$ siege -b -c 1 -r 2000 http://127.0.0.1:7777
Transactions: 2000 hits
Availability: 100.00 %
Elapsed time: 1.40 secs
Data transferred: 0.39 MB
Response time: 0.00 secs
Transaction rate: 1428.57 trans/sec
Throughput: 0.28 MB/sec
Concurrency: 0.95
Successful transactions: 2000
Failed transactions: 0
Longest transaction: 0.01
Shortest transaction: 0.00
Transactions: 2000 hits
Availability: 100.00 %
Elapsed time: 1.31 secs
Data transferred: 0.39 MB
Response time: 0.00 secs
Transaction rate: 1526.72 trans/sec
Throughput: 0.29 MB/sec
Concurrency: 0.94
Successful transactions: 2000
Failed transactions: 0
Longest transaction: 0.01
Shortest transaction: 0.00
http://sheddingbikes.com/posts/128082938...
That's Zed Shaw, who famously left the Ruby community for Python, and the Pythonista's still aren't sure if that's a win or not. It is nice to see someone take a scientific wack at the issue and in one spot he says:
But, if your server is the kind with lots of idle connections just hanging out, then epoll wins not poll. That's because your active/total ratio is < 0.6 most of the time.
Where MUDs would clearly fall into the second category. Still, I've been looking at epoll as an untapped panacea – which apparently, it aint.