Re: Network Monitoring

Sure. Say this is a trading floor environment with external brokerage firms. If there is latency normally of 20ms and there is an issue that causes that to be 50ms or more for an amount of time, trading information may be invalid and someone could be looking for compensation. Today, Reactive troubleshooting must now be done to figure out what happened and traced back to network, server or application. With the 5 minute trending, that is a process that may or may not produce a good answer. They are aware of the impact of constant polling on the network.

Sent from my iPhone using mb"Box Mail - http://mboxmail.com

________________________

On 2009-05-09 08:21:46 -0400 Dale Shaw <dale.shaw_at_gmail.com> wrote:

>
> Perhaps you could elaborate on the "Milliseconds delay can cost money
> in this case" aspect, so people fully appreciate the requirement. It's
> not as though collecting data every 5 minutes means you lose
> visibility of everything that happens in between polling cycles, but
> clearly interface stats like bits per second get averaged out.
>
> Have you considered the impact of such frequent polling, both in terms
> of management traffic and managed device performance overhead?
>
> cheers,
> Dale
>
> On Sat, May 9, 2009 at 9:33 PM, Bill Mckenzie <bmckenzie_at_hotmail.com> wrote:
>> The critical needs are ability to scale but just as important, polling in
>> 1/5/10 second intervals. Milliseconds delay can cost money in this case, so
>> it's very important for constant polling.
>>
>> On 2009-05-09 07:22:22 -0400 Dale Shaw <dale.shaw_at_gmail.com> wrote:
>>
>>> We use Statseeker in a couple of managed networks --
>>> http://www.statseeker.com
>>>
>>> It does a really good job, scales very well and is super easy to
>>> integrate, but data is collected and reported in five minute
>>> intervals.

Blogs and organic groups at http://www.ccie.net
Received on Sat May 09 2009 - 08:31:47 ART