When ping reply is too long?

[Marcin_9]

Hi. When It can be said that ping time response is too long. I mean when I have to serwers and lets say they are in the same switch what response time should be maximum acceptable?
I know that it can be ridiculous question but I would like to know if this subject can be discussed in this way?.

[Peter Godron · Accepted solution]

Marcin,
the rule is :
If you notice that the response is slow, then it is slow.
In other words:
There is no clear rule on how fast ping should respond, but if it slower then you you need it to be it is too slow.

[Arunvijai_4]

Hello,

It could be due to slow DNS resolution and network lagging. Check the below threads for more information,

http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=960924
http://forums1.itrc.hp.com/service/forums/questionanswer.do?threadId=958691

-Arun

[Marcin_9]

Thanks

[Arunvijai_4]

http://forums1.itrc.hp.com/service/forums/helptips.do?#28

[rick jones]

Indeed, the question is rather like asking "How long is a piece of string?" :)

Slow DNS would affect when the _first_ of the pings was sent but should have absolutely no effect on the response times of the pings themselves.

Apart from that, "ping time" will be a function of a number of things:

1) The distance between the two systems - both in terms of physical distance and number of intervening devices
2) The path length of their networking stacks
3) Their CPU loading
4) The loading of the network path between them
5) The settings on the NICs

Some NICs have interrupt coalescing parameters that can minimize the number of interrupts, and increase aggregate throughput, but at the expense of unloaded response time. An example with which I am familiar is on Linux with the e1000 driver which configures the NIC initially to do no more than 8000 interrupts per second and so gets no more than ~8000 on a netperf TCP_RR test.

[Bill Hassell]

The network path will likely affect the ping response times the most. If the systems are on the same physical network (2 ports on the same switch) then ping times will be just a few milliseconds. Toss in a router to another network and the time will increase a bit depending on how busy the router might be. If the two computers are separated on a WAN, then ping times can be dozens to hundreds of milliseconds. And occassionally, the ping response can be affected by a massively busy server. The server had a workload of 50 to 75 with a system load of 99% (due to a kernel bug) and ping times were 700ms to over 2000ms.

The question implies that there is something you can do to fix a long ping time. If the computers are responding correctly, then you have to remove the components between them that are "too slow". Naturally, over a WAN or the Internet, this isn't possible. Internet ping times can bounce all over the place as a different path is often selected by routers and load balancers.

What makes more sense is to find 'normal' values and then use these values as guides.

[wah]

I think that you should be troubleshoots end-node connectivity by tracing the paths between the two end nodes on the network. Some visual network monitor tools can help you to evaluate the network usage condition, e.g. 3Com® Network Supervisor.

Also, ensure that the DNS configuration is correct (e.g. likes '/etc/hosts'. when the system performs a name lookup operation, it normally searches the hosts file.)

How about the "telnet" to this host? if the system take a long time to establish the telnet session between the client and server, it should be related to DNS configuration.

[Muthukumar_5]

check this,

traceroute

to find where the problem is starting. It will take more time at that point.


Another reason is hostname to ip-address resolvation. Check up resolvation information as,

hosts: entry details in /etc/nsswitch.conf file.

dns entries for /etc/resolv.conf and /etc/hosts file.

--
Muthu