I'm studying the Amazon EBS–optimized instances document trying to apply it to a specific scenario.
I'm using a c6a.2xlarge instance with a single gp3 volume (2 TB) attached. According to the document the c6a.2xlarge instance is listed as one of those instance types that support maximum performance for only 30 minutes at least once every 24 hours.
About the baseline performance of c6a.2xlarge, the second table says the following:
- Baseline throughput=265.25 MB/s
- Baseline IOPS=8333
About the maximum performance of c6a.2xlarge, the first table says the following:
- Maximum throughput=833.333 MB/s
- Maximum IOPS=26667
On the other side, the gp3 volume can be configured with:
- Min: 125 MiB, Max: 1000 MiB. Baseline: 125 MiB/s per TiB.
- Min: 3000 IOPS, Max: 16000 IOPS
I have a workload that requires sustained maximum performance for longer than 30 minutes, but I cannot afford a different instance.
I'm trying to intersect the information reported above, so my questions are:
- If I modify the
gp3volume with exactly 265 MB/s of throughput and 8333 IOPS I'm almost guaranteed to have such performance 24/7 ? - Is there any benefit in increasing the throughput up to 833 MB/s and/or the IOPS up to 16000 given the fact that such performance can be sustained just for max 30 minutes in 24 hours ?
