they’ve
- … but most enterprises still can’t fully use AlwaysOn – when we’d ask what capabilities they were running on AlwaysOn, it was pretty limited. They’re using the replication features, but the real power of AlwaysOn is doing read/write split and sending traffic to readable secondaries. Very, very few of the 1000 or so users running AlwaysOn that we talked to were using readable secondaries. The hitch? They have to recode the app, and most don’t have the time, ability, or risk tolerance. When they heard we do read/write split automatically, with no app changes, their eyes lit up.
- people are struggling to deploy active/active data centers – we also talked to folks about their HA environments. A lot noted that disaster recovery (DR) was losing favor. Rather than “recover” from a failure, most were looking for tips on how to design for continuous availability instead. The database failover portion is particularly tricky – or rather, keeping the application up during that database failover is tricky. Enabling customers to move operations from one data center to another, without app downtime, is something ScaleArc helps customers achieve, and interest in active/active ops has definitely increased this past year.
- maintenance windows are getting increasingly hard to call – we love asking folks about maintenance windows (we enable zero downtime maintenance for our customers, so asking about these windows is like peeking into the past for us). Most noted that Saturday night is the only remaining time – and even then, they said, it’s challenging. Even the companies that aren’t global find the burden of getting approval for application downtime to be very painful.
- use of the cloud is still limited for mission-critical apps – a lot of folks talked about using the cloud for backup and recovery. Several were using it for app development. But very few had mission-critical apps running in the cloud – any cloud. The hesitation in cloud adoption tended to fall in two areas – often together: performance and uptime. One customer noted the instability of server availability made him hesitant to move database workloads to the cloud in particular.
The good news is people are leveraging more capabilities of Modern database architectures. The challenge remains having the application tier take advantage of those more sophisticated capabilities. Being able to share customer success stories from folks including Dell, Microsoft IT, and Nasdaq – how they’re using database load balancing software to take advantage of Modern SQL – made for some great conversations at both PASS and Ignite.comments powered by Disqus