Unfortuantely, the delay isn't in the ping send, but in the receive.
Putting the send time in the packet isn't much help. Even if we could
have senders put in the time (and it was accurate) it wouldn't help.
We don't know that Zenoss would have a problem collecting ping latency,
but we suspect that it might. It would be interesting to capture the
latency to see how good it is.
Zenoss gets a lot of it's scalability by using massive amounts of I/O
parallelism. So when we ping hosts, we may be pinging 50-100 at a
time. That is, there are dozens of outgoing packets sent while we are
waiting for responses. This is just spiffy if you are just trying to
validate the state of your network and you have 1000-5000 devices. If
you increase your OS buffers as suggested in the admin guide, you will
get amazing performance. But you may get an artificially high latency
as zenstatus grinds down the ping responses from 100 machines. Let's
say it takes 0 time for a device to respond. Lets say it takes
zenstatus 0.005 seconds to perform the system call that fetches a ping
packet from the OS. With 100 responses, it will take 0.5 seconds to
pull the last response out of the OS buffer. That's an unacceptable
delay if you are measuring ping latency, but completely acceptable if
you are trying to validate 5000 individual devices in 60 seconds.
I have a feeling that "latency" ping measurements will need a different
sampling method. Either less frequently or more distributed.
Fortunately Zenoss performance collection can be configured to do both.
-Eric
John Neiberger wrote:
The only better way I could think of it would be to use a ping utility
that uses timestamps. Then all you need to do is record the delta
between the pings and the responses based on the timestamp in the
packet.
On 7/28/06, Erik A. Dahl <[EMAIL PROTECTED]> wrote:
We thought about this more. The problem with using the time that we
have from zenping is that it is pinging many devices at once
asynchronously. This means that any ping instance may sit in the OS
buffers for a while before we get to processing it. Therefor, any
time we would report would have a high degree of error as a
percentage of the time measured. The only way I can think to measure
would be to ping one device at a time. Anyone have other thoughts?
-EAD
On Jul 28, 2006, at 3:50 AM, Emmanuel Gomez wrote:
> I'm more of a prospective than current user of Zenoss. That said,
> I would
> be very interested in collecting and reporting latency data.
>
> My interest is also for the purpose of managing a VoIP rollout.
>
> Thank you,
>
> Emmanuel Gomez
>
> -----Original Message-----
> From: [EMAIL PROTECTED]
> [mailto:[EMAIL PROTECTED] On Behalf Of Dennis B. Hopp
> Sent: Wednesday, July 26, 2006 3:40 PM
> To: General discussion of using zenoss system
> Subject: Re: [zenoss-users] Performance Metrics
>
> On Wednesday 26 July 2006 07:49, Erik A. Dahl wrote:
>> john,
>>
>> We have the history of ping up/down but not latency. I have never
>> found it useful. We have the data bug aren't currently logging it.
>> Would other people be interested in this info?
>>
>
> I would be greatly interesting in this. Specially since we are
> getting
> ready to roll out a lot of VOIP.
>
> --Dennis
> _______________________________________________
> zenoss-users mailing list
> [email protected]
> http://lists.zenoss.org/mailman/listinfo/zenoss-users
>
>
> Confidentiality Notice: This e-mail message including attachments,
> if any,
> is intended only for the person or entity to which it is addressed
> and may
> contain confidential and/or privileged material. Any unauthorized
> review,
> use, disclosure or distribution is prohibited. If you are not the
> intended
> recipient, please contact the sender by reply e-mail and destroy
> all copies
> of the original message. If you are the intended recipient, but do
> not wish
> to receive communications through this medium, please so advise the
> sender
> immediately.
> _______________________________________________
> zenoss-users mailing list
> [email protected]
> http://lists.zenoss.org/mailman/listinfo/zenoss-users
_______________________________________________
zenoss-users mailing list
[email protected]
http://lists.zenoss.org/mailman/listinfo/zenoss-users
_______________________________________________
zenoss-users mailing list
[email protected]
http://lists.zenoss.org/mailman/listinfo/zenoss-users
_______________________________________________
zenoss-users mailing list
[email protected]
http://lists.zenoss.org/mailman/listinfo/zenoss-users
