How do you solve a problem like performance?

posted by Tomas Mcinernery on June 01, 2007

It’s been an interesting week, various meetings have started me thinking about things outside of the pure mobility realm so I thought I might put some thoughts down and see what comments we get. Performance is the name of the game today; how does an enterprise continue to ensure platforms perform to its best ability?

Why does my PC take so long to boot? is quite possibly the most difficult question IT support teams face. In my experience that’s because it’s just never a simple answer. The software platform contains so many various components a ‘slow’ platform (and lets not even get started on the actual definition) makes easy solutions tough to find. As an enterprise IT shop it’s a key responsibility to maintain platform performance even on older hardware.

The status quo here is once a system us up and running a user doesn’t think it needs to be maintained, and in the most part I would agree but like your car it needs a service to check the tyres, fix the air conditioning etc once a year at least. The issues around performance start to arise because IT shops design one build which is deployed and updated periodically. What happens to that build after it’s been in the environment after some time? Sometimes the answer to that question that’s been addressed. A strategy around performance needs to be in place to ensure those platforms can keep meeting users requirements.

Reporting and benchmarking is the first step here. Work to have a good understanding of your platform performance across the user base. Put some guiderails in place for typical performance that can be used to warn when performance slips. The benchmarking that creates the guiderails may not just be boot speed and fragmentation but include amount of calls logged, age of hardware and granular details like average processes running on the operating system.

User education can help here; making users aware applications run with file access overheads or are intensive on memory may help, teaching your employee to use the application correctly for long-term (i.e. how to maintain their applications, or perhaps only run it when they need it) will help. Interestingly I have seen some enterprises allow users to help with driver/patching performance themselves for example allowing them to update their own BIOS. I wouldn’t recommend it normally but it’s your call. Work to communicate simple messages around applications and maintenance activities, make sure it’s something, as always, the user can take with them and reference. Intel IT has developed tools where users can visit a web site and generate a quick report about what they could do to speed up performance – it’s proved very popular.

Keeping the system current is where some of the issues lie. It’s natural for a compound effect to hit the platforms after hundreds of various updates, users end up with different versions of the original build that may include applications of their own, specific business unit tools, new drivers from some pilot or other etc. All these additions to the platform need to go through one process of quality assurance which doesn’t just test that the applications technically work together but that they also function well under load. Security updates are unfortunately unavoidable. Keeping a system secure should be major goal of IT teams, however this can impact performance. Streamline the upgrades where possible and work to update builds to install these automatically – this will save you time when deploying new systems.

Patching applications and driver upgrades can cause another compound issue with performance. Automated tools either purchased as part of a management suite is great asset have; it can work with on board hardware tools (like VPRO) Even on older systems have OEM tools which can support upgrading drivers/application patches. The challenge then is that the hardware OEM software tends to be targeted specifically to their hardware or licensing costs can hit.

Quality of service (QoS) is an interesting question here; as more enterprises start to move towards applications that require higher priority on memory access and CPU utilisation overall platform performance becomes important. Tools are available (some native to the operating system) that can prioritise the application stack, this may be an option to ensure that your QoS sensitive applications get what they need, it may also help lower the priority of applications that don’t require as much as they think.

Does your IT shop have a focus on client performance? How to you maintain and educate employees on keeping their system fresh? Have you started looking at QoS on the platform?

Comments (7)

tagged: chimps, performance