2 ms·
I once had to debug the scaling performance of a MPI-based simulation algorithm on cheap linux machines with TCP. I finally collected a TCP trace and showed it
by dekhn 6y ago
I once had to debug the scaling performance of a MPI-based simulation algorithm on cheap linux machines with TCP. I finally collected a TCP trace and showed it to the local expert who said: "hmm, 250ms delay right there.. that's the TCP retransmit timer... you're flooding the ethernet switch with too many packets and the switch is dropping them. Enable <such and such a feature>."
Since then I've always kept various constants in human RAM because it helps root cause.
- whoisburbansky 6y agoWhat are some examples of other such constants you have hanging around?
- dekhn 6y agoa good one (although it's really a functional constant that has to be determined per-system) is the amount of time it takes to look up data in local RAM vs RAM that is attached to the other CPU socket (IE, over the system bus); it's 50% longer. I typically remember all the TCP timing constants, the seek time of a hard drive, everything in "Programmer's rule of thumb" (how long it takes for data to travel by light from CA to NL, etc)