3 ms·
I've worked at a shop that used both bare-metal servers from OVH as well as VMs from AWS. Using the EC2 instances were infinitely simpler, we could easily autom
by karmakaze 6y ago
I've worked at a shop that used both bare-metal servers from OVH as well as VMs from AWS. Using the EC2 instances were infinitely simpler, we could easily automate so much of it and manage using load balancing and scaling services. AWS instances were the default choice as the scaling up and down was a huge benefit as well as the simpler management and additional services available.
The only places we used OVH bare metal servers was for bulk. Our ElasticSearch cluster had so much RAM that the EC2 cost was very prohibitive. For our DB servers we couldn't at the time get the same performance raid SSDs/controllers and vertical scaling (256GB, 512GB). For our disposable webapp servers we could have used either and since the ops was already worked out having used OVH before AWS kept it that way with a bit more ops and a smaller infra bill. When you have hundreds of hardware hosts, they're failing monthly and it's up to you to image a new one and add it to the cluster.
To have only one way to do things, it also meant that Chef was used instead of immutable instance creation which was a pain to keep everything in sync.
My thought process is basically: (1) small number of instances -> cloud VMs, (2) medium instances w/ founder(s) time but no capital -> bare metal, (3) mature app or huge number -> bare metal