Profiling deliverables

Application Transaction Response-Time Baselining /Benchmarking

Whenever a project requires a detailed understanding of how an application is currently performing, or how it will perform once deployed or subject to new deployment conditions, it is essential that the application is first benchmarked or baselined. This stage provides the valuable reference against which all future tests and profiles can be  analysed and compared.

The Diagnosys Application Profiling service can help you achieve this baseline understanding in very short timescales, and without the need for complex, or expensive test-bench set-up's.  Using passive data capture techniques and specialist application analysis tools, we can deliver a baseline profile of almost any TCP-IP based application with the following typical detail and value:

Transaction Response-Time & Network Traffic Profile

click on image to enlarge

Client-Server Network Traffic Flow

click on image to enlarge

Client-Server Response-Time Analysis

click on image to enlarge

This information provides a base-level understanding of how the application, and its main user transactions behave and perform across the network. Further analysis reveals more information about how the application itself behaves, its impact upon the network, and where user response-time performance bottlenecks may be occurring.

Network Capacity Planning

One of the most important challenges that any enterprise business will face when deploying an application across a distributed network infrastructure, is knowing how much WAN network bandwidth is required to support business user response-time performance objectives. Traditionally, organisations without complex performance testing tools, have either estimated bandwidth requirements or taken a "suck it and see" approach, usually resulting in either inadequate capacity, or wasted over-capacity.

Diagnosys, through its Application Profiling service, can offer a simpler, more intelligent approach.

Once the application has been profiled, and a baseline model of the main transactions obtained, we can then perform a number of "What-If?" analyses on the data to provide predictive analysis of response-time performance under different LAN/WAN configurations. The wide-ranging performance characteristics of different applications, result in widely varying performance profiles under different LAN/WAN deployment configurations. The application profiling service provides a very effective way to quickly and accurately understand how the transaction response-time any n-tiered TCP-IP application will vary with differing values of network bandwidth, latency, or load.

The following screen-shot illustrates an example of a response-time prediction for a database query transaction running between a client and server.  The top (green) bar is the baseline response-time performance with all client-server tiers connected via a 100Mbps LAN, with minimal network latency and load.  The following graph bars reveal the impact on response-time performance once bandwidth is restricted (in this case at 64kbps, 128kbps, and 2Mbps).

click on image to enlarge

The following alternative predictive analysis, reveals how sensitive the application is to varying values of end-to-end network latency (50, 100, 200 ms), together with increasing network bandwidth (from 64kbps to 1.5Mbps).

click on image to enlarge

Both of these analyses reveal how sensitive the application is to network bandwidth and latency, and provides a very accurate guide as to what the optimum value of network bandwidth is required to hit a required response-time performance target.

In many cases, the predictions reveal that, in its current state, no matter how much bandwidth is applied, the application will never meet the required level of performance specified by the business. In this case, changes will need to be made within the application itself, so as to change its baseline performance profile, and its behaviour when deployed across a WAN. The information contained with the application profiling report will provide clues to the application developers, DBA's and/or technical architects as to what changes to the application code are required to meet the response-time performance objectives. Typical examples of such issues, include database queries that fetch data on a row-by-row basis across the WAN, instead of implementing more efficient  server-side stored-procedure queries.

Application Performance Troubleshooting

When users or customers start complaining about poor performance of business applications, you need to fix the problem fast! However, identifying precisely what the problem is, where in the system it is, how the problem manifests itself, when it occurs, and why ... Is no simple task. The most usual question asked is ... "Is it the application - or the network?".

The Diagnosys Application Profiling service can usually get to the root of the problem in much faster timescales, and require far fewer resources, than traditional performance troubleshooting techniques.

The first step would be to obtain a baseline profile of the application under study.

The next step would be to analyse the profiles of the transactions included in the profiling study. This stage reveals the true performance characteristics of the application working across the network, and immediately helps identify where in the application code, the network or end-to-end client-server system the performance bottleneck is occurring.

click on image to enlarge

In this example, the profile analysis reveals that the main problem resulting in poor user response-times is a delay within the application - specifically a delay in the processing of a SQL "SELECT" query statement. The analysis therefore rapidly reveals which part of the application code is to blame, and assists the in-house development/DBA staff decide on the best remedial action to improve application response-times.

Typically, attempting to troubleshoot application performance issues requires multiple people and resources from across the IT organisation to inspect each and every component of the application system under isolated conditions (e.g. application, network, servers, etc.). Usually, these separate troubleshooting activities do not reveal a holistic picture of the entire system, and therefore increases the complexity and time required to identify what and where the problem is.

The Diagnosys Application Profiling service offers an extremely efficient approach to isolating performance problems quickly, and delivering precise information to help your IT teams understand what remedial action is required.

Please call us now on +44 (0) 870 2405519

or, for more information, email us at info@mydiagnosys.com