测试：性能测试概念 1

 

1.1       Performance Description

Here we divide all features into two groups , Single instance feature and Multi-instance features.

 

1.1.1        Single instance feature

One step here means that when one instance of the feature runs, there are no other instances running at the same time.

 

Two numbers are used here to describe the performance

 

1 Response Time: The time that applications use to finish exact task with a certain load.

How-to: 

If the feature is a complex set (BP full publish), a little Program should be used here to ensure the record of the time past accurately. It will record the time before and after the feature runs.

If the feature is only to run a sp, the VSTS is a good choice; the duration column recorded by this tool is accurate enough.

           

 2 Resource Utilization: The resource utilization when tasks run.

How-to: Windows administrative tool Performance is used here to record utilization data. The counter should be used in different environments and different aims.

 

When the feature mainly run at SQL server: 

·         SQLServer:Buffer Manager

o        Buffer cache hit ratio

·         SQLServer:General Statistics Object

o        User Connections

·         SQLServer:Locks

o        Lock Requests/sec

o        Lock Timeouts/sec

o        Lock Waits/sec

o        Number of Deadlocks/sec

·         SQLServer:Memory Manager

o        Memory Grants Pending

·         SQLServer:SQL Statistics :

o        Batch Requests/sec

o        SQL Compilations/sec

o        SQL Re-Compilations/sec

·         Memory

o        Available Bytes

o        Page Reads/sec

o        Page Writes/sec

·         Processor

o        % Processor Time

When the feature mainly run in Windows:

Processor

·         Processor Utilization

ASP.NET Apps

·         .NET Request Queue Length

·         .NET Request Wait Time

·         .NET Requests / Sec

Memory

·         Available Mbytes

.NET CLR LocksAndThreads

·         Contention Rate/Sec

·         Total # of Contentions

·         # of current Logical Threads

·         #of current recognized Threads

.NET CLR Memory

·         Induced GC

·         # of Pinned Objects

·         Finalization Survivors

·         Gen 0 heap size

·         Gen 1 heap size

·         Gen 2 heap size

·         % Time in Garbage Collection

.NET CLR Exeptions

·          # of Exceps Thrown

·         # of Exceps Thrown/Sec

ASP.NET

·         Request Execution Time

·         Requests Queued

·         Worker Process Restarts

ASP.NET Applications

·        Cache API Entries

·        Cache API Hit Ratio

·        Cache API Turnover Rate

·        Output Cache Entries

·        Output Cache Hit Ratio

·        Output Cache Turnover Rate

·        Requests Executing

·         Requests Failed

Web Service

·         Current Anonymous Users

·         Connection Attempts/Sec

·          Current Connections

·         GET Requests/sec

·         POST Requests/sec

·         Bytes Sent/sec

·         Bytes Received/sec

Network Interface (specific)

·          Output Queue Length

·         Packets Outbound Errors

·         Packets Received Errors

·         Packets Received/Sec

·          Packets Sent/Sec

 

 

1.1.2        Multi-Instance feathers

 

1 Response Time: The time that application uses to finish exact task with a certain load.

            How-to:

All times of instances with the feature should be recorded and than the average time is what we need. The way is same as single instance features.

                       

 2 Resource Utilization: The resource utilization when tasks run.

            How-to: The same as single instance features.

 3 Throughout: The throughput that application uses to finish exact tasks within a unit time.

Before we give “How-to”, I want to explain that why we should get the throughput in addition to response time. In Multi-Instance feathers, many instances will run together, there are two types of time we could record, total time of all instances and duration time of each instance. The second one is considered in response time, it will show how fast the server could run the feature. Now here is a question, shall we calculate total time from response times? No, you can’t. The total time will lower than response times, the difference between them is more or less which lies on the affiliation between the instances of the features and the performance of the servers. The lower the total time is, the lower the affiliations between the instances are and the stronger the servers are.

How-to: The start time and end time should be recorded by “SQL Server Profiler” or Virtual Studio 2005.

Throughput = load number / Total time

 

 

1.2       Test Scenario

One Scenario = One Environment + One Testing Objects + Test Type

A table should be created to show Scenarios. Test results should fill in the table later and the test report should be linked in the table too.

  

1.3       Priority (0, 1, 2, 3… the lower, the more important)

All scenarios should have a priority to lighten the most important scenario.

If it’s updated version, the changed feature should have a high priority.