I was talking with someone about their installation of SQL Server and they were bragging about how they were using a cloud-based solution (really doesn’t matter even which one) where the systems were completely managed for them.
From backups to performance to just about every aspect of their SQL Server installation, it was taken care of. It was saving big dollars by using this approach vs. full-time staffing and they felt confident that they had outstanding talent watching over their systems (and they do).
My question comes though when you have an issue – be it some sort of breach to deal with or a programmatic error or even insider issues with your databases. Prevention issues aside, it can and does happen in small and big ways. So, in situations where you’re in an urgent triage and recover situation, is reliance on a cloud provider an issue?
It seems to me it can blunt the knowledge and expertise you have on-staff because worst case, they’re been replaced with the service. Best case, they may have split responsibilities now and feel less of a need, and indeed have less involvement in solutions deployed to these environments. Sure, they know how it all works, but intimate, in-your-face knowledge of the nitty gritty bits for recovery, restore, digging around in logs and such – all of that, along with the intricasies of how the application works, is designed and how it talks to the data platform… I would think all of those pieces and parts could easily be subject to atrophy a bit.
Add to that that your provider will definately be up to speed on the deployment of the dataplatform, but NOT the interfaces and such there, and you’re faced with a possible significant challenge in triaging issues that may come up. In fact, there may be actual barriers to recovery and barriers to troubleshooting that you don’t face when you own the platform in terms of managing it and running it.
I’m not crying wolf, I’m merely suggesting that this should be a planning point in deployments, I think. I think we need to be cognizant of the fact that we have to still be very fluent in the systems and environments we build, even if they are supported or even partially supported by the cloud. Know how the pieces fit together. How do you recover them, what are the recovery options, what if you just want one. little. piece. of information? Can you get it? How? What tools do you have for digging around and getting into a data issue? Might be malicious, might not be, but how do you find out what happened and what can you do about it once you’ve determined that.
Lastly, what’s the response time? It’s one thing to say “they have the most outstanding support – their people really know what they’re doing…” but it’s another to add to that “we just have to work with them in our maintenance window.”
I’m not suggesting you don’t have options, I’m simply saying that staff responsibility for education and testing and such seem almost constant, even if you’re using a software or platform as a service. To assume it’ll just be handled by outside resources in times of crisis may not be an entirely realistic expectation.