This matches my testing. Aurora is fast enough that we actual started to contend on the heap locks, and had to tune our batch orders. Too bad network-attached i/o is so slow.
Note that they were running psql on ec2 instances, which implies bare metal with SSDs.
Yes you can get actual hosts in AWS if you pay for them.
The article also implies that they were never able to get psql to replicate effectively.
Whomever their DBA was was couldn't do it, so they were like "fuck it, let's move to Aurora." Their database brought no actual value, so it make sense for them.
They likely were running data on EBS volumes instead of bare metal SSDs, due to ease of recovery (a failed instance does not lose data on the attached EBS volumes). You can only run your DBs on bare metal SSDs if you are prepared to lose a node’s data completely.
In fact, many instance types no longer have any ephemeral storage attached and it’s a default practice to use EBS for root and data volumes.
There are some instance types that have extremely fast EBS performance (EBS io2 Block Express), which has hardware acceleration and an optimized network protocol for EBS network I/O and offers sub-millisecond latency. However, these are expensive and get even more so if you go up in IOPS.
They were on AWS already, they just moved from "the most expensive" to "the second most expensive" alternative.
I don't think that these numbers stand if you compare with a PostgreSQL on a bare metal server with NVMe SSDs attached to it.
Yes you can get actual hosts in AWS if you pay for them.
The article also implies that they were never able to get psql to replicate effectively.
Whomever their DBA was was couldn't do it, so they were like "fuck it, let's move to Aurora." Their database brought no actual value, so it make sense for them.
No, this doesn't imply SSDs.
EC2 instances do not have access to SSDs but to EBS volumes which is a much slower distributed network block storage medium.
Not sure about their PostgreSQL replication issues but lots of companies manage to make it work without a hitch.
In fact, many instance types no longer have any ephemeral storage attached and it’s a default practice to use EBS for root and data volumes.
There are some instance types that have extremely fast EBS performance (EBS io2 Block Express), which has hardware acceleration and an optimized network protocol for EBS network I/O and offers sub-millisecond latency. However, these are expensive and get even more so if you go up in IOPS.
Using EBS seems like a total anti-pattern for DB workloads.