Overview
Performance-at-a-Glance is the technique we've developed to clearly illustrate the performance bottleneck in any end-to-end system.
The diagram above shows a typical example. Our customer processes credit card applications. The users were experiencing intermittent response times of 20+ seconds against an expectation of 10 seconds. The belief was that the cause was either the Managed Network or delay in responses from the External Credit Rating Agency.
To determine the cause of the problem, advance7 consultants attached passive trace capture tools to various points on the network. Working alongside the users they waited for a problem. As soon as the problem occurred, the consultant marked all traces using one of our tools called Tranmark.
advance7 then processed the captured trace data with the help of another of our tools called the Event Correlation Engine. The Correlation Engine matches trace, log and Tranmark events enabling us to map a request from end to end.
The output from the Correlation Engine enables us to produce the Performance-at-a-Glance diagram and identifies abnormal events such as lost packets.
We use this method to analyse in-house business systems and Internet facing websites. The application does not have to be bespoke. We have looked at the performance of many commercial products.
The analysis can be as granular as you want. If you need to see the delay across individual network components we can show you the response time contribution of switches, routers, firewalls, load balancers, etc.
|
Advantages
- Definitive
Alternative methods rely on capacity management data, the interpretation of which is often subjective. Because the Performance-at-a-Glance
shows precisely where the time was spent for individual transactions the findings are definitive.
- One shot
You may only get the problem once per hour, day or week. Because we take a snapshot at the precise time of the problem, we only need
to capture one example. From this we can tell you precisely what happened at the time of the problem.
- Drill-down
We gather full network trace data enabling us to drill-down into a problem. For example, if the majority of delays are in the database, we can
extract the full query. If it appears that the major delays are across the network, we can determine if packets are being dropped or if the
network is overloaded at the precise time of the problem.
- Optimisation
By studying the captured data, we can determine if the infrastructure is configured for optimal performance.
- Program logic
Further study of the data enables us to identify flaws in the program logic that can be fed back to developers.
- Live environment
The measurement tools we use are totally passive. This means that
we can measure a live environment without causing disruption,
so you don't have to recreate the problem in a test lab, which
is slow, expensive and rarely works.
- Perfect IT upgrades
How many times have you had to make an upgrade decision based on anecdote or biased advice from a supplier? Because we have the Correlation
Engine we can process very large amounts of diagnostic data. We can produce a Performance-at-a-Glance diagram spanning, say, an 8-hour period.
That gives you perfect information on which to base upgrade decisions.
|
