Hi all, just wanting a little sanity check before implementing and getting hard test numbers. Hopefully your input can help prioritize this work.
We currently build a lot of very large in house software, and do so with beefy build servers in CI, but we want to close the feed back loop further and are looking at ways to speed up build time, especially for developer workstations. I'm looking if anybody has recent experience with distcc or ccache, and if they are worth the technical overhead to implement. Sadly due to the legacy nature of some of the code base and focuses on other products splitting the large monolithic applications into smaller, quicker building libraries is not feasible at this time.
All build and development systems are on 10G lines with cut through switches, so latency will be relatively low. The average systems are beefy (2x8 core xeons at 2.7GHz, HT left enabled so 32 cores). Sadly, only the CI systems have Fusion-IO cards and the systems in the distcc config will have a pair of 10k RPM SAS drives in a HW raid 1.
Do you all thing it's worth continuing down this path for raw numbers, or does your money side with me that seek time + network overhead will cause more delay than help with such powerful processors?
Thanks in advance!
[link][12 comments]