4 ms·

I want to see an article like this, but written from a Fortune 500 CTO perspective It seems like they all abandoned their VMware farms or physical server farms

I want to see an article like this, but written from a Fortune 500 CTO perspective It seems like they all abandoned their VMware farms or physical server farms for Azure (they love Microsoft). Are they actually saving money? Are things faster? How's performance? What was the re-training/hiring like? In one case I know we got rid of our old database greybeards and replaced them with "DevOps" people that knew nothing about performance etc And the developers (and many of the admins) we had knew nothing about hardware or anything so keeping the physical hardware around probably wouldn't have made sense anyways

ndriscoll 2y ago
Complicating this analysis is that computers have still been making exponential improvements in capability as clouds became popular (e.g. disks are 1000-10000x faster than they were 15 years ago), so you'd naturally expect things to become easier to manage over time as you need fewer machines, assuming of course that your developers focus on e.g. learning how to use a database well instead of how to scale to use massive clusters. That is, even if things became cheaper/faster, they might have been even better without cloud infrastructure.
jrs235 2y ago
>we got rid of our old database greybeards and replaced them with "DevOps" people that knew nothing about performance etc Seems a lot of those DevOps people just see Azures recommendations for adding indexes and either just allow auto applying them or just adding them without actually reviewing it understanding what use loads require them and why. This also lands a bit on developers/product that don't critically think about and communicate what queries are common and should have some forethought on what indexes should be beneficial and created. (Yes followup monitoring of actual index usage and possible missing indexes is still needed.) Too many times I've seen dozens of indexes on tables in the cloud where one could cover all of them. Yes, there still might be worthwhile reasons to keep some narrower/smaller indexes but again DBA and critical query analysis seems to be a forgotten and neglected skill. No one owns monitoring and analysing db queries and it only comes up after a fire has already broken out.

ndriscoll 2y ago

Complicating this analysis is that computers have still been making exponential improvements in capability as clouds became popular (e.g. disks are 1000-10000x faster than they were 15 years ago), so you'd naturally expect things to become easier to manage over time as you need fewer machines, assuming of course that your developers focus on e.g. learning how to use a database well instead of how to scale to use massive clusters. That is, even if things became cheaper/faster, they might have been even better without cloud infrastructure.

jrs235 2y ago

>we got rid of our old database greybeards and replaced them with "DevOps" people that knew nothing about performance etc Seems a lot of those DevOps people just see Azures recommendations for adding indexes and either just allow auto applying them or just adding them without actually reviewing it understanding what use loads require them and why. This also lands a bit on developers/product that don't critically think about and communicate what queries are common and should have some forethought on what indexes should be beneficial and created. (Yes followup monitoring of actual index usage and possible missing indexes is still needed.) Too many times I've seen dozens of indexes on tables in the cloud where one could cover all of them. Yes, there still might be worthwhile reasons to keep some narrower/smaller indexes but again DBA and critical query analysis seems to be a forgotten and neglected skill. No one owns monitoring and analysing db queries and it only comes up after a fire has already broken out.