Capacity Planning

[joe_91]

What does it take to be a capacity planner? I currently do sys admin stuff and do a lot of performance monitoring using various tools (measureware, perfview, glance, top, sar, vmstat,iostat,lsof,tusc,q4 etc). what are the additional things i should know about to become a capacity planner? I thought it was an extension to the sysadmin job. Is that correct?

Thanks

Joe

[Jeff_Traigle]

In general, I'd say it takes not much more than what you're doing. Capacity planning essentially means taking the performance metrics and monitoring trends over time. As time passes, how much space is being used by databases/applications, how much memory utilization increases, etc.? At the rates of change you discover, how long will it be before additional resources are needed? Are there enough resources available for some bozo to add another database or application to a server right now? :) Depending on the size and dynamics of the environment, capacity planning can be a full time job unto itself or it could be just a small piece of a sysadmin's overall duties.

[Ivan Ferreira]

Yes, its an extendion to sysadmin job, but there are a lot of things to consider when you do a capacity planning. Also, mathematics and statistics are very used, and there are techniques to get a very accurate capacity planning.

My best recommendation is to obtain the following book:

Performance by design: Computer Capacity Planning by Example

By Daniel A. Menasce, Virgilio A.F. Aleida, Lawrence w. Dowdy.

[Steven E. Protter]

Shalom Joe,

Capacity planning is a dark art. The dark art of predicting the future.

You take your current applications and check the requirements of the next versions if the information is available. You ask the managers to tell you how many bodies will be working of the systems in the years ahead.

You plug all of these things into your brain and hope you can find the servers to meet your needs.

Then someone doesn't tell you about some new project or application and it all goes out the window.

Its part of the core of being a sysadmin.

SEP

[Geoff Wild]

I'd say yes it is an extension of a sysadmin.

The other things you need to have, understand who your customer is and what they want.

From ITIL, Capacity Management is:

"To understand the future business requirements (the required Service Delivery), the organization's operation (the current Service Delivery), the IT infrastructure (the means of Service Delivery), and ensure that all current and future capacity and performance aspects of the business requirements are provided cost effectively."



Rgds...Geoff

[Bill Hassell · Accepted solution]

Capacity planning may turn out to be much more than monitoring current apps and buying more or bigger boxes to run them. Capacity (which is typically performance) many affect several areas beyond the box such as disks, backup hardware, networking, even power and battery backup. As mentioned, there is a lot of magic involved trying to figure out what will happen in 3, 6 or 12 months in the future. depending on your company, this planning may be triggered by an aquisition(s) of another company, company initiative to reduce costs through server consolidation, even new government regulations that double or triple the storage (and backup) requirements.

I mention backups a couple of times because it is all too common to add a few terabytes of storage without a clue as to how to backup the data. This is where the amount of data and the backup timeslot are no longer compatible. Swapping a smaller tape drive with a larger drive often fails because the new tape drive has significant perforamnce requirements to meet throughput specs. A slower machine (slow CPU and/or slow disks) may reduce the tape speed to less than 1/10 of the drive (due to stream dropouts).

Similarly, data transfers over the network may suddenly increase and cause significant delays. Or the company decides to move the servers to some remote location without characterizing the bandwidth needed for the systems. 100Mbit local network speeds will be incredibly expensive to implement when the systems are hundreds of miles away.

An IT initiative might be to convert all local storage (JBOD and small arrays) into a large storage farm. But fiber cards and switches are expensive and giant arrays with large caches are also expensive. So the choice might be to buy a network storage box without evaluating the big drop in data throughput due to slow LAN speeds (100Mbit is VERY slow for disks).

Just some of the areas that need examinationg and consideration.

[joe_91]

Thanks for some real great answers!!!!


Best Regards

Joe