Re: Network very very slow... was iSCSI and jumbo frames

[mlelstv%serpens.de@localhost (Michael van Elst)]

rvp%SDF.ORG@localhost (RVP) writes:

>$ sudo dd if=/dev/sd0 of=/dev/null bs=1m count=100
>100+0 records in
>100+0 records out
>104857600 bytes transferred in 11.883 secs (8824168 bytes/sec)

The block device is always slower as everything is funneled through
the buffer cache in tiny blocks (usually 2kbyte).


>If I use the raw disk, I get:
>$ sudo dd if=/dev/rsd0 of=/dev/null bs=1m count=1000
>1000+0 records in
>1000+0 records out
>1048576000 bytes transferred in 9.722 secs (107855996 bytes/sec)

>Performance close to reading the raw /dev/rdkN partition underlying it.

>If I mount /dev/sd0 (an ext2fs partition) and read a file, I get:

>$ sudo dd if=/mnt/tmp/noto-ttf-20170410.tgz of=/dev/null bs=1m
>775+1 records in
>775+1 records out
>813225648 bytes transferred in 7.803 secs (104219613 bytes/sec)

>Again, good performance.


Same here when the iSCSI server (an Intel N3150) connects to
its own volume:

Reading:
1073741824 bytes transferred in 7.140 secs (150384008 bytes/sec)

Writing:
1073741824 bytes transferred in 9.254 secs (116030022 bytes/sec)

Copying a volume to a second one (on a different phyiscal disk):
1073741824 bytes transferred in 16.315 secs (65813167 bytes/sec)


Things look different if I reduce the buffer size from 1024k
to 64k:

Reading:
1073741824 bytes transferred in 12.178 secs (88170621 bytes/sec)

Writing:
1073741824 bytes transferred in 13.627 secs (78795173 bytes/sec)

Over a real network with higher latencies than loopback the effect
is much larger, especially for reading,

-- 
-- 
                                Michael van Elst
Internet: mlelstv%serpens.de@localhost
                                "A potential Snark may lurk in every tree."