Network World
Friday, November 21, 2008
DNSstuff.com
Get information about your IP
IP Information
50+ On-demand DNS and network tools

App Performance View

Navigation

The Rows and Columns Performance Management Game Board

There are two sides to performance management: rows and columns. We ask you to build a table of rows and columns in our mind - like a grid. We ask that you visualize this table only because actually drawing such a table is very complex.

The columns are resources that must be managed.  It takes a lot of resources to deliver a screen of information to the user. These resources are typically managed by element managers and enterprise management systems. The enterprise often has a huge investment in these tools. Furthermore, the tools generally match the system architecture and organization chart. There are management systems for the network, servers, database, storage, etc.

This is referred to as silo management. There are good reasons to manage each silo independently - the technologies are very different. You wouldn't want the network expert trying to figure out why one server with 10 users seems to be much slower that the other server with 1000 users.

Smart enterprises know that they can't rely on each silo operating completely independently, so they implement "end-to-end" silo management. These systems provide an integrated view of all the silos on a single screen. The need for dashboards and simplification arises from the fact that an enterprise may easily have 6 to 12 or dozens of silos in operation, and each silo may encompass tens or hundreds of devices.

The rows are a completely different view of the system. This is the view as seen by the actual user's traffic. A user session is made up of a flow of packets from the client machine through some of the elements of each silo and back again. This is often referred to as the horizontal cut or flow management view of the particular elements actually serving the user at a given time.

A single user task (traffic from when the user hits "Enter" to when the screen updates) takes up many rows. Each row is a series of packets or protocol events that bounce between a client and a server to complete the task.  We refer to these individual round-trips as turns.  In a business-to-business application built on a modern web-based architecture each user task typically requires 20 to 100 turns or rows in the table.  That is 20 to 100 discrete measurements of events and performance at the transport layer.

The complete table of rows and columns provides the data necessary for comprehensive performance management (PM) but can easily have hundreds of columns and thousands of rows for just one hour of operation.  The table in your mind may now have even more entries.  The number of cells in this table is in the millions!  Now you know why you may get headaches from time to time.

The table of rows and columns can only be managed with automated software.  Software that provides a good view of each column and then integrates the columns into an end-to-end view is providing Infrastructure Performance Management (IPM).  Conversely software that can identify application flows and report appropriate performance characteristics that show the user's experience are providing Application Performance Management (APM). 

You need a balance of both views.  We recommend that you look at your management software capabilities and assess if you have an equal dose of both views.

One way to look it

Useful answer?
0

Yes, this is a common way to look performance. The nice is that it is easy to run all kinds of statistical checks and (visual) results when data is presented this way but, if the causes and results are not well understood, can also be very misleading. That's why it is often used by marketing, right numbers but wrong (misleading) conclusions!
The problem is that many of the measurement results are not independent - enhance one and something else or maybe the whole gets worse. This happens because performance management is seen too often equal to capacity management, which it isn't. Performance management is the tactical part of capacity management which is strategic. Only taking a larger (longer) view of performance can find a balanced view of performance problems.
Maybe that would be next good article?
Just reading the numbers doesn't help much when next holiday peek, end of quarter reports, new product or a moderately changed application, new storage array or even a new vendor software version kick in and by then it is too late to start analyzing why? That might work for social and gaming systems and networks but can be very bad for business. There is no magic in tools, expensive or not, the user must understand what they really want.

Comment viewing options

Select your preferred way to display the comments and click "Save settings" to activate your changes.

Post new comment

The content of this field is kept private and will not be shown publicly.
  • Allowed HTML tags: <a> <em> <strong> <i> <b> <cite> <code> <ul> <ol> <li> <dl> <dt> <dd> <blockquote> <br /> <br> <p>
  • Lines and paragraphs break automatically.
  • You can use BBCode tags in the text.
  • Web page addresses and e-mail addresses turn into links automatically.

More information about formatting options

CAPTCHA
This question is for testing whether you are a human visitor and to prevent automated spam submissions.

About Peter Sevcik and Rebecca Wetzel

NetForecast is an internationally recognized engineering consulting company that benchmarks, analyzes, and improves the performance of networked data, voice, and video applications.

RSS feed XML feed

Sevcik and Wetzel's archive.

The opinions expressed in this Weblog are those of the writer and may not represent the opinions of Network World.

Advertisement: