Fitting Workload



Post anything related to mainframes (IBM & UNISYS) if not fit in any of the above categories

Fitting Workload

Postby pinkstonmp » Sun Jul 04, 2010 5:04 pm

Lets say I have two pieces of business, one that is currently running in one of our mainframe environments and another that is proposed but from the same client. What do I need to validate to make sure these two pieces of work could be put together on the same mainframe environment.

????

1. MIPS Rating
2. Current MIP usage
2. z/OS version

Anything else key????????
pinkstonmp
 
Posts: 2
Joined: Sun Jul 04, 2010 4:59 pm
Has thanked: 0 time
Been thanked: 0 time

Re: Fitting Workload

Postby steve-myers » Sun Jul 04, 2010 8:54 pm

This sounds a lot like a class homework assignment.
steve-myers
Global moderator
 
Posts: 2105
Joined: Thu Jun 03, 2010 6:21 pm
Has thanked: 4 times
Been thanked: 243 times

Re: Fitting Workload

Postby dick scherrer » Sun Jul 04, 2010 8:59 pm

Hi Steve,

It quite possibly is - but this is a large part of why this forum exists. . .

From the top of the page:
A Help & Support Forum for Mainframe Beginners and Students
Hope this helps,
d.sch.
User avatar
dick scherrer
Global moderator
 
Posts: 6268
Joined: Sat Jun 09, 2007 8:58 am
Has thanked: 3 times
Been thanked: 93 times

Re: Fitting Workload

Postby pinkstonmp » Mon Jul 05, 2010 1:21 am

Was just looking for some pointers not a lecture, guess I will look for a different forum where respondents actually have the best interest of the board at heart.
pinkstonmp
 
Posts: 2
Joined: Sun Jul 04, 2010 4:59 pm
Has thanked: 0 time
Been thanked: 0 time

Re: Fitting Workload

Postby dick scherrer » Mon Jul 05, 2010 2:05 am

Hello,

Was just looking for some pointers not a lecture,
If you believe that somewhere between your first 2 submissions there was a lecture, you will have a very frustrating time in IT (or anywhere else in the business world).

As i mentioned to Steve, this particular forum (ibmmainframeforum.com) is for students and other beginners. Capacity Planning (which is what you asked about) is not given to "newbies" - except as possibly a class exercise - and the "rules" are defined as part of the exercise.

When you post a question that is classwork or if you are simply curious, it is best to say so in the initial post. . . It can make a big difference as to who might respond as well as what kind of reply is posted.

There are some forums out there that are very friendly - unfortunately 1/2 to 3/4 of the technical replies are between questionable and simply incorrect. . . And an inexperienced person with a question will not know the difference.

And, yup, i guess this was a lecture. . . :)

Good luck,

d
User avatar
dick scherrer
Global moderator
 
Posts: 6268
Joined: Sat Jun 09, 2007 8:58 am
Has thanked: 3 times
Been thanked: 93 times

Re: Fitting Workload

Postby steve-myers » Mon Jul 05, 2010 4:18 pm

I'm reluctant to discuss the capacity planning issues. The points raised in the initial post are far too simplistic. Applications seldom, if ever, have a "MIPS rating" because their CPU requirements are too dependent on the data they have to process. In any event this is just one element of at least three elements in capacity planning. The other two elements are I/O usage and storage usage. The "MIPS rating" itself is too simplistic, particularly if the application must respond to input very quickly, even if the apparent CPU requirement is not very great.

In any event, most of the time z/OS is superb at absorbing new workloads without falling apart. "Mainframe" type computers have been running at very high CPU utilizations since at least the late 1970s when enough storage became available so that multiple applications could really run together.

The z/OS release is not usually a very big issue with applications directly prepared by most users. Most users are slow to take up significant new OS features, and IBM is usually very careful to make sure that application prepared for z/OS Release 1.11 (for example, the most current z/OS release) will run properly under earlier releases unless the application is using a feature that is new in z/OS 1.11 and is not supported under an earlier release.

Vendor written software is a different story for several reasons. Most major vendors are running the most recent z/OS releases, and usually have access to z/OS releases before they're generally released. This serves the vendor, IBM, and their common customers well. The vendor can be ready with release dependent updates when IBM sends out a new release. CA, for example, will generally issue a press release the day IBM sends out a new z/OS release that CA's products are ready for the new release, and they are usually telling the truth, though sometimes not the whole truth. The vendor can often spot new release problems in these early releases and get them repaired, even if the problem has little or nothing to do with the vendor's products. The other problem is the vendor can get fixated on current releases and put in product updates that require the new release, even if their customers are not running the new release, which is often the case.
steve-myers
Global moderator
 
Posts: 2105
Joined: Thu Jun 03, 2010 6:21 pm
Has thanked: 4 times
Been thanked: 243 times

Re: Fitting Workload

Postby Robert Sample » Sun Jul 11, 2010 9:03 pm

You have asked a question that is extremely difficult to answer not just because you have simplified a very complex subject way too much, but also because you have not provided anywhere near enough information. For example, if the new workload involves system products not currently running in the LPAR (such as CICS and MQSeries, for example), you have to factor in the increased system overhead for the additional system products as well as the application load. And where do you get the numbers for the application load from? Another system such as Unix does not directly translate into z/OS performance. Even going from another z/OS system may create some differences since the application may have been tuned on the other site and you don't have those years of tuning expertise to make the same adjustments to your system. Furthermore, there are many dimensions to capacity planning that you didn't mention -- channel contention, device contention, paging, backup and recovery, disaster recovery, tape library capacity, batch windows, etc, etc, etc.

Basically, while you can do some estimating, you must be aware that all you are doing is generating a guess. Until you actually get the new workload on the box and see how things play out, you cannot know for sure that the capacity is there.
Robert Sample
Global moderator
 
Posts: 3719
Joined: Sat Dec 19, 2009 8:32 pm
Location: Dubuque, Iowa, USA
Has thanked: 1 time
Been thanked: 279 times


Return to All other Mainframe Topics

 


  • Related topics
    Replies
    Views
    Last post