I want to monitor application response times across the enterprise. I have investigated synthetic transactions and various commercial programs, but it seems to me that the best way in my situation is to write monitoring features into the software we use (which is mostly in-house developed).

I have read that there is an Application MIB, but all the literature I found on it is really old.

Does anyone know if that MIB actually went anywhere, and whether I can write an RFC-compliant snap-in for the Windows 200 MIB?

Thanks, Ben

-- Ben O'Loghlin (ben@tenfold.com.au), July 31, 2002

Answers

Ben, Rather than write your own monitoring features, you could define and create your own metrics, collect the output of the metric across the enterprise at various established tiers. Of course these tiers in the enterprise will be the deciding factor in definition of the metric. Such as end-user response, network chat, server response, etc. Compuware Corporation's, Vantage Product Suite has this ability, either end-to-end or at a point solution. Their professional services may also develop a rather inexpensive solution, via this Custom Counter package, if it's just one application you wish to monitor. Going down the road of SNMP is really your best choice for an in- house solution, however you will find lots of out-of-date info in this segment, since it's a rather cylindrical technology, it's industry support comes and goes every few years.

-- Mike Korzenowski (mrbaxtor@yahoo.com), August 01, 2002.

Are you familiar with Application Response-time Monitoring or ARM? I believe this may be what you are describing. Essentially it is timing transacations at the point of initiation within the application code. This is achievable for applications that you have access to the source code. However for many commercial apps synthetic transactions will be the only way to go.

Keep in mind also that when performing application response time monitoring for a mult-system application you should approach monitoring utilizing a portfolio approach. That is to say, you may use a combination of monitoring techniques based on what it is you are monitoring for. If it is end-user experience for SLA's then end- to-end type transactions monitoring (a.k.a "robotic" or "synthentic") can be a very viable solution based on ease and cost to implement/maintain.

If transaction level monitoring for capturing diagnostic data for use in potential problem situations, then a more extensive performance monitoring solution will be called for which should cover the various components participating in the transaction(s).

-- Wiley Vasquez (wiley_vasquez@bmc.com), August 04, 2002.

Ben, I was co-chair of the IETF Working Group for Application Management. That's the group that developed the applmib and sysapplmib. I agree with Wiley that ARM is probably closer to what you are looking for. However, in general, I recommend that IT organizations develop their own monitoring solutions. Normally, in the long-run a commercial product will be more cost effective. Ultimately, the right answer depends upon the specifcs of your environment.

Rick Sturm

-- Rick Sturm (sturm@enterprisemanagement.com), September 24, 2002.

Speak to Sithumina I Solutions to get the answer. Try www.sithumina.com

-- Nivantha (nivantha@optusnet.com.au), April 20, 2004.

are you still looking for application response time in REAL time and not synthetic or playback which only give you a feel of what happened after the horse has bolted?

go to www.end-endsoftware.com

you will be surprised to see what you are looking for

-- Brian Peck (bpeck@end-endsoftware.com), May 11, 2004.