Our News/Articles/Opinions/Technotes from the world of IBM Collaboration

Lotus Traveler – Latest Device Updates (Feb 2012)

It’s worth noting this article published by IBM to make sure you’re on an appropriate release of Traveler for your users’ devices.  (and also for information on upcoming Traveler device support!)

Steven Vaughan – Domino People Ltd

The BES for Domino Extended Threading Model

Did you know that there is a default number of threads (40) on a Domino BES server that allow for connections to remote target mail servers which are used to read/write to users mail files?

Performance problems can occur if you have multiple target mail servers with a good number of target users. The Domino BES server by default will automatically allocate threads by target mail server numbers. A simple explanation is as follows:

Server A – 5 Users – 5 Threads Allocated out of 40 available – Ratio 1 Thread per user

Server B – 10 Users – 10 Threads Allocated out of 40 available – Ratio 1 Thread per user

Server C – 15 Users – 15 Threads Allocated out of 40 available – Ratio 1 Thread per user

Server D – 50 Users – 15 Threads Allocated out of 40 available – Ratio 0.3 Threads per user

From the example above, clearly Server D with 50 users will have potential performance issues with the BES server unable to allocate enough threads to facilitate the high user number which means there may be delays in delivering messages to handheld devices. The basic way of putting this is each BES will assign a number threads to mail servers(assuming they’ve at least one BlackBerry user) within your environment.

Now here’s a simple explanation of the difference between the two threading models:

The default threading model assigns the default pool of 40 threads purely by the number of mail servers.

The extended threading model assigns threads dynamically by the number of users on each mail server. (You can also increase the number of total threads in your pool for your BES beyond the default 40).

I use the following rule of thumb for when the default threading model will work efficiently for you if you’ve a BES server connecting to:-

-Only one mail server/cluster with a user base 100 or less


– Multiple mail servers /clusters with a similar number of BlackBerry users per server/cluster. (again with a total user base no greater than 100).

So the the conclusion from that is that it won’t work efficiently for you if  you’ve:

-More than one mail server/clusters with  differences in BlackBerry users per server/cluster


-A mail server with greater than 100 users (ie beyond that point 40 threads become insufficient).

If you’ve more than 100 users and you want your BES to be efficient, you need to start thinking about increasing your thread pool. One caveat is to be very careful increasing the number of total threads in your thread pool.  Do so incrementally and make sure the resources on your BES can handle an increase.

What does this mean for my end user?

The threads are used for every interaction with BlackBerry handhelds.  So if a mail server is “under served” by threads it will mean delays the delivery of mail to handhelds.

You also need to take into account, that not all threads assigned to a mail server will be effectively usable for most of the users on that mail server. There are no real hard and fast numbers with this, a number of factors (how big your mailfiles are, whether they’re on a custom design, the network connectivity to your BES) contribute to how quickly a mailfile is “read” by BES.   But usually, from experience most mail servers will have at least a small number of mailfiles who hold onto BES threads for significantly longer than others, so effectively several threads are “locked” from the majority of your users.  If your mail server isn’t adequately served by the remaining threads this will lead to mail delivery delays to your other users.

RIM’s explanation of the threading models and extending threads is here.

Let us know if you’ve found this useful.  There is much more detail I can go into on how to know if your users are “locking” threads consistently, how to measure if changes you’ve made to the threading model are effective and  and how to handle completely hung threads better than the built in alerts RIM provide.

If you’d like us to look at your environment for you and recommend improvements that could be made to make your BES threading model more efficient then please do contact us

Cormac McCarthy – Domino People Ltd