CMG Forum

Capacity Management Certification

Fri, 5 Mar 2010 17:43:45 GMT

Can anyone recommend a capacity management certification or training track? Does anyone have experience with the ITIL Practioner Certificate in Capacity Management offered by Metron? Thanks Nell ...

Re: UNIX Processor Q Length vs Processor Busy

Mon, 22 Feb 2010 19:40:05 GMT

Tim, Thanks for the educational response. It seems to me that many performance problems are sudden. A minor new software version goes live, an additional user community comes online, a query goes nuts. So knowing the recent typical range of values for performance metrics and comparing them to current values may be more useful than carefully planned thresholds. Time series analysis might be ap ...

Re: List of parameters?

Sun, 21 Feb 2010 18:14:42 GMT

Rather than a list, which is so vendor-hardware-config dependent...consider this: computer systems have four basic resource categories of interest for your purposes - CPU-related metrics, Memory-related metrics, I/O-related metrics, and Network-related metrics. Evaluation of the capacity of these resources is generally focused on measures of utilization whereas evaluation of service is focused on ...

Re: Building Capacity Mgmt Software Requirements

Sun, 21 Feb 2010 15:50:27 GMT

These are very good responses. I would like to offer one more suggestion, and that is to consider multiple, inexpensive, and simple tools rather than one large and complex toolset. Integrate them into your solution also as simply as possible. Some of the larger and more comprehensive products demand a long learning period for the capacity planner because of their complexity and the tasks related ...

Re: Capacity Management evaluation

Sun, 21 Feb 2010 15:32:36 GMT

Yes, I agree with Shana. These are large and expensive tools and they don't come with fully featured trial versions that you simply download with a zip file as you would in the Windows utility world. For these high-end products you have to start a 'relationship' with the vendor and often they want to be closely involved in your evaluation process. This can be a good thing since they have a big ...

Re: WLM Resource Capping

Sun, 21 Feb 2010 15:23:07 GMT

Sometimes your goals for WLM can't be achieved or are too aggressive, and in those situations, a more modest goal has a better overall effect. You can have too many high level goals for too many tasks. Make a (10% - 80% - 10%) distribution with 'most favored' and 'least favored' tasks in the upper/lower 10%. Also, with WLM, you can have too many service classes and policies for adjustments and m ...

Re: UNIX Processor Q Length vs Processor Busy

Sat, 20 Feb 2010 17:38:48 GMT

I think your objective is not an analysis of Q length and busy, but finding patterns that predict "trouble". Based on the other comments, clearly, Q length and CPU Utilization in isolation or in comparison to a fixed threshold of 'not troubled' values is not useful for your purposes. The analysis becomes one of hypothesizing multiple causal models to 'explain' measured values rather th ...

Re: IBM MF MIPS offload to open environmet

Sat, 20 Feb 2010 16:13:56 GMT

I've seen work go both ways, i.e. mainframe to distributed and vice versa. The technical reasons for hosting on the mainframe are (1) Availability, (2) High relative I/O content of workloads, and (3) workloads that can exploit I/O parallelism. These are dimensions of value, not cost, since cost per unit of work is almost always higher for mainframes. For apps heavy with Java code, etc. specialty ...

Kudos

Fri, 19 Feb 2010 20:49:22 GMT

Wanted to personally thank Margaret Greenberg and the entire Measure IT editorial staff for the Mario Morino story in this month's issue of Measure IT. Mario was a great influence on nearly every CMG 'dinosaur', and it was great to see him acknowledged once again. Dr Bernie ...

IBM MF MIPS offload to open environmet

Fri, 19 Feb 2010 20:32:40 GMT

Hello, We are evaluating a process of offloading some MIPS (~3000) of CICS COBOL from Z10 to Clarity (CICS compatible) and Microfocus COBOL on HP Superdom (INTEL ITANIUM) machine. I know that proper sizing depends very much on the specific workload. But I wonder if there are some numbers/benchmarks/calculation methodology for MIPS to ITANIUM cores ratio. Avi ...

SYSPLEX environment

Fri, 5 Feb 2010 14:24:12 GMT

In a SYSPLEX environment what are possible causes for high priority critical batch jobs to wait for CPU resources more than lower priority normal class jobs? ...

CMG marketing, e.g. wikipedia

Thu, 7 Jan 2010 10:42:59 GMT

Hi, At the latest CMG meeting there was some discussion of the ways in which CMG could be brought to the attention of more people. One avenue is the Wikipedia page at [url:34297iz6]http://en.wikipedia.org/wiki/Computer_Measurement_Group[/url:34297iz6] though I noticed that it is a bit outdated (e.g. refers to "CMG'08 will be held in Las Vegas"). I would encourage anyone to ...

Dutch NL CMG to start

Thu, 7 Jan 2010 10:35:43 GMT

Hello to all, and especially dutch speaking members. I am happy to announce that a small team is formed to set up the dutch CMG chapter. If you speak dutch, and/or live in the netherlands, i urge you to join the discussion at [url:1ke07yt7]http://cmg-nl.blogspot.com[/url:1ke07yt7] and subscribe yourself to the mailing list by sending a blank message to [url]mailto:nlcmg@getresponse.co ...

Re: Clustering already highly available servers

Thu, 7 Jan 2010 10:24:09 GMT

The basis statistics goes like this. If the failure modes are completely independent, you can multiply the probabilities for downtime. So if one server is down 10% of the time, then the probability of two being down at the same time would be 10% times 10% = 1 %, giving an uptime of 99%. In your case that would lead to an uptime of 99.999999 %, although it is very unlikely that the failure modes ...

Clustering already highly available servers

Tue, 5 Jan 2010 16:05:41 GMT

Some modern UNIX servers are claimed to be four nines reliable (99.99% -IBM P series) I.e highly available. If I cluster trwo together what improvements in reliability/availability will i get? I presume I will not get mainframe levels (5 x 9's), but can't seem to find out the exact maths to use. ...

WLM Resource Capping

Tue, 22 Dec 2009 21:55:49 GMT

In our installation, a couple of service classes were assigned to a single resource group (limit set at SU's/sec). Both the service classes were defined with the importance of 2 and velocity goals of 50 and 20 respectively. The high velocity service class is not meeting it's goals when the resource group gets capped. The low velocity service class has more tasks than the high velocity service c ...

Re: Reporting - Should it be Easier?

Tue, 15 Dec 2009 13:29:26 GMT

- The messy, time-consuming part of reporting is marshalling the correct data. So anything that makes it easy to get the right data would be helpful. - Having a customizable template with all the right context would be helpful. For example, a place for the object of the report (server name, application, transaction) , the time frame, metric selected, author, source of data, time the report was pr ...

Re: SSDs and the end of seek time

Mon, 14 Dec 2009 20:05:00 GMT

Hi Carl, I just started looking at these boards and noticed this topic. Dr Gilbert Houtekamer (who co-wrote the MVS I/O Subsystems book with Pat) presented a paper and case study at CMG last week on SSD's. I assume you can get the paper off of the CMG site but let me know if not and I'll send you a copy. john.baker@intellimagic.net