search

michalpolkowski / alt-f

Automatically exported from code.google.com/p/alt-f
0 stars 0 forks source link

Dropped packets #25

Open GoogleCodeExporter opened 9 years ago

GoogleCodeExporter commented 9 years ago 
What steps will reproduce the problem?
1.Start Samba on NAS
2.Start Intel NAS performance test
3.run ifconfig

What is the expected output? What do you see instead?
There are some dropped packets.
eth0      Link encap:Ethernet  HWaddr XX:XX:XX:XX:XX:XX
          inet addr:192.168.7.14  Bcast:192.168.7.255  Mask:255.255.255.0
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:6203605 errors:258 dropped:258 overruns:0 frame:0
          TX packets:8345210 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000
          RX bytes:1426737278 (1.3 GiB)  TX bytes:447621543 (426.8 MiB)
          Interrupt:21

There are no dropped packets on vendor (1.09) firmware in the same conditions.

What version of the product are you using? On what operating system?
Alt-f 0.1B6 on top of vendor firmware.

Please provide any additional information below.
Samba performance of Alt-f is about 75% of vendor's firmware.
Still analyzing results...

Your work is great! Easy-to-use, with a lot of function firmware.

Original issue reported on code.google.com by Mike.Sem...@gmail.com on 2 Jan 2011 at 2:52

GoogleCodeExporter commented 9 years ago 
[deleted comment]
GoogleCodeExporter commented 9 years ago 
-Have you run ifconfig before starting the tests?
-Are you running NFS? mountd uses udp, which can lead to dropped/duplicated 
packets.
-Are you on a Gbit network? Direct connection or using a switch/hub/router?

-----------------------------------------------------------------------------

> Samba performance of Alt-f is about 75% of vendor's firmware.

This statement needs more "substance" and is out of scope, use a separate issue 
for it, please.
-How do you measured it? Empirical testing? Using the intel NAS tool? Does it 
have a global index? I don't think so. Could you post the results on all tests?

-What filesystem were you using? ext2/ext3/ext4? same fs on dlink firmware 
tests and on Alt-F tests? using RAID 0/1/5/JBD? Same disks on both tests? The 
host computer was running linux or MS?

Thanks

Original comment by whoami.j...@gmail.com on 6 Jan 2011 at 5:13

GoogleCodeExporter commented 9 years ago 
-Have you run ifconfig before starting the tests?
yes, same tests with 1.09(ffp with ssh only) & alt-f(not flashed, default 
settings with SAMBA only) firmwares, with restart between tests.
-Are you running NFS? mountd uses udp, which can lead to dropped/duplicated 
packets.
No, only SAMBA was in use
-Are you on a Gbit network? Direct connection or using a switch/hub/router?
Gbit network with D-Link switch.

PS. I'll create separate issue for SAMBA performance soon.

Thank you for your work!

Original comment by Mike.Sem...@gmail.com on 6 Jan 2011 at 6:41

GoogleCodeExporter commented 9 years ago 
Tested with B7, the same issue:
dropped packets on B7 firmware versus 1.09.

ifconfig on DNS before test:
# ifconfig
eth0      Link encap:Ethernet  HWaddr XX:XX:XX:XX:XX:XX
          inet addr:192.168.7.14  Bcast:192.168.7.255  Mask:255.255.255.0
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:271 errors:0 dropped:0 overruns:0 frame:0
          TX packets:319 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000
          RX bytes:30584 (29.8 KiB)  TX bytes:149456 (145.9 KiB)
          Interrupt:21

lo        Link encap:Local Loopback
          inet addr:127.0.0.1  Mask:255.0.0.0
          UP LOOPBACK RUNNING  MTU:16436  Metric:1
          RX packets:4 errors:0 dropped:0 overruns:0 frame:0
          TX packets:4 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:0
          RX bytes:392 (392.0 B)  TX bytes:392 (392.0 B)

and after:

# ifconfig
eth0      Link encap:Ethernet  HWaddr XX:XX:XX:XX:XX:XX
          inet addr:192.168.7.14  Bcast:192.168.7.255  Mask:255.255.255.0
          UP BROADCAST RUNNING MULTICAST  MTU:1500  Metric:1
          RX packets:5038960 errors:95 dropped:95 overruns:0 frame:0
          TX packets:6373456 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:1000
          RX bytes:1058652623 (1009.6 MiB)  TX bytes:3356284536 (3.1 GiB)
          Interrupt:21

lo        Link encap:Local Loopback
          inet addr:127.0.0.1  Mask:255.0.0.0
          UP LOOPBACK RUNNING  MTU:16436  Metric:1
          RX packets:4 errors:0 dropped:0 overruns:0 frame:0
          TX packets:4 errors:0 dropped:0 overruns:0 carrier:0
          collisions:0 txqueuelen:0
          RX bytes:392 (392.0 B)  TX bytes:392 (392.0 B)

there are 95 errors, nothing similar with 1.09 firmware with exactly the same 
tests.

Original comment by Mike.Sem...@gmail.com on 21 Apr 2011 at 9:25

GoogleCodeExporter commented 9 years ago 
I'm not really worried with this issue.

Dropped packets is not really an issue when using tcp, as the packets are 
retransmitted. Only if using VOIP or other realtime protocol it could be an 
issue; even in that case, the fraction of dropped packets, if compared with the 
total number of received packets, is negligible (0.001%).

Dropping packets is indeed a simple way to slow down the transmitter, and might 
be attributed to different congestion techniques used in linux 2.6.12 (dlink fw 
is based on, and probably using "reno") and linux 2.6.35, that uses "cubic".

Without further details, namely in what intel test the packets are dropped, and 
what the test is stressing-on, I can't do nothing.

I tested a flood ping on a gigabit network with MTU=9000 (ping -s 9000 -f nas) 
and no dropped packets appeared. I only observed dropped packets when the ping 
packet size exceeded the configured MTU on the nas.

Original comment by whoami.j...@gmail.com on 24 Apr 2011 at 11:04