13.1.2 Investigating the cause

This subsection explains how to check the performance by using statistics.

Organization of this subsection

(1) Check application program performance
(2) Checking the performance of user functions

(1) Check application program performance

The figure below shows how to use statistics to check application program performance. The detailed procedure is provided below.

Figure 13‒2: General procedure for checking application program performance

(a) Check the trend in average performance

Check the trend in average performance from the following statistics items:

Statistics type	Item	Column name
Statistics (`eads_stats.csv`)	Average request processing time	`RequestProcessingAverageTime`

Statistics type

Item

Column name

Statistics

(eads_stats.csv)

Average request processing time

RequestProcessingAverageTime

(b) Identify the cause of poor performance

Identify the possible causes of poor performance by checking the following statistics item:

Statistics type	Item	Column name
Statistics (`eads_stats.csv`)	Average internal processing time	`InternalProcessingAverageTime`

Statistics type

Item

Column name

Statistics

(eads_stats.csv)

Average internal processing time

InternalProcessingAverageTime

(c) Analyze the cause of poor performance

Analyze the possible causes of poor performance identified above.

■ If the EADS server is the cause

If the EADS server might be the cause, check the following:

Determine if accesses to the EADS server are concentrated by checking the following statistics item:

Statistics type	Item	Column name
Statistics (`eads_stats.csv`)	Number of requests accepted (`put`, `create`, `update`, `replace`, `get`, `remove`)	`RequestCount`

Statistics type

Item

Column name

Statistics

(eads_stats.csv)

Number of requests accepted (put, create, update, replace, get, remove)

RequestCount

Check the Java logs for any garbage collection.
Check the OS information (such as the CPU usage and memory usage)

■ If the network is the cause

If the network might be the cause, check the following:

Determine the amount of communication by checking the following statistics items:

Statistics type	Item	Column name
Statistics (`eads_stats.csv`)	Maximum size of receive data between EADS client and EADS servers	`CSReadMaxSize`
	Total size of receive data between EADS client and EADS servers	`CSReadTotalSize`
	Maximum size of send data between EADS client and EADS servers	`CSWriteMaxSize`
	Total size of send data between EADS client and EADS servers	`CSWriteTotalSize`
	Maximum size of receive data between EADS servers	`SSReadMaxSize`
	Total size of receive data between EADS servers	`SSReadTotalSize`
	Maximum size of send data between EADS servers	`SSWriteMaxSize`
	Total size of send data between EADS servers	`SSWriteTotalSize`

Check the network status.

To Page Top

(2) Checking the performance of user functions

The following figure shows the general procedure for checking the performance of user functions by using statistics. The detailed procedure is provided below.

Figure 13‒3: General procedure for checking the performance of user functions

(a) Check the trend in average performance

Check the trend in average performance from the following statistics item:

Statistics type	Item	Column name
User function statistics (`eads_function_stats.csv`)	Average user function execution time	`FunctionExecuteAverageTime`

Statistics type

Item

Column name

User function statistics

(eads_function_stats.csv)

Average user function execution time

FunctionExecuteAverageTime

(b) Identify the cause of poor performance

Identify the possible causes of poor performance by comparing the following statistics items:

Statistics type	Item	Column name
User function statistics (`eads_function_stats.csv`)	Average user function internal processing time	`FunctionInternalProcessingAverageTime`
Average user program processing time in the user function	`UserProgramProcessingAverageTime`

Statistics type

Item

Column name

User function statistics

(eads_function_stats.csv)

Average user function internal processing time

FunctionInternalProcessingAverageTime

Average user program processing time in the user function

UserProgramProcessingAverageTime

(c) Analyze the cause of poor performance

Analyze the possible causes of poor performance identified above.

■ If the EADS server is the cause

If the EADS server might be the cause, check the following:

Determine if accesses to the EADS server are concentrated by checking the following statistics item:

Statistics type	Item	Column name
User function statistics (`eads_function_stats.csv`)	Number of times the user function was executed	`FunctionExecuteCount`

Statistics type

Item

Column name

User function statistics

(eads_function_stats.csv)

Number of times the user function was executed

FunctionExecuteCount

Check the Java logs for any garbage collection.
Check the OS information (such as the CPU usage and memory usage)

■ If the user program in a user function is the cause

If the user program in a user function might be the cause, check each user function's statistics to determine which user function is the cause.

■ If the network is the cause

If the network might be the cause, check the following:

Determine the amount of communication by checking the following statistics items:

Statistics type	Item	Column name
Statistics (`eads_stats.csv`)	Maximum size of receive data between EADS client and EADS servers	`CSReadMaxSize`
	Total size of receive data between EADS client and EADS servers	`CSReadTotalSize`
	Maximum size of send data between EADS client and EADS servers	`CSWriteMaxSize`
	Total size of send data between EADS client and EADS servers	`CSWriteTotalSize`
	Maximum size of receive data between EADS servers	`SSReadMaxSize`
	Total size of receive data between EADS servers	`SSReadTotalSize`
	Maximum size of send data between EADS servers	`SSWriteMaxSize`
	Total size of send data between EADS servers	`SSWriteTotalSize`

Check the network status.

To Page Top