Review of SOS
Shawn M. Gordon
SOS is a system performance monitoring package from Lund Performance Solutions that is available for both MPE V and MPE/iX operating systems. While there have been various products of this type for the HP 3000 over the years, SOS is showing itself to be one of the survivors, as well as being one of the easiest to use, and easiest to understand. As I said in my review of Performance Gallery from Lund Performance Solutions, your opinion of a system performance monitoring tool is almost as personal as your choice of text editors. Having used about four different system performance tools I am in a good position to make a comparative judgment.
I tested SOS on my HP-3000 series 37 with 2 meg of memory and 150 meg of disk space. Unfortunately the only way to measure the hit of SOS on the system, was with SOS. I am sure I got a fair measure however.
As an example of how important knowing what is happening on your system, I want to tell you a story that another system performance software provider related to me. He had been called into a location that HP had been doing performance monitoring for a month. What HP had determined was that they needed a bigger CPU, and more memory. Well, my friend came in and ran his performance tool for a couple of days, and guess what he found out. Turns out their applications required that the user log on again to access each application, this resulted in hundreds, if not thousands of logons a day, and as most people are aware, logons are the most CPU expensive process you can do. So by changing their system to use a menu processor and avoid the logons they were able to quickly and cheaply fix their problem. This should serve as a valuable example to how a performance monitoring program can pay for itself, now, on to the review.
You can run SOS right out of the box as it were. You can instantly get an overview of your system with the first screen, as seen in figure 1. What is very nice here, is how SOS will immediately tell you what the ‘Hog’ process is, that is, the process that is the biggest consumer on your CPU. You are also presented with a function key that allows you to immediately zoom into the ‘Hog’ and see what it is doing, as seen in figure 2 (due to graphic type conversions, some of the english text may not come out on the screen snapshots). What you should also note, is the use of fairly easy to understand english sentences that describe the current state of your system for each measurable item. This makes it so easy to tell where performance problems may reside and allow you to quickly go after them.
Once you are in SOS it is a simple matter to drill down through various types of information using the function keys. The function keys are highly dynamic, and change to whatever is appropriate for your given position in the program. There is even an appedix in the back that shows a map of the function keys, and what key get’s you to the next level, this was really a nice touch. The HELP function key is available from anywhere in the system. It is also very quick and easy to get back to your starting point. Typically pressing function key eight will keep taking you back until you finally exit.
Most performance monitoring software will provide basically the same data with greater or lesser degrees of difficulty finding it. So what I am going to do is go over some of the screen snapshots briefly, which are by no means a comprehensive list of the available screens, but just a little cross-section, then I am going to go over some of the more esoteric features, to try and give you a good overview of the software. Keep in mind that all these snapshots are on a small Classic system, you will get a bunch of other information typically on a Spectrum class machine.
Figure 3 shows the CPU Detail, and all the different ways the CPU was being consumed in the last time slice. Figure 4 shows the Disc usage detail, with my system having only one drive this is a fairly short list, but it is a comprehensive overview of the disc usage. Figures 5, 6, 7 and 8 are really some of my favorites, and from what I have seen they are unique in the system performance tool arena. The provide trend history for each measurable unit from the time you started up the program, to the time you decide to display it. Function key seven provides a facility for searching through this list if it get’s to long, but the great thing here is being able to get an immediate trend without having had to set up some background process to collect it before hand. This helps those of us who just need to look at it for a half hour or so to find out what the heck is going on. I really found these screen useful.
Figure 9 depicts a nice overall picture of just about everything in one screen. From this point there are all sorts of other options that you can access or drill down to. In the are under “System Performance Advice” is where you would see the easy to understand, plain english recommendations from SOS, as I said earlier, this may have been lost in the graphics conversion to monochrome. The manual describes each of the parameters that is presented, so if you don’t understand what some metric is telling you, you can just look it up and get a nice clear explanation.
The historical analysis module allows you to accumulate performance trend information and then view it in a number of ways. You can go through the data interactively, print reports, export selected data to an ASCII file to download to a PC, graph data, determine averages based on user-definable criteria, and exclude specific periods, days and dates from analysis. There is a HOLIDAY file that allows you to define dates that are to be used as holidays
The reporting process uses something that LPS call’s ‘The Report Card Module’. This is an off-line analyzer of logged SOS performance data. These reports are easily selected through the menu that is presented, and the reports are very easy to read and understand. This is really a great little tool, and the advice reports are terrific, they give you nice clear messages about what the problems were during the interval, which should lead you quickly to a fix.
Application Workloads are another unique feature of SOS. This allows you to create a named group of logical entities that you want to track, an example would be;
FINANCE !NAME of group (max 12 characters) BOTH !Group TYPE (JOB/SESSION/SYSTEM/BOTH/ALL) PROG=AP001.@.@ !PROGRAM SPECIFICATIONS PROG=AR001.@.@ !Prog specs are optional USER=BILL,MGR.FINANCE !USER SPECIFICATIONS (optional) USER=LAYNE,MGR.ACCT@ !Partial "@" is valid PRSLO=5.0 !Prompt Response time Service Level of 5 seconds !At least one blank line (required) EDITORS !Begin defining a second workload SESSION !Only interactive ones count (SESSION) PROG=EDITOR.PUB.SYS PROG=QUAD.UTIL.SYS LDEV=20 - 45 !Only prots 20 - 45 included FRSLO=2.0 !First response time service level of 2 seconds
With this set up you can be alerted when there are various problems for a workload, display information about the response and consumption by the workload, as well as being able to use the Report Card to view and extract information based on the workload. Overall, this is a very powerful feature.
The last unique feature I am going to cover (but not the last in the software) is the Advice Module. The purpose of this module is to provide plain English “one-liners”. Once you get the message number you can get more information if you need it from the documentation. You can get in and modify the Advice Module messages if you want to tailor them to your environment. You can also provide a list of TELL recipients, or a TELLOP for various messages when certain criteria are hit. This gives you an ‘alarm’ type feature so you don’t have to wait till the users call and complain.
I have really only scratched the surface here, there is all sorts of information provided by SOS, as well as the ability to do look at what users have a file open, stack traces on a particular pin, MPE dispatcher queue priority altering, and many many more.
Usability and Installation
The software uses the standard installation of restoring a job, streaming the job, reloading the tape, and you are done. This all proceeded without incident, and was quickly done.
As I have said, the software is extremely easy to use, even for the novice. It would behoove you know at least a little bit about the computer system, otherwise, even with all the english descriptions, you won’t know what is going on.
SOS chugs right along giving you all the information that it can, and it seems to do this without crashing or crashing your system. This is a more impressive feat on a Classic where there is no AIF (Architected InterFace), or MI (Measurment Interface) to access to get system information. You have to get it all yourself, and that can be pretty darn tricky. So in other words, this was solid as a rock. It’s been nice to review so many high quality products lately.
From what I can tell SOS is as easy on your system as it can be. If you set a scan rate down to a few seconds, or constantly hit the return key to get a new scan, you will have a negative impact on performance. As long as you use the product in a normal manner, you will see very slight overhead involved with SOS. Any performance tool is going to impact your performance if you try to scan to much, this is due to the overhead involved in collecting the information, and the fact that they run in a high dispatcher queue, so they will always get preferential treatment for a slice of CPU time.
Support and Documentation
The support from LPS has always been good for me in the past, and SOS was no exception to this. Everyone I talked with was knowledgeable and helpful. I didn’t have any problems with the software, just some general questions. My only real complaint on this part is the lack of an 800 number.
The documentation is well written, informative, and easy to read. LPS usually tries to explain in english why things are what they are. This helps take you from being a monkey that reacts to certain alarms, to someone who truly understands what is happening, what causes it, and how to fix it.
There is a minor cosmetic bug in the HELP facility however. LPS uses the alternate character set to do line drawing to help make the screen readable. What I found was, when I went into help, they weren’t sending the cursor to the bottom of the screen, so you end up with some of the help text overwriting some of what is on the screen, and sometimes being converted to graphic characters due to the line drawing.
I do want to say however, that the help subsystem is VERY nice. It looks and acts like the HELP you get in MPE, but it is more complete with better drill down keywords. Now maybe that doesn’t excite you, but it is consistent with what everyone should already be familiar with as well as being complete and easy to use.
I’m sorry if I make anyone mad with this statement, but of all the system performance tools I have used over the years, this one is by far the most comprehensive, and easy to use. Even if you don’t understand the nut’s and bolts of system performance, this tool should allow you to identify, and hopefully fix what is wrong with your system. The use of clear english descriptions, as well as a massive, comprehensive on-line help system make this an amazing combination of power and ease of use.
The Advice Module is really an amazing piece of work, and is just one of the things that make this product shine. The high convertibility allows you to make it available to people who you may not normally want in such a powerful utility. And of course the historical analysis helps bring it all together with all the information necessary to tune your system into a lean, mean, processing machine.
If you have gotten the impression that I like this product, you are right. Buying something like SOS before you need it is being proactive and keeping your user community from lynching you. The problem is, no one seems to see a use for it until after they start having problems. In all my years, I have never seen a system that didn’t start having problems at some point, whether due to new demands being placed, or buggy programs. You will have to evaluate for yourself if you can budget for it, but I highly recommend it.
SOS version E.02
Lund Performance Solutions
440 First Ave. S.E., Suite 3
Albany, OR 97321
FAX : 503-926-7723
Call, write or FAX to receive a demo of the software. HP-3000 software on any tape format you require, one indexed 234 page manual. Software price is $1895 – $7195, support is $540 – $1000 depending on CPU size. They also offer free annual performance reviews.