BeagleBone slow inbound net I/O

Bernd Walter ticso at cicely7.cicely.de
Thu Mar 12 13:34:42 UTC 2015


On Wed, Mar 11, 2015 at 04:51:15PM -0400, Brett Wynkoop wrote:
> Greeting-
> 
> So I am finding that network reads from my nfs server are timing out on
> the BeagleBone.  I am having no similar issues with other nfs clients
> on my network.  
> 
> In an attempt to eliminate NFS as a possible reason for slowness of
> transferring the ports tree into the system I am now transferring by
> piping tar through nc.  This is after I had a failure with rsync
> inbound on the same data.  The transfer is going VERY slow.  Many times
> slower than my transfers out of the BB to my 10.1 server.
> 
> Have I managed to find a network driver issue?  Any ideas how to gather
> more information to help get to the bottom of things?
> 
> Netstat seems to provide nothing useful:
> 
> [wynkoop at beaglebone ~]$ netstat -i
> Name    Mtu Network       Address              Ipkts Ierrs Idrop    Opkts Oerrs  Coll
> cpsw0  1500 <Link#1>      00:18:31:8c:a5:22        0     0     0        0     0     0
> cpsw0     - 199.89.147.0  beaglebone            1117     -     -      511     -     -
> lo0   16384 <Link#2>                              36     0     0       36     0     0
> lo0       - localhost     ::1                      0     -     -        0     -     -
> lo0       - fe80::1%lo0   fe80::1%lo0              0     -     -        0     -     -
> lo0       - your-net      localhost               36     -     -       36     -     -
> [wynkoop at beaglebone ~]$ 
> 
> I do not believe the transfer is limited by disk i/o because 
> zpool iostat 5 shows inferior speed as compared to when I did the copy from 
> the sd card to the usb stick.

Use gstat to monitor disks.
If the flash disk is cause the slowness then you likely see a large
L(q) and high ms/w.
Do not care about %busy - this is an interesting value, but missleading.
It is not unlikely for small flash based devices to have problems to
keep up with writing - big SSD with many flash chips and decent
controller chips balance writes over chips to overcome this issue.

-- 
B.Walter <bernd at bwct.de> http://www.bwct.de
Modbus/TCP Ethernet I/O Baugruppen, ARM basierte FreeBSD Rechner uvm.


More information about the freebsd-arm mailing list