| 1 |
On Wed, Dec 21, 2011 at 11:43 PM, Donnie Berkholz <dberkholz@g.o> wrote: |
| 2 |
> I looked into this 6 or 7 years ago. It wasn't feasible unless you were |
| 3 |
> on an extremely high-speed, low-latency network, beyond what was |
| 4 |
> typically accessible at the time outside of universities and LANs. Could |
| 5 |
> be worth exploring again now that 25-100 mbps connections are becoming |
| 6 |
> more common. |
| 7 |
|
| 8 |
I tried messing around with this with Amazon EC2. The problem was |
| 9 |
that due to latency I only really saw the benefit for VERY high levels |
| 10 |
of parallelization (think -j25+).. However, make isn't actually |
| 11 |
"distcc-aware" so it just runs 25 jobs of anything in parallel. So, |
| 12 |
anytime a makefile launched a ton of java or python jobs the host |
| 13 |
ground to a halt as it wasn't distributed and it was way more than the |
| 14 |
host could handle (especially java - which swapped like there was no |
| 15 |
tomorrow). |
| 16 |
|
| 17 |
If somebody were to do a distcc-ng for a large cluster one of the |
| 18 |
problems to solve would be having it not run jobs in parallel if it |
| 19 |
couldn't actually distribute them. |
| 20 |
|
| 21 |
Rich |
Archives