From 898357e35f0dc53882b6dbcfe6b5b1359a5130fb Mon Sep 17 00:00:00 2001 From: Roger Ng Date: Thu, 3 Apr 2025 11:17:45 +0000 Subject: [PATCH] Remove the wrong required I/O size to reach max IOPS with throughput limit --- docs/performance.md | 2 +- 1 file changed, 1 insertion(+), 1 deletion(-) diff --git a/docs/performance.md b/docs/performance.md index ab0c0e25..3450c3a8 100644 --- a/docs/performance.md +++ b/docs/performance.md @@ -137,7 +137,7 @@ The indicative figures below were measured using the [CT hammer tool](/internal/ - t3a.small (2 vCPUs, 2 GB Memory) - General Purpose SSD (gp3) - - IOPS: 3,000 (3 KiB I/O) + - IOPS: 3,000 - Throughput: 125 MiB/s The write QPS is around 400. The bottleneck is the high I/O wait at the EBS volume. The "Volume IOPS exceed check" monitoring metrics goes up to 0.8 unit. This is because the AWS deduplication makes use of the [on-disk bbolt key/value store](/storage/bbolt/dedup.go). The Aurora MySQL CPU utilization is around 10%. The EC2 CPU utilization is around 40%.