strange Ping responses

J

jerb

Limp Gawd
Joined
Nov 27, 2004
Messages
322
Im getting a wierd response from a cisco 2900 series switch.

if I ping it directly, i get a response of abut 40ms ... horrible!
however if I ping 1 hop up i get a response of >1ms ...?

in fact I went 6 hops up (as far as I can go and still stay on my network) and im getting responses of 1 - 5 ms... how can this be?
 
Wait.. are you pinging the 2900 or another device from the 2900?

Did you do a trace route to see what path is being taken?

Obviously, the more hops you encounter, the more latency is injected into the network. As well as loaded routers/switches etc.

Lots of variables, need more info.
 
As Wes said, we would need to know more about the topology.. how many redudant links do you have, etc..

It's possible that going there is <1ms, but the return route is different and takes a higher latency link.

If you ping the 2900 from the router that has the connected managment VLAN, is it 1ms or 40? If it's 1ms, then you have a return route issue.
 
not a problem.

heres how the network looks

3800xl -> 2900 -> 2950 -> windows xp box

I am sitting at the windows box at the end of the chain and I am pinging the 2950 (ip address 172.16.13.232) and its telling me that its taking 40ms to reach the destination, thats an awefully long time considering the cabling is good and I have a direct connection to the switch

however If I ping the router which is the 3800 series (IP address 172.16.12.1) it tells me that its taking 1-5ms to get back to me

I checked the logs of all three devices, I dont have any loop back problems, duplex errors, etc. why does it take longer to ping the switch that I'm directly connected to that the router (or even the switch) up stream from it (since the packets have to pass through the 2950 to get to the 3800)
 
I should add, we are not utilizing the full capabilities of the cisco switches. there are no vlans or communities setup.
 
communities on a switch? Huh?

anyways, your latency is a bit high but im curious as to how much traffic this switch forwards? Do keep in mind that transit traffic is switch and processed WELL before traffic destined for the processor so this is why your pings vary in latency. When you ping the switch the processor is queuing that request(request cycles) at a low priority while trasit traffic is given high priority.

Check the proc on the switch: show proc cpu

I have some 3524's at home(a tiny bit better than the 2950) and the thing constantly sits at about 50% utilization. I can ping it 4 states away through an ipsec over GRE tunnel and the average latency is about 30ms.
 
the thing is when you are pinging the switch you are basicically pinging its software right?
because the switching fabric is usually up front and then the software is behind that and can modify the way the fabric works with ACLS and vlans, etc. Sounds to me if the delay isnt being added when you ping a device on the other end that its take 40ms because the CPU is busy.

Is network perfomance or reliability being affected? If not I wouldn't worry about it. How many pcs are behind the same 2900 (not the 2950, total behind the 2900)?
 
xphil3 said:
communities on a switch? Huh?

anyways, your latency is a bit high but im curious as to how much traffic this switch forwards? Do keep in mind that transit traffic is switch and processed WELL before traffic destined for the processor so this is why your pings vary in latency. When you ping the switch the processor is queuing that request(request cycles) at a low priority while trasit traffic is given high priority.

Check the proc on the switch: show proc cpu

I have some 3524's at home(a tiny bit better than the 2950) and the thing constantly sits at about 50% utilization. I can ping it 4 states away through an ipsec over GRE tunnel and the average latency is about 30ms.
Click to expand...

SNMP communities

xphil3 said:
communities on a switch? Huh?

anyways, your latency is a bit high but im curious as to how much traffic this switch forwards? Do keep in mind that transit traffic is switch and processed WELL before traffic destined for the processor so this is why your pings vary in latency. When you ping the switch the processor is queuing that request(request cycles) at a low priority while trasit traffic is given high priority.

Check the proc on the switch: show proc cpu

I have some 3524's at home(a tiny bit better than the 2950) and the thing constantly sits at about 50% utilization. I can ping it 4 states away through an ipsec over GRE tunnel and the average latency is about 30ms.
Click to expand...
 
the switch doesnt transmit much traffic at all. there are maybe 20 machines attatche to it which grab small text files for CNC machines once an hour or so. none of the machines are configured to allow for internet access by the machinists in the shops so the traffic is next to nothing.

its very reliable, I'm not seeing many errors in any of the logs. I happened to ping it from one of the stations and I was curious about why that number was so high.

its not reporting its cpu usage or memory usage to solarwinds so Im goinng to have to play with the snmp settings later and see what I can dig up.

thanks for all the info guys
 
OmegaAvenger said:
SNMP communities
Click to expand...
AHHHH... community strings. Never heard anyone say it like that, I thought he was talking about bgp communities on that switch, which isn't possible....

OmegaAvenger said:
the thing is when you are pinging the switch you are basicically pinging its software right?
because the switching fabric is usually up front and then the software is behind that and can modify the way the fabric works with ACLS and vlans, etc. Sounds to me if the delay isnt being added when you ping a device on the other end that its take 40ms because the CPU is busy.
Click to expand...
You are pining software, but that software requires CPU cycles, even more becuase the target of the ping is on the switch itself. ICMP to the CPU is given lowest priority.... same theory goes for pinging routers and packets being punted to the RP for processing. The switch fabric in this case would not come into play as the switch is not forwarding traffic from port to port.

Jerb, if its not impacting anything then dont worry about it. More often then not in networking we experience many unexplainable things.
 
You must log in or register to reply here.
Back
Top