03-08-2008 07:04 AM - edited 03-05-2019 09:37 PM
I have two dell server 2950-III dual quad-core processors
3.0Ghz with 8GB RAM and 1TB SATA drive. Dell_1 has an
ip address of 192.168.1.10/24 Dell_2 has an ip address
of 192.168.1.20/24. Both of the dell is connected to
a Cisco 2960 Catalyst switch copper Gig. I am running
Redhat Linux ES 3 on these servers. I hard code the
interface to 1000/full
When I perform FTP between the servers, I can get about
800Mbps throughput. That's the good part.
Now, I have a Cisco 2621 (64RAM/16F) I connect both
F0/0 and F0/1 to the catalyst 2960. The router is
running IOS version 12.3(24). I set both the interface
of the router and the catalyst to 100 full-duplex.
I give F0/0 192.168.1.1/24, F0/1 192.168.2.1/24. I
give Dell_2 192.168.2.10/24 with the gateway to
be 192.168.2.1. Dell_1's default gateway is 192.168.1.1.
My FTP transfer is peaking out at 5Mbps between
Dell_1 and Dell_2 across the 2621. The CPU on the Cisco
2621 peaks at 99% cpu utilization. I see no
errors on both the catalyst switchports and on the router
interfaces. I thought I could get much better on the
Cisco 2621 than 5Mbps throughput. With
either SecureFTP (sFTP) or SecureCopy (scp), the through put drops to 2Mbps.
In other words, it gets worse.
Anyone know what the throughput for Cisco 2621 router?
IOS on the router is c2600-ik9o3s3-mz.123-24a.bin.
Thanks.
03-08-2008 08:52 AM
David
According to a router performance doc from Cisco the 262x may get up to 12 Mbps. These figures are based on 64 byte packets when no services are configured and all traffic is CEF switched. The throughput drops to less than 1 Mbps when traffic is process switched. Depending on your traffic (and FTP should be sending maximum size packets) and how the router is configured your performance would be different. And given your observation that router CPU goes to 99% it strongly suggests that something is configured that is making the router do something that is impacting performance.
I am not clear in your post whether the switch ports where the servers are connected are in the same VLAN or whether you configured separate VLANs for them.
HTH
Rick
03-08-2008 09:09 AM
Rick,
- When both servers are in the same VLAN,
I get 800Mbps throughput,
- There is no other services configured on
the router. This is a test router.
- I have CEF enable,
- Dell_1 and F0/0 is in vlan 10 and Dell_2
and F0/1 is in vlan 20. I get 5Mbps
throughput with 99% CPU.
- I repeat the test with a Cisco 7204VXR
and I can only push 40Mbps and 99% CPU. I
have CEF enable in this case as well.
Any ideas why?
David
03-08-2008 09:40 AM
David
If all it were doing were forwarding the traffic and only using CEF I would certainly not expect the CPU to go that high. The high CPU suggests that the router is doing something other than just CEF forwarding packets. Is there any possibility that it is logging anything about the traffic? Is there any possibility that it is having to fragment packets?
HTH
Rick
03-08-2008 10:05 AM
this is what I did to the router:
1- write erase,
2- reload
3:
conf t
hostname C2621
no logging on
no logging console
no logging buffered
no logging monitor
ip cef
int f0/0
speed 100
dup full
ip address 192.168.1.1 255.255.255.0
no shut
ip cef
int f0/1
speed 100
dup full
ip address 192.168.2.1 255.255.255.0
no shut
ip cef
I'm still getting 5mbps throughput at 99%
CPU. On the VXR7204, repeat with the same
process, throughput is at 40Mbps, with CPU
at 99%.
In both cases, with sFTP and scp
applications, throughput is around 20mbps
on the VXR7204 and between 1-2mbps on the
2621.
If I replace the Cisco router with a pair of
Checkpoint SPLAT Active/Active ClusterXL
firewall, I get wire-speed file transfer at
about 90Mbps.
03-08-2008 10:24 AM
David
Thanks for posting config information. That certainly seems to indicate that logging is not an issue. I do not know why it would need to fragment or do anything to the packets but I am wondering what would drive CPU so high. I wonder if you would clear the counters, run a transfer and check the number of packets in on each interface against the number of packets out on the other interface?
HTH
Rick
03-08-2008 10:53 AM
Rick,
I already tried what you did and the # of
packets in is almost the same as the # of
packets out on the other side.
I dont' think packet fragmentation is
an issue because I run tcpdump on the
servers on both side I do not see any
fragmentation packets at either end.
03-08-2008 11:06 AM
When you are doing these transfers and you look at those 100 meg router uplinks are they pretty much buried ? You also have to remember you have gigabit nics and you are trying to shove all that information up a 100 meg pipe , so I calculate at best you would get maybe 12 megabytes per second transfer with overhead . In your first post are you talking bits or bytes ????
03-08-2008 11:15 AM
12 megabytes per second = 96 megabits per sec.
I get about 5 megabits per second which
is not a lot.
I even change the speed on the Dell servers
from 1G to 100m and the same thing on the
catalyst 2960 as well. Same result.
03-08-2008 11:33 AM
Please post what you are seeing
1) Clear the counters from all interfaces
2) Initiate the transfer once again
3) Post the show interface, this would give us a bit more information.
You may also post the show processes while the transfer is taking place.
Thanks
__
Edison.
03-08-2008 12:24 PM
03-08-2008 12:34 PM
http://www.cisco.com/en/US/products/hw/routers/ps359/products_tech_note09186a00801c2af3.shtml
Your high IP Input is of concern, the above might assist troubleshooting it.
PS:
Process switching could account for your low file transfer rates.
[edit]
PPS:
The packets within the input queues could be of concern too. See Input Queue Drops in: http://www.cisco.com/en/US/products/hw/routers/ps133/products_tech_note09186a0080094791.shtml#topic2
[edit2]
Regarding Input Queue Drop, above, none noted as being dropped, I think we still might be concerned about number shown within input queues.
03-08-2008 12:21 PM
When both the 2621 and 7204VXR max out their CPU, interrupt CPU is also about maxed out?
What NPE is within the 7204VXR?
When you jump across either router, are the packets 1500 bytes?
PS:
The performance numbers we have for a 262x are: 1.5 Kpps process switching and 25 Kpps fast switching.
Ethernet packet size to bandwidth table (for gig):
Gig
Packet Size (Bytes) 64 128 256 512 1024 1518
Theoretical Maximum Kpps 1488 845 453 235 120 81
Remember for file transfer, traffic is bidirectional, need some pps for return ACKs.
03-08-2008 01:54 PM
Maybe there is something you can figure out from this doc.
http://www.cisco.com/en/US/products/hw/routers/ps359/products_tech_note09186a00801c2af3.shtml
03-09-2008 10:17 AM
I actually read all of those links prior to
posting my question in the forum.
Any more ideas anyone? Thanks.
Find answers to your questions by entering keywords or phrases in the Search bar above. New here? Use these resources to familiarize yourself with the community: