Lotus Traveler: Tuning English

Da wiki ebasso.net Wiki
Ir para: navegação, pesquisa

Lotus Traveler Thread Pools

Description

Lotus Notes Traveler has four thread pools that can be tuned:

  • Prime sync threads - Determine if changes to user mail files need to be synchronized to user devices.
  • Device sync threads - Synchronize data between Domino mail servers and user devices.
  • Worker threads - Used internally in the device synchronization process.
  • HTTP threads - Used by the HTTP task, one per HTTP connection.


GENERAL: Optimizing Name lookup

A good way to optimize Name lookup, is to configure the minimum number of characters in the search string required to perform a search.

NTS_LOOKUP_MIN_LENGTH=3


Traveler Availability Index

Availability Index (AI) used by Traveler servers in the pool to balance load.?

The AI range can be from 100 to -200
  • 100 to 0 Normal processing range (base AI 100)
  • 0 to -100 Traveler server is in constrained state (base AI 0)
    • Constrained state is when the server recognizes it is overloaded and starts send 503 server

busy responses in attempt to recover.

  • 00 to -200 Traveler server is unavailable (base AI -100)
    • Administrator has marked server as unavailable using tell command.

? The AI is computed by subtracting the larger of the memory usage and CPU usage from the Base AI. Native memory is the percentage of Native memory use vs physical memory Average CPU percentage over last 15 minutes.

System stat results

System stat results

Default Values

Variables at file Domino data directory\traveler\cfg\NTSConfig.xml.


Thread pool name

Default Value

TSS_PRIMESYNC_THREADS

10

TSS_SYNC_THREADS

50

WORKER_THREADS

100 (should always be double the TSS_SYNC_THREADS value)

HTTP Threads

  • 100 on Windows 32-bit machines
  • 400 on Windows 64-bit machines)


How to determine the current number of active devices

Checkout for Push.Devices.Total statistics. To find out if you need to change this parameter, run the following command

> tell traveler stat show

...
 Push.Devices.Total = 225
 ...


How to determine the current number of active HTTP threads

The current number of active HTTP threads allocated by the Domino web server at startup can be found in the Server document on the Internet Protocols -> HTTP tab in the Number active threads field.

Guidelines

The following are general guidelines for the relationships between the number of threads in each of the thread pools.

  • Prime sync threads should be much less than the number of device sync threads; normally the prime sync default of 10 threads should be adequate.
  • Worker threads should be double the number of device sync threads.
  • HTTP threads should be 1.2 times either the number of device sync threads or the number of ActiveSync devices, whichever is larger.

Statistics

Statistics

The first set of statistics is the device synchronization GetThreadDelayTime histogram, where you can see if there is a long delay before a device synchronization thread is allocated. This is an example of a device synchronization thread histogram:

  • ThreadPool.GetThreadDelayTime.DS.00-02
  • ThreadPool.GetThreadDelayTime.DS.02-05
  • ThreadPool.GetThreadDelayTime.DS.05-10
  • ThreadPool.GetThreadDelayTime.DS.10-Inf

Interpretation

  • ThreadPool.GetThreadDelayTime.DS.00-02 = 571071
  • ThreadPool.GetThreadDelayTime.DS.02-05 = 45
  • ThreadPool.GetThreadDelayTime.DS.05-10 = 29
  • ThreadPool.GetThreadDelayTime.DS.10-Inf = 17

In this example the first line indicates that the majority of the device synchronization starts in the 0 to 2 second range. This implies that the current device thread count is sufficient for the load.

Reference


Statistics

The second set of statistics is the number of device synchronizations that have finished with various return codes (DeviceSync.Count.xxx). The return codes are:

  • 200 - Successful
  • 408 - Device did not respond before the server terminated the session
  • 409 - Device started a new session which caused this session to be terminated
  • 500 - Unknown Error
  • 503 - Server Busy

Interpretation

  • DeviceSync.Count.200 = 43427
  • DeviceSync.Count.408 = 199
  • DeviceSync.Count.409 = 44
  • DeviceSync.Count.500 = 5
  • DeviceSync.Count.503 = 4

DeviceSync.Count.503 shows the number of synchronizations that were not allowed to start and were ultimately aborted because a device synchronization thread could not be allocated to run the synchronization. This value should be much smaller than the DeviceSync.Count.200 value, which shows the number of synchronizations that have finished with a return code of 200 (Successful). If the DeviceSync.Count.503 value is too high, you should increase the number of device synchronization threads, which should reduce the occurrence of synchronizations that result in a server busy response.

The following is an example of the DeviceSync.Count stats. In this case, the ratio of 503 results (4) to 200 results (43427) is very small, which indicates there is no need for additional device synchronization threads.

Reference


Cache User Info - HTTP Maximum Cached Users - 64 bits only

To optimize the response time keeps Domino design elements in memory.

Tune

To change the parameters of Web Cache, Open the server document and go

Internet Protocols → Domino Web Engine → Memory Caches.

and change the field value

Field

Description

Maximum cached users

Enter the number of users to cache. The default is 64. After a user successfully authenticates with a server, Domino stores in memory the user's name, password, and the list of groups to which the user belongs. Use this field to increase the number of users for whom Domino stores this information.


Statistics

  • Push.Devices.Total

Interpretation

  • Push.Devices.Total = 2455

Reference

Example

To find out if you need to change this parameter, run the following command

 > Show stat domino.cache.user*
Domino.Cache.User Cache.Count = 68
Domino.Cache.User Cache.DisplaceRate = 0
Domino.Cache.User Cache.HitRate = 73.258795221749
Domino.Cache.User Cache.MaxSize = 64


Observing these numbers we can deduce that the value of elements (Count) has reached the maximum (MaxSize), so elements in this cache are discarded ( DisplaceRate).

The optimal value for theHitrate is near 100%, andDisplaceRate is 0%. Increasing the value of the fieldMaximum cached designs improves response-time.


Deciding which server performs the lookup

By default, the lookup is performed against the user's mail server. This is done to reduce the configuration needed on the Traveler server and to provide more consistent results with a Notes client.

You can change this behavior to perform the lookup operation against the Traveler server. To do this, set the following parameter in the notes.ini file:

  • NTS_TRAVELER_AS_LOOKUP_SERVER=True

Ver também