[Info-vax] VSI has purchased PERFDAT from HPE

[Norm Raphael]

-->
From: IanD via Info-vax <info-vax at info-vax.com>
To: info-vax <info-vax at info-vax.com>
Sent: Thu, May 16, 2019 9:05 am
Subject: Re: [New Info-vax] VSI has purchased PERFDAT from HPE

I have and I found the software really good with the ability to incorporate various sources of data as well, like storage etc

Compared to T4, it's light years ahead

I also used PSPA many years ago, which I must say I enjoyed. I particularly liked the advisory aspect of it because it had the ability to bundle certain events together and classify them rather than measure strictly by typical VMS performance metrics. Hot files I remember being a helpful measure (yes, you can get this from cache data).
This was like nearly 12+ years ago I think now, so I'm struggling to remember much about it. 
I used PSPA running over X Windows and some of the analysis collection could take 8+ hours which caused all sorts of problems because we were behind a jumphost which was set to pass through X Windows but the port was only allowed to be open for 8 hours so it would get closed if it didn't detect traffic, which it often didn't

Perfdat compared to PSPA is snappy, integrated and drags in storage data easily so you get a unified view across VMS and storage. I seem to remember it also could bring in switch data too.
In terms of visuals, I remember PSPA was nicer looking and that's with many years between when I used PSPA and in the past 3 or so years using perfdat, so my memory might be favoring PSPA wrongly

Perdat support was very good. I contacted them by email a few times and got concise, specific answers to an issue I had. The response was quick too. I did however get the feeling that it might have been a 1 person show but that is only my impression based on, well, based on, my impression only. No facts to back up that view at all

So my experience history with VMS performance monitoring tools has been

1. Native VMS monitor and trying to extract data myself using cludges, scripts, voodoo and luck

2.Then used PSPA and was amazed at what a seemless all encompassing data performance reporting tool could do. I still hold foud memories of PSPA, probably because of the difference between a ready made solution and having to wrangle data out of the native VMS tools

3. Then I used T4 because the company I worked for didn't want to pay for anything and I wasn't prepared to smash my head against the wall and develop all those cludges and half arsed roll your own performance monitoring tools again. T4 it's ok if you have nothing but it's piggy backed on Excel for some bits and is hardly a highly polished product. It is however free and great for free

4.Then I used perfdat and was rather surprised at how functional it is.
It doesn't have all the performance recommendation components that PSPA had (or I didn't use them) and you have to find these issues out yourself by getting to see the trends in your data.
However, having it all displayed visually makes this a million times easier than just purely numeric representatiion because you see various different areas all together on a single graph. Perdat is very good at times series collation and plotting and the fact that you can drag in storage metrics as well makes it very good. We didn't use any switch related data though which I believe it can also work with

>From what I remember, I think licensing was fairly cheap for perfdat? I am very reluctant to quote a figure because I didn't do any of the purchasing, but $1500/year seems to ring a bell but I'm probably wrong on this

However, with the horsepower available on most systems today, how much one needs such in-depth performance tools I don't know. 
It's not like when I used PSPA and we had systems, applications, databases that required you to constantly be looking at performance trying to wrangle every ounce of performance out of your system, these days it's about investigating a fairly rare event to determine what caused a problem and to make sure it doesn't happen again versus the need for continual tuning

I'd rather see perdat and monitoring type tools which are snapshot driven have event based functionally added, like when I used to use dectrace for looking into ACMS and RDB performance. I found for that sort of drilling down, dectrace was far more useful

Perfdat is awesome for seeing trends but things like dectrace are far more useful for getting to the crux of a specific performance hit

Horses for courses I guess

Sorry for rambling, I have fond memories of the various tools I used for VMS performance monitoring over the years. The highlight would be using dectrace for details, PSPA for lovely visuals, perdat for ease of performance data collation, T4 for being free and monitor and self derived tools for teaching endurance and wishing that VMS system monitoring was easier to work with in terms of data output to other performance investigation tools.
Now a lot of places spit data out to Splunk and use that for log scanning, performance trending and event notification
<--

Well, back in the day, I used PAWS from Perfcap Corporation http://www.perfcap.com which is a multi-platform tool that include/ includes OpenVMS. IMHO it's worth a look.
_______________________________________________
Please reply to norman.raphael at ieee.org
 Norman F. Raphael"Everything worthwhile eventually degenerates into real work." -Murphy