I just start playing with EC2 today and these are some useful notes and observation. The small standard instance is quite slow, only scores 50 under Unixbench-WHT. Our quads E5405 are around 410+, Q9550 is at 650-750.
I also was confused about S3 versus EBS. The console management makes life much easier, but could be improved even further. This is what you do after attaching a EBS volume to an instance:
mkfs -t ext3 /dev/sdx
mkdir /mnt/data-store
mount /dev/sdx /mnt/data-store
For CentOS AMI (ami-0459bc6d), you need to run this to be able to customize the AMI and build your own
depmod -a
modprobe loop
As for performance, it’s pretty disappointing indeed. I ran Unixbench WHT 4.1 and small instance only got 51 (CentOS 5 32-bit) and the large instance got 111 (CentOS 5 64-bit). If you run an instance 24/7, it’s much more expensive for the same performance of a dedicated server. Our quad core servers usually get 600+ on the same test. The price you pay for EC2 is “on-demand”, meaning that you need to automate it to the point scaling up during peak is seamlessly and not having to run a single or even multiple instances all the time.
Leave a Reply