How much of your database workload is going to be in a public cloud?

  • And not just your standard SQL servers

  • In my line of work there is always going to be data we cannot put in the cloud.

  • Possibly between 20-70% in GOV Cloud, but public cloud would most likely be 0%.

  • Same here! Anyway, it's all sitting in isolated networks, and it should definitely stay that way Wink

  • I'm reminded that the Cloud means you are just using someone else's servers and infrastructure, that you can't control. 

  • I worked on a hospital network for seventeen years and saw how patients and medical providers required access to their data 7x24.  And I've seen too many cloud-based total outages that affect patients and medical staff for hours, or even days; there's nothing good about the cloud then.  There will always be a need for critical data  to remain accessible 7x24, and the only way I've seen that happen is when our employees are 100% in charge of the servers & pathways that data uses.  We have multiple data centers connected by multiple resilient 10G private lines.  The more we rely that infrastructure, the better our uptime has been.  And as we came to rely on the cloud, the more times we saw our downtime increase.  Sure, the cloud is popular.  It's (maybe) even practical--for smaller companies who can't afford multiple data centers and the depth and scope of employees who maintain & support the systems.  But if someone dies because of human error in a router upgrade, or because ransomware from the Internet isolated their data, putting everything into the cloud wasn't worth the cost savings over local data centers & the staff they require.  But SOME things can be in the cloud--as long as there's a fast and reliable failover solution that redirects the traffic back to our data centers in the event of a national problem affecting DNS or BGP or OSPF.  I haven't seen that yet, but I can imagine one.  Business types monitoring the bottom line like the cloud for its lower costs.  It doesn't take nearly as many local employees to operate & maintain it.  It can have fast-healing redundant server solutions.  And someone else does all the work.   But . . . putting your data in someone else's server eliminates your ability to verify its safety and security.  And that's something that Personal Health Information can't tolerate when it becomes vulnerable to single points of failure the end customer (hospitals) can't see or correct.  Patient data that is required to get timely and correct decisions from health care providers can't be made subject to unknowable / unreliable problems caused by other companies and people who aren't in our chain of operations or command.  When one person can make a mistake, or intentionally perform a bad act that affects access to that data--and when we can't know who that person is--what do you tell the family of someone who died because their doctor didn't have access to their health care record through that mistake or act of anger?  Or more practically, what do you tell that family's lawyers when they come to sue you and your company?

  • My company closed down several of our datacenters and has a directive that all new projects are cloud based.   We keep a small 2-6 node VMware cluster in every warehouse/shipping facility for the apps they use because latency really sucks for their scanners.  We spend a staggering amount of money for it but the leadership is hoping that the cloud first strategy will drive change in the organization.  It certainly did drive a lot of turnover,  so maybe that counts? 

  • Datacenter will be a dinosaur for most businesses 3 years from now. Even for those of us with compliance/security requirements... they will shift.

  • Last Summer we moved to a new data center, it wasn't the boards first choice of where we needed to go. 4 years ago they asked us to price out moving to the cloud. We modeled a 100% shift, a 'only a few things' on prem, and a 50% move and 100% on prem for 2, 5 and 10 year forecasts total cost of ownership. The dev work to get 100% in the cloud took too long, so its cost was outrageous, but also impossible. Get 'lots' to the cloud was do-able to shrink our footprint in the data center, but if you need some on-premise, you can get a lot of bang for your buck by leveraging some of the same tech that AWS and Azure do. The new data center looked cheap compared to moving a significant workload. That said, we are hybrid, and compute workloads get evaluated routinely to see where they should be. Mostly its feast and famine stuff and stuff with little ties to other systems that make sense today.