AWS GP3 Volumes
AWS made the following announcement at Reinvent2020
As noted by our king of snark Corey Quin, this has some great cost implications to AWS customers!
Why should I care
We have all accepted the model of gp2 volumes, to get more IOPS we increase our volume size. This however leads to massive over provisioning of either storage or wasted IO allocation that will never be consumed. No one wants to pay for io1, so we just accept and move on.
Now customers have more options!
Another value add is this can all be done without re-creating your gp2 volumes!
Its not always all about S3
At my company we have migrated our Splunk Enterprise instance to AWS. As part of that migration we have implemented SmartStore which enables us to leverage S3 as a backend for long term data retention.
Block storage is still required on the indexer EC2 nodes to keep hot data buckets available for active searches. Depending on the data volumes being ingested this can get quite large.
Our Splunk infrastructure is made up of an index cluster, search head cluster and other supporting bits and pieces, which is costing about 300USD daily for my EBS usage of around 70TBs.
Kicking the tires with gp3
I kicked off the migration process for most of my gp2 volumes on the 1st of December. Yes it took some time to migrate the large ones, but here is the result!
As the graph shows, a significant drop in my daily cost for block! Now at about 234USD daily which is about 22% saving
Keeping up to date with releases from AWS is a full time job, but rewarding when it saves me money.
I recently joined the FinOps foundation (for free). There is some really great content available for anyone managing workloads in the cloud. Go check it out https://www.finops.org/
There is always a catch....
Some people have reported some teething issues with gp3 volumes around status reporting and a bug regarding billing. Should this happen raise an AWS support ticket.
Hope someone else finds this useful.