Articles
Whats New?
Database
General Interest
Industry
Performance
Reporting
Security
Upgrades
Kick the Dog
Subscribe | Authors | FAQ
Activate Digital Subscription

Connected! Our Newsletter

Add to Google Reader or Homepage
Putting the P in EPM
Explore each layer of EPM for opportunities to improve performance

Posted on 1/28/2007 (originally published in print on 3/1/2004)

by Andrew Spidle

Performance management for any process can be a daunting task and Enterprise Performance Management (EPM) is no exception. A few years ago, when applications were applications and data warehouses were data warehouses, we did not have such a tough time deciding which rule of thumb to follow. Well, today my thumb hurts. It hurts figuratively for all the rules of thumb that I have broken and literally from pushing way too hard on that little area on either side of the bridge of my nose next to my eye. Frustration can only begin to describe the feeling. Unfortunately, I have not found the magic rulebook either. Some of us may be in good shape with quick running processes while others are wondering if their process will ever finish. Hopefully, I will be able to shed some light on the areas that have helped me in the past. A word of warning though: every environment is different and my luck may or may not be transferable to another environment.

One of the first things to do is to educate. Everyone needs to be educated. About what, you ask? EPM, I say! Is it an application or a data warehouse? EPM has components of both and acts like neither. Therein lies both the problem and the source of frustration. The care and feeding of your new or old EPM must be left to a group of people with expertise in many areas. Again, all these folks need to be on the same page.Your system administrator may see EPM as an application, while your DBA may see it as a data warehouse. Well, they are both right and both wrong. The simplest way I have found to get a handle on this education process is to form a performance tuning/monitoring team. With a performance team in place, accountability, responsibility and action can take place to manage EPM and to continue the education process outside of the EPM area.

Forming the team with the right members is important to the team’s success and credibility. Executive sponsorship is essential. This ensures EPM gets the proper level attention and funding for its ongoing operations. Representatives should be selected from all stakeholders in EPM. It is important to have system administrators, data warehouse DBAs, developers and even end users on your team. This team should be staffed with quality individuals who are able to direct and prioritize the performance management efforts. Prioritization is important since there are so many places in EPM to look for performance maximizing opportunities. This team cannot be afraid to get dirty and do some problem solving on the spot either. A balance of action and management is needed from this team. Despite the differing needs and philosophies of the various people on your team, one guiding principle can usually be agreed upon. EPM is there for the users to use. Without the user, it becomes very difficult to generate the ROI called for by the business case or justify the expenditure for ongoing maintenance at all.

Now that you have your team built, you are ready to jump in and start tuning. Where do we begin? Some may say to check the SQL first. Others may look toward the tool used. Actually, anyplace is a good place to start, I believe. This may be a question for your performance monitoring team, right? However, I will begin from the bottom and work my way up.

A solid foundation for any EPM instance begins with hardware. Hardware such as CPUs, tape drives, SAN disks, RAID arrays and network cards are all examples included in this layer. At a high level, we need to look for the warning signs of a bottleneck. This may be indicated by high CPU utilization, memory swap rates, I/O rates, page faults or whatever indicator is available. There are literally hundreds of statistics that can be used to flag a problem. Some problems may be hidden without an obvious indicator. The goal for hardware performance is to provide as many pathways for the data to flow through as possible.

To continue reading this article you must have a current VP1 Subscription.
Already a Subscriber?

Become a VP1 Subscriber

or

Activate your Digital Subscription

© Copyright 2007 VP1 - All other trademarks are the property of their respective owners.