BMC Communities Banner

Mainframe Revolution

2 Posts tagged with the reduce_cpu tag

zIIP it

Posted by Jonathan Adams Dec 4, 2009

Guest post by Bronna Shapiro, Solutions Marketing Manager

 

Fully loaded mainframe processing can be expensive. IT organizations rarely have extra money to spend, and our economy has every shop tightening its belt.

 

So when a problem comes along, you must zIIP it.

 

Offloading work to zIIP engines can result in savings of up to $7,000 per MIPS because software running on specialty engines is exempt from related charges such as license costs, upgrades, and maintenance fees.

 

The choice seems obvious: maximize the use of specialty engines to reduce cost and free up general purpose processors for other work. However, because only certain types of processing are eligible to execute on these specialized engines, you must identify which workloads are eligible for zIIP processing and then move those workloads. BMC products can help guide you in identify the best candidates for using zIIPs.

 

Offloading CPU cycles to zIIPs during peak hours will likely give you the best bang for your buck. Offloading work overnight, when you have processing capacity available, may not provide significant payback.

 

BMC products automatically exploit zIIPs. In one case, a customer was able to reduce 50% of the MIPS required for monitoring z/OS.

 

Remember that everything comes at a cost. Look at your heaviest workloads and capacity needs. Consider the costs of purchasing the zIIP engines and rewriting applications to take advantage of them. Each zIIP you install still requires maintenance and power. Work with your vendors to understand what products are most suited for zIIPs. Weigh the costs against the potential savings.

 

And when appropriate, zIIP it. zIIP it good.

The postings in this blog are my own and don't necessarily represent BMC's opinion or position.
| More
0 Comments Permalink

Guest post from James Davis, Customer Support Representative

 

Lowering CPU usage and costs are hot ticket items these days. To help reduce CPU usage, vendors and IBM work together to make mainframe processing more efficient. Here is one example.

 

Shared queue monitoring in WebSphere MQ for z/OS involves communication with DB2, and that the resource cost of this communication can be as high as 15% of total CPU utilization. BMC discovered that a "reset queue statistics" command for a private queue causes a query to be executed against DB2 to satisfy the potential for a generic name.  This happens for both shared and private queues, even though private queues have no relationship to DB2. The cost of the unnecessarily executed SQL can be very high, depending on the number of shared queues defined and queues selected for monitoring.

 

BMC reported this issue to IBM, who accepted it as an area for improvement and issued a fix. When the APAR is made generally available, users can expect significant savings if shared queues are present in their monitored environment.

 

By working together, IBM and BMC were able to optimize the interaction between DB2 and WebSphere MQ when resetting queue statistics...and reduce CPU usage for monitoring WebSphere MQ queues by 66%!

 

  The postings in this blog are my own and don't necessarily represent BMC's opinion or position.
| More
0 Comments Permalink