BizTalk 2006 / R2 Publishing Throttling State #2 (imbalanced message publishing rate, input rate exceeds output rate)

Our next destination in BizTalk Throttling exploration drive is inbound Throttling states, publishing throttling state #2 Throttling due to imbalanced message publishing rate (input rate exceeds output rate).

Throttling state # 2, due to imbalanced message publishing rate, input rate exceeds output rate:

This is the most simple scenario to understand, here imbalanced message publishing rate means the ratio of the incoming messages by outgoing messages or vice versa.

The following is the condition of the Throttling state:
- (Message publishing incoming rate / Message publishing outgoing rate) * Rate overdrive factor in %

This condition belong to rate base throttling, For inbound (published) messages, BizTalk Server throttles publishing of messages if The Message publishing incoming rate for the host instance exceeds the Message publishing outgoing rate * the specified Rate overdrive factor (percent) value. The Rate overdrive factor (percent) parameter is configurable on the Message Publishing Throttling Settings dialog box. Rate based throttling for inbound messages is accomplished primarily by inducing a delay before publishing the batch of messages into the MessageBox database. No other action is taken to accomplish rate based throttling for inbound messages.

In following general scenario, I found the the throttling state can be possible in environment.

- very high demand to process messages and less available resources to process those messages.
- high use of co-related messages and self-correlated ports, and multiple subscribers for single received message.
- Slow outbound adapters.
- imbalance between host configured for inbound adapter and host configuration for XLANG / host configuration for outbound adapter.
- high processing complexity.



Understanding about Publishing throttling threshold parameter,

Minimum number of samples: Minimum number of messages BizTalk Server will sample for the Sampling window duration before considering rate-based throttling. If the actual number of samples in a sampling window fall below this value then the samples are discarded and throttling is not applied. This value should be consistent with a rate at which messages can be published under a medium load. For example, if your system is expected to handle 1,000 documents per second under a medium load, then this parameter should be set to 1,000 * Sample window duration in seconds (or more precisely, 1 * Sample window duration (milliseconds)). If the value is set too low, then the system may experience a throttling condition under low load. If the value is set too high, then there may not be enough samples for this technique to be effective.
Enter a value of zero to disable rate based inbound throttling.
The default value is 100.

Sampling window duration (milliseconds): The time-window measured in milliseconds, which is used to calculate the publishing rate based on the samples collected. The duration should be increased if the latency required for publishing a single message is high.
Enter a value of zero to disable rate based inbound throttling.
The default value is 15,000.

Rate overdrive factor (percent): This controls how much higher you allow the request rate to be than the completion rate before a throttling condition occurs. For example, if messages are being published at a rate of 200 per second and this parameter is set to 125, then the system will allow the publication of up to 250 messages per second (125% * 200 = 250) before applying throttling. Specifying too small a value for this parameter will cause the system to throttle more aggressively and could lead to over-throttling. Specifying too large a value for this parameter will cause under throttling and prevent the throttling mechanism from recognizing a legitimate throttling condition.
The default value is 125.

Maximum throttling delay (milliseconds): This is the maximum delay BizTalk Server will impose on a message instance due to throttling. The actual delay depends on the severity of the throttling condition.
Enter a value of zero to disable inbound throttling.
The default value is 300,000.

Triggering Mechanism of State #2 publishing throttling: In the sampling window duration, if the incoming/outgoing ratio excceds the configured parameter "Rate overdrive factor" value then BizTalk will be in throttling state.

Throttling Action: Block the publishing thread for a dynamically computed time period until the Message Publishing Incoming Rate is at par with the Message Publishing Outgoing Rate * the specified Rate overdrive factor (percent) value. By blocking the publishing thread, BizTalk can reduce the queued up messages, I have noticed many time the queued up inbound messages has very bad impact on the memory as well as the processing on the BizTalk artifacts. And the impact of the queued up messages is showing exponential memory & utilization which can also impact indirectly overall to every processing in BizTalk. So by blocking publishing thread, BizTalk is reducing incoming message flow and keeping flow healthy.

Sample collection of BizTalk Perfmon data and Graph: You can view in below graph that when the publishing throttling rate condition is setisfying the system in throttling mode #2, also you can analyze as per the sample window state through it.



If you would like to have the perfmon logs data, biztalk throttling configuration snap and graph in compress file for studying or analyzing it further for study which I have simulate for study, please write me to nilayparikh@gmail.com

Hope this article helped you to understand state#2 of throttling for you.

Related articles posted by me on the same blog:
BizTalk 2006 R2 - Throttling - Perfmon Parameters - My Experience (Tuesday, January 15, 2008)
BizTalk Server 2006 / R2 Throttling Mechanisms (Friday, January 4, 2008)

MSDN Resource Link:
http://msdn2.microsoft.com/en-us/library/aa559591.aspx
http://msdn2.microsoft.com/en-us/library/aa559893.aspx
http://msdn2.microsoft.com/en-us/library/aa559628.aspx
http://msdn2.microsoft.com/en-us/library/aa578302.aspx
http://msdn2.microsoft.com/en-us/library/aa547859.aspx

Thanks for BEAR me during the article. Your valuable feedback and suggestion are welcome to nilayparikh@gmail.com

BizTalk 2006 R2 - Throttling - Perfmon Parameters - My Experience

BREAKING NEWS:These days.. Dr. Time Pass is spending most of time to understand and implement best throttling scenarios :-).

Let me come back to official language. Few day back, I saw lots of throttling in our BizTalk environment. My enthusiasm drag me to the further exploration in throttling mechanism. Let me grab opportunity to document some cooked things.

Since last weeks, I saw my BizTalk environment is facing suddenly unusual throttling behaviors so I tried to find the root cause behind the throttling state in my BizTalk environment.

The environment is facing lots of message delivery throttling during pick processing hours, and due to the delivery throttling state, and delays our whole environment get stunned. After this incidence, I start taking this functionality very seriously and now digging lots of stuff related to it, here in this article I am going to share my experience as well as the understanding I got after going through many books, articles and MS-KBs.

As we all BizTalk professional aware about the inbound message processing and the internal BizTalk instruments takes care about the inbound message management. In short if I like to explain the word "inbound message" then I can stat "messages towards messagebox" as inbound messages, in BizTalk messagebox is the heart of any environment or implementation. We were facing problem due to unsymmetrical messages' incoming publishing rate and messages' outgoing publishing rate, I noticed in our environment incoming message rate is very high in compare to outgoing message (here outgoing means subscription to messagebox, we are talking about deliveries), in resultant the number was increasing for the messages in memory. BizTalk recognized and our environment moves towards throttling state but still throttling was also not able to make the system stable and put it again towards stable node. Due to the very heavy message rush and less outgoing rate very high number of the messages were waiting in queue for processing. We saw almost 300-700 seconds as delay time during the processing. And finally the nightmare came to true :-0, and counter called "Days without Servility 1" need to reset. The thing I would like to convey is the requirement of the optimized configuration parameter for the throttling while setting up any host. I will stat 'must' recommend for setting-up a set of well configured and planned host for the pick time processing for the critical interfaces. I don't want to drive my article into implementation designing, but I would like to come back to the throttling scenarios.



In this article I would like to focus around perfmon counters, few important ratios, throttling states.

Let's dig first to some state counters, state counters shows the state of the current situation in your environment under monitor. BizTalk 2006/R2 comes with very useful perfmon state counters.

some high (*) counters like High database session, High database size, High in-process message count, High message delivery rate, High process memory, High system memory, High thread count, these all counters represents only a state by showing value like 0 or 1, where 1 the respective focused area cross the configured or warning level. These shows very useful information which should be monitor in your Biztalk environment.

In-process message count, Active instance count are two message queue counters which show the Number of in-memory messages delivered to the XLANG engine or the outbound messaging engine that are not yet processed and instances are in active state in memory referring to the EPM or XLANG engine, respectively.

Specialized message delivery counters like Message delivery delay (ms), Message delivery incoming rate, Message delivery outgoing rate, Message delivery throttling state, Message delivery throttling state duration, Message delivery throttling user override. These all perfmon counters should be monitored regularly during the pick processing to understand the behavior of your environment and the coming out patterns to understand and prevent the failures.

Message delivery delay (ms): The current delay in ms imposed on each message publishing batch (applicable if the message publishing is being throttled and if the batch is not exempted from throttling).

Message publishing incoming rate: Number of messages per second that are being sent to the database for publishing in the given sample interval.

Message publishing outgoing rate: Number of messages per second that are actually published in the database in the given sample interval.

Message publishing throttling state:

• 0: Not throttling
  
• 2: Throttling due to imbalanced message publishing rate (input rate exceeds output rate)
  
• 4: Throttling due to process memory pressure
  
• 5: Throttling due to system memory pressure
  
• 6: Throttling due to database growth
  
• 8: Throttling due to high session count
  
• 9: Throttling due to high thread count
  
• 11: Throttling due to user override on publishing

Message publishing throttling state duration: Seconds since the system entered this state. If the host is throttling, how long it has been throttling; if it is not throttling, how long since throttling was applied.

I would like to place basic example of some perfmon data and counters in graph representation.



I will try to explore on Message publishing throttling states on my coming articles.

Thanks for BEAR me during the article. Your valuable feedback and suggestion are welcome to nilayparikh@gmail.com

Related articles posted by on the same blogs.
1. BizTalk Server 2006 / R2 Throttling Mechanisms

BizTalk Server 2006 / R2 Throttling Mechanisms

BizTalk 2006 / R2 comes with good relief in terms of performance and throttling mechanism, if you work close to administrating and performance tuning job for BizTalk environment or any messaging system then you can visualize the complexity of pick processing time scenarios in any messaging system. Here with the article I am focusing to the particular BizTalk throttling scenarios and their solution. I have derived below contact by referring few books as well as my personal experience while working on performance criteria for BizTalk environment.

The below cases might not cover all the possibilities throttling scenarios of live BizTalk environment, but I have tried to collect as much from my memory.

Case: High message ratio for delivery rate / completion rate

The higher side of the ratio value indicates the service is unable to handle the rate of inbound/incoming messages to the environment. As solution we can throttle message delivery and make available as much resource as possible to BizTalk Server artifacts like XLANG and outbound transports. As a result of improved message completion rate, message latency rate and IO operation rate for message box will be reduced exponentially, also we can get more benefit by cleaning up the queue at out-bound end point.

Ideally environment should maintain ratio value around 1.05 – 1.07.

We can get monitor message delivery rate and message completion rate through various instruments like WMI / OM object / Perfmon counters / etc.

Case: High publishing ratio for request rate / completion rate

High ratio is indicates of the message box being unable to cope up with the load. To improve such scenario in our environment probably we can block publishing threads to slow down the rate and also we can indicate the service class to slow down the message-publishing rate. In the case there is no such solution available apart of scaling out the SQL Server environment where message box has been deployed, if the environment is facing such problems concurrently then it will be appropriate decision to scale-out the SQL Server environment.

I don’t have accurate observation for the value of the ratio, but I am sure that every organization must need 1 to called reliable messaging services :-).

To get the ratio value, we need to monitor entry and exit of commit batch call.

Case: Process memory exceeds a threshold

If process memory exceeds a threshold then probably your environment and batch process need steep memory, if you are getting such alerts or messages regularly then you should think about the steep memory for the environment. In this scenario, it has affect components like XLANG and all kind of transportation (adapters/pipelines). It also indicates services to dehydrate and shrink catch.

By monitoring private bytes of the process we can get clear idea about the memory threshold for the individual processes. In BizTalk hosts can be treat as processes.

Case: System memory exceeds a threshold

Same possibilities we can assume as we discussed for case “Process memory exceeds a threshold” above.

Here you can monitor the physical memory and trend through perfmon.

Case: Database session being used by the process exceed a threshold count

You can throttle publishing inside the BizTalk environment. To improve such condition, we need to tune the SQL Server Database properly and need to scale up/out as per our threshold requirement. In this scenario XLANG and all inbound transports will get affected. There is not such quick resolution available for the scenario but we can reduce/block the external/idle sessions available on SQL Server. If BAM Notification / BAS are configured to your environment and their database are also running on same server then you can disable / suspend those less priorities services to resume the message processing to normal.

Such scenario can affect XLANG objects as well as all type of inbound transports.

You can monitor session on message box by Perfmon counter / SQL Server Counters / Management Studio / etc. For cluster environment you can monitor session per message box.

Case: Process thread count exceeds a particular threshold

Throttle publishing, delivery and it indicates to reduce the thread pool size.

Such scenario affect XLANG objects as well as all kind of transports (adapters / pipeline). There is no direct solution for the scenario, long term solution could be scale-out the server box and suggestion will be to migrate 64x or itanium (or equivalent) processors server boxes. To handle some pick processing time during the day, we can provide more thread and priority to the host processes handling the threshold.

You can monitor count like threads per CPU to undertake such scenarios / cases.


BizTalk Server 2006 / R2 is handling such throttling automatically but we can get much improved performance by giving some favor to the mechanism. To perform auto throttling BizTalk Server uses the configuration parameters, I will try to come up with detail review of each parameter in next coming article.

Thanks for BEAR me during the article. Your valuable feedback and suggestion are welcome to nilayparikh@gmail.com.

Understand BizTalk – Performance Counters – Bottleneck Analysis (Counter: Low % Disk Idle on SQL Server High Avg. Disk Queue Length on SQL Server)

Counters: Low % Disk Idle on SQL Server High Avg. Disk Queue Length on SQL Server

Area of Analysis:
- Check whether the tracking database and Message box are on the same disks.
- Check whether the data and log files are on the same disks.
- Check the log sizes.

Possible Solutions:
- Use a SAN.
- Ensure that the tracking and Message-box databases are on different servers.
- If they are on the same server, ensure that they are on different disks.
- Ensure that the data and log files are not sharing the same disks.
Make sure the BizTalk agents are enabled on the server. The agents copy tracking data from the Message box to the tracking database. They also back up the databases and clean up the logs.

Understand BizTalk – Performance Counters – Bottleneck Analysis (Counter: Low % CPU Idle on SQL Server)

Counter: Low % CPU Idle on SQL Server

Area of Analysis:
- Check if DBA changed any of the default database settings set by the BizTalk installation.
- Max Degree of Parallelism are set to off Auto-Update Statistics and Max Degree of Parallelism are set to off and 1 respectively on purpose.

Possible Solutions:
- Minimize the number persistence points in your orchestrations.
- Use static methods instead of wrapping nonserializable components in atomic scopes.
- Avoid using Parallel shapes, except when needed.
- In a multi-Message box scenario, ensure that you have at least three Message boxes. The master Message box is doing all the routing to secondary Message boxes, which is CPU intensive. Whenever you are using multiple Message boxes, Microsoft’s distributed Transaction Coordinator (DTC) is involved; therefore you need to jump from one Message box to three instead of only two to offset that overhead.

Understand BizTalk – Performance Counters – Bottleneck Analysis (Counter: - Low % CPU Idle on BizTalk Server)

Counter: Low % CPU Idle (On BizTalk Server)

Area of Analysis:
- Too many hosts running on the server.
- Improper use of custom pipelines.
- Custom components requiring optimization.

Possible Solutions:
- Isolate receive, process, and send functionality into different hosts and run host instances on different servers within the BizTalk Server Group.
- Move message transformation out of orchestrations to your ports to avoid the creation of new copies.
- Move message filters to your ports and receive locations.
- Optimize your schema. Large schema reduce performance.
- Use distinguished fields in orchestrations rather than properties or XPath.
- Use pass-through pipelines whenever possible.

I will surely come back again on more performance & bottleneck analysis.