Is it bad that latency for our internal LAN varies greatly within the 15 - 40 ms range?

Question

I've always thought the "typical" latency for a small office wired Ethernet network would be <5 ms. However, pinging the default gateway yields an average of 14 ms, and any host is around 30 ms.

What could be the cause of these poor latencies? Does it indicate a network problem, or just slow hosts? (I would not think it is the hosts because they are all in that 30 ms range, and some of them are quite new).

Also, what does the variance indicate?

Here is an example:

Pinging 192.168.10.134 with 32 bytes of data:

Reply from 192.168.10.134: bytes=32 time=70ms TTL=128
Reply from 192.168.10.134: bytes=32 time=59ms TTL=128
Reply from 192.168.10.134: bytes=32 time=53ms TTL=128
Reply from 192.168.10.134: bytes=32 time=53ms TTL=128

and another

Pinging 192.168.10.136 with 32 bytes of data:

Reply from 192.168.10.136: bytes=32 time=14ms TTL=128
Reply from 192.168.10.136: bytes=32 time=35ms TTL=128
Reply from 192.168.10.136: bytes=32 time=13ms TTL=128
Reply from 192.168.10.136: bytes=32 time=13ms TTL=128

Finally, is this latency even a problem?

We are using Gigabit switches with an ASA 5505 as the router/firewall/gateway.

EDIT 0: QoS Rule

access-list qos_in extended permit udp any any range 49152 49214
access-list qos_in extended permit udp any any eq 10060
access-list qos_in extended permit tcp any any eq 10032
access-list qos_in extended permit tcp any any eq 10064
access-list qos_in extended permit tcp any any range 49152 49214
access-list qos_out extended permit tcp any any eq 10064
access-list qos_out extended permit udp any any eq 10060
access-list qos_out extended permit udp any any range 30000 31000
...
priority-queue inside
  tx-ring-limit 3
priority-queue outside
  tx-ring-limit 3
...
policy-map qos_out
 class qos_out
  priority
policy-map qos_in
 class qos_in
  priority
service-policy global_policy global
service-policy qos_out interface inside
service-policy qos_in interface outside

That's certainly very slow for a single-site LAN. What network devices are between the computer you're pinging from and the remote computers - is routing being done, for instance? — Shane Madden, Mar 26 '12 at 18:08

score 1 · Answer 1 · answered Mar 26 '12 at 18:06

1

Is it normal? Not really. If you are using inexpensive switches it could be due to switching backplane congestion. Whether it is a problem depends on your environment and the networks role in end user productivity.

answered Mar 26 '12 at 18:06

Thomas G

631
4
6

score 1 · Accepted Answer · answered Mar 26 '12 at 18:07

1

An ICMP ping may not be indicative of actual latency, as many networks have QoS enabled which may reduce the priority of ICMP pings.

answered Mar 26 '12 at 18:07

gekkz

4,229
2
20
19

I do have a QoS rule set to discriminate between voice and data traffic on our network... I'm posting this above. – tacos_tacos_tacos Mar 26 '12 at 18:11
Do ICMP packets receive lowest priority during QoS application? – tacos_tacos_tacos Mar 26 '12 at 18:16
From what you added, you are only doing priority queuing for certain TCP and UDP ports so yes ICMP since it is not matched by your QoS ACLs would still get best effort delivery after all priority queued traffic was processed first. – Thomas G Mar 26 '12 at 19:55
Your QOS policies are layer 3. You state above that these hosts are in the same subnet so the QOS policies will never be applied. How many switches are between the hosts? Are you running spanning tree protocol? – resmon6 Mar 27 '12 at 13:00

Is it bad that latency for our internal LAN varies greatly within the 15 - 40 ms range?

2 Answers2