Benchmarking amazon EC2 vs Google compute engine

Question

For EC2: t2.micro, host in Oregon

For compute engine: n1-standard-1, in west-us

Both in ubuntu 16.04, single instance. I installed nvm, then node js 7.7.2 and initiated a simple server:

require('http').createServer(function(req, res) {
  res.end('Hello world')
}).listen(8080)

Then from my local machine, I used wrk for bench marking

wrk -t12 -c400 -d10s http://myinstanceaddress

Result, average of 5:

EC2: ~1200 requests/s
Compute engine: ~500 requests/s

I expected that compute engine n1-standard-1 would perform better than EC2 t2.micro since the former has more CPU and power. However, the result indicated otherwise.

So, my question is: is this simple benchmark accurate, what other factors that I need to consider? If the benchmark is accurate, does this mean AWS EC2 t2.micro actually performs better than google compute engine n1-standard-1 despite the later has more power (CPU and memory as advertised)?

E. Anderson E. Anderson · Accepted Answer · 2017-03-14T04:30:58

Have you checked the network path (e.g. latency) between the two nodes, using a tool like ping?

I'm guessing that you're actually being limited by the network path somewhere along the way.

Benchmarking amazon EC2 vs Google compute engine

1 Answers