[ENBD] 2.4.32 conclusions

[Peter T. Breuer]

"Also sprach Anders Blomdell:"
> 
> > "Also sprach Peter T. Breuer:"
> >> What does top say is the process doing the work, and what does strace
> >> say it is doing? :-).
> 
> Client:
> 
>   dd if=/dev/zero of=/dev/ndk bs=64k
> 
>   3161 root      15   0   440  440  380 D  8.9  0.0   1:40.63 dd
>   3138 root       9   0  1024 1020  888 S  6.3  0.1   0:07.86 enbd-client
>   3139 root       9   0  1024 1020  888 S  5.0  0.1   0:07.85 enbd-client
>   3140 root       9   0  1024 1020  888 R  3.0  0.1   0:07.88 enbd-client
>   3141 root       9   0  1024 1020  888 S  2.3  0.1   0:07.87 enbd-client

That's OK, although dd is waiting on i/o (not surprisingly).

> Server:
>   2012 root      18   0  2064 2064 1612 S 16.0  0.2   0:27.00 enbd-server
>   2013 root      10   0  2064 2064 1612 D 11.7  0.2   0:26.95 enbd-server
>   2014 root      10   0  2064 2064 1612 D  9.3  0.2   0:26.45 enbd-server
>   2011 root       9   0  2064 2064 1612 D  3.3  0.2   0:25.91 enbd-server
> servers does a lot of recv's and writes (20-40 MBytes/s)

The servers, which are entirely userspace, look in worse shape. Three of them
are starving! They are writing to disk and can't get back acks as fast as they can
write.

Peter