load testing report document - university of … · web viewwhen you insert new tables they may...

TTU010

Load Testing Report Document

Version 1.026/10/2015

TTU010Load Testing Report Document Version 1.0 26/10/2015

Contents1 Introduction...............................................................................4

2 Load Test Scenarios and Results:................................................52.1 Scenario 1: Course Timetables.....................................................................52.2 Scenario 2: Building Timetables...................................................................82.3 Scenario 3: Combined Courses Timetables................................................122.4 Scenario 4: Room Bookings........................................................................15

3 Conclusion...............................................................................20

of 21


Document managementRole Unit Name

Technical Architect Development Technology

Emilio Perez

Other contributors

Version controlDate Version Author Sections Amendments

26/10/2015 1.0 Emilio Perez All Initial draft.

Some keyboard shortcuts have been added to this document: Alt-P changes to paragraph style; Alt-N changes to normal style; Alt-R changes to preformatted style; Alt-C changes to code character format, use Ctrl-Spacebar to clear; Alt-1, Alt-2 and Alt-3 create headings; Alt-B creates a bulleted list and Alt-0 (zero) creates a numbered list. You can also create appendices using the Shift-Alt-1, Shift-Alt-2 and Shift-Alt-3 shortcuts.

When you insert new tables they may appear too big because of the formatting of the default paragraph style. To fix this issue, select all rows and change to normal style (Alt-N). You can change the heading style of a table by placing the cursor anywhere inside it, selecting the Design tab and clicking one of the first three table designs.

To update the table of contents and any cross-references, hit ctrl-A to select all then F9, choose ‘update entire table’ and click OK. You might also want to select ‘update fields before printing’ under File > Options > Display > Printing Options.

of 21


1 IntroductionThe TTU010 Load Testing was carried out on both the old and new Test environments in order to compare the performance of the upgraded new environment with the previous version environment.

As on previous years the Load Testing mainly measured the response times of the web page requests and the servers (Web server and Database server) load. There are 4 scenarios that cover the main user operations on the Web Timetables and the Web Room Booking services.Each Test Scenario was kept running for 30 minutes in order to have enough operations samples and monitor the server load for a prolonged period of time.

of 21


2 Load Test Scenarios and Results:

2.1 Scenario 1: Course Timetables

Old Test Environment:

Summary:

Response Times Distribution:

Response Times Over Time:

of 21


Transactions per second:

New Test Environment:

of 21


Summary:




of 21


Comparison and results:

As we can see on the graphs of both old and new environments, the new environment got slightly better response times in general. Note that the new environment got higher Maximum response times because the requests had not previously been cached as it is the case with the old environment where all the tests were run a few times before the recorded test on this document.At the expected load, the web server responded within reasonable good times and the CPU usage in both environments was below the expected threshold, although significantly better on the new environment:

CPU usage avg Web server / DB server (Old Env) CPU usage avg Web server / DB server (New Env)40% / 50% 25% / 30%

After a load pattern of 40 timetables per minute sustained for 30 minutes, both the Web and the Database servers did not show any sign of increase on the CPU and memory load. In fact, the memory usage remained constant during the tests, hence there was no need to include it on the document.The average response time for this scenario (Display Course Timetables) was below 0.50 seconds in both environments.

of 21


2.2 Scenario 2: Building Timetables


Summary:



of 21




Summary:


of 21




of 21



In this Test Scenario the new environment got better minimum response times again. However, the response times over time graph shows that the new environment had more instances where the response times were above 1 second.

As on the Test Scenario 1 the CPU usage in both environments was below the expected threshold and exactly the same on both environments in this particular scenario:


After a load pattern of 20 timetables per minute sustained for 30 minutes, both the Web and the Database servers did not show any sign of increase on the CPU and memory load. Memory usage was also constant throughout the test.

The transactions per second graph also shows a better throughput per second than the old environment as the new environment reached up to 5 transactions per second and the old environment only reached 3.

The average response time for this scenario (Display Building Timetables) was below 0.50 seconds in both environments.

of 21


2.3 Scenario 3: Combined Courses Timetables


Summary:




of 21



Summary:


Response Times over Time:

of 21




of 21


Test Scenario 3 showed better minimum response times. However, the response times over time graph shows that the new environment had more instances where the response times were above 1 second. As these differences in response times are so small it does not have an impact on the average response times.

In this Scenario the CPU usage in both environments was below the expected threshold and significantly lower on the new environment servers:


After a load pattern of 8 timetables per minute sustained for 30 minutes, both the Web and the Database servers did not show any sign of increase on the CPU and memory load. Memory usage was also constant throughout the test.

The average response time for this scenario (Display Combined Courses Timetables) was below 0.70 seconds in both environments.

of 21


2.4 Scenario 4: Room Bookings


Summary:



of 21


Transactions per Second:


Summary:


of 21



Transactions per Second:

of 21



The Test Scenario 4 showed better overall performance on both Response Times and CPU usage for the New environment in comparison with the Old environment.

In this Scenario the CPU usage in both environments was again below the expected threshold and significantly lower on the Web server of the new environment:


After a load pattern of 8 bookings per minute sustained for 30 minutes, both the Web and the Database servers did not show any sign of increase on the CPU and memory load. Memory usage was also constant throughout the test.

The average response time for this scenario (Web Room Bookings) was below 0.15 seconds in both environments, and the average Room booking total time was below 6 seconds assuming an automatic Room Booking (in this case, using a Jmeter script).

of 21


3 ConclusionThe overall response times for the four test scenarios were very similar in both environments. However, the new environment showed slightly better response times mainly on the “minimum response times” of the script requests. Also, the CPU usage of the Web server on the new environment was significantly better on Scenarios 3 and 4.

Although the Load Testing does not reflect all the variables that can affect the performance on the Live service, however it has shown a positive result for the new upgraded environment over the old one.

of 21

load testing report document - university of … · web viewwhen you insert new tables they may...

Documents