Thread count tuning
Hi!
We are running a weblogic 5.1 server for jsp/servlet/ejb behind an apache web
server, via the wls/apache bridge. As the number of users increased we expereienced
slow response times. We checked thread dumps from peak hours and saw that all
servlet threads were busy writing to network sockets. Still, if we check the internet
line usage, we are not using up all our bandwith.
We've managed to solve this by adding more threads to our wls server (36 servlet
threads, 46 totally). This has however slowed down our database somewhat since
we have to increase the connection pool size with the number of server threads.
So my question would be: is there some other, more elegant, way to solve this?
Having 46 execute threads seems pretty much (15 is recommended from BEA). Ideally
I would like the wls threads to write their results as quickly as they can to
the apache server, disconnect from the socket and leave actual work of writing
all the data back to the end user's browser to apache. My network knowledge is
pretty limited so I'm not sure if this is possible.
/Mattias
Why do you need to have the maximum thread count equal to the maximum number of simulatneous users? You'll never end up with a system that has 60,000 threads.
If you are doing traditional-style HTTP, then typically you can support many thousands of users with just a few threads. For example, our SPECjAppServ er submissions use about 40 threads to handle 4000 clients. It all depends on how many requests each client actually makes per second, and how long each request takes. If the clients are idle for long periods of time, you could handle 10s of thousands with just a few threads.
The keep-alive tuning is also critical in that scenario, as the server will eventually start disconnecting clients, who will re-create the TCP connections. That works seemlessly from a functional point of view, but isn't necessarily the optimal thing for performance.
On the other hand, if you're trying to use a Web 2.0-style HTTP request where you simulate server pushes by having the servlet request block, then yes, in 8.1 you will need one thread per client, but you'll be bound by OS and other system constraints on the number of threads the process can have. You'd be much better off in that case using the Asynchronous Request processing features being added to AS 9.1 (see Sun's project glassfish for details).
For information on configuring the thread count for 8.1, see:
http://docs.sun.com/source/819-0215/httpservice.html
In particular -- the section on Configuring HTTP Service Request Processing Threads.
Similar Messages
-
How do you modify the default Execute thread count in Weblogic Server 9.2?
How do you modify the default Execute thread count in Weblogic Server 9.2?
How can you tune the starting number of weblogic.ExecuteThread on server startup and/or set minimum number?
Is there an option from the console?
Please let me know.
ThanksSelf tuning will automatically manage the threads but however you can still control the min and max by adding the min and max values for each instance either directly adding in config.xml or through JVM settings
1) Modifying the config.xml
Just add the following line(s) to each server definition :
<server>
<name>AdminServer</name>
<self-tuning-thread-pool-size-min>100</self-tuning-thread-pool-size-min>
<self-tuning-thread-pool-size-max>200</self-tuning-thread-pool-size-max>
</server>
2) Adding some JVM parameters
It's safer the following way :
add the following option in your command line : -Dweblogic.threadpool.MinPoolSize=100
Regards
RR -
Changing thread count in WebLogic 10.3
Hi
We have migrated from WebLogic 8.1 to 10.3.
how do i change the thread count in weblogic 10.3? i saw options in weblogic 8.1 and 9.2 but not in 10.3.
Can anybody throw some light on this.From WL9.0 onwards, the Weblogic thread management mechanism is changed.Now no need to set the thread count. WLS itself is intelligent enough to do the self tuning of the thread pool.
To prioritize and schedule your application requests,you can create work managers and can associate request classes/constraints.You can refer the below article to understand the workload management using work managers.
http://www.oracle.com/technology/pub/articles/dev2arch/2006/01/workload-management.html
Inigo Prince -
Thread Count Queue Length in Negative and Too many standby Thread
We are using Weblogic server 9.2.2 with 1 admin server and 4 managed server . Currently in one of the servers I could observe that there are around 77 standby threads.
Home > Summary of Servers > server1 > Monitoring > Threads > Self-Tuning Thread Pool
I could see that the "queue length" is negative (-138) and self tuning standby thread count is 77. Large number of threads STANDBY thread persists during the busy time of the business hours where as other servers are fully utilized.
Is it normal to have negative queue length and so many STANDBY threads? As for JMS queue negative oracle had already acknowledged that it is a bug. Thanks.
Edited by: 855849 on May 1, 2011 7:19 AM
Edited by: SilverHawk on May 12, 2011 8:12 AMYesterday an Oracle Consultant acknowlegded that it is a bug. There was a patch issued for Negative count in the JMS queue count and now this. Thanks for the reply by the way.
-
Consumers count = thread count ?
Hello,
I've implemented on a wl cluster (10 6.1sp4 instances, patched with CR091195)
the approximated distributed destination pattern.
In my case the destination is a queue.
Every wl instance has thread count = 100.
I have 1 mdb, non transactional, and I haven't configured the max-beans-in-free-pool.
It happens, when I deploy the MDB, targetting it to the cluster, that I can see
on every wl instance (I have a monitoring program):
JMSRuntime.connectionsCurrentCount = 1
JMSConnectionRuntime.sessionsCurrentCount = 100 (I have only a "connection4116")
JMSDestinationRuntime.consumersCurrentCount = 100 (I have only 1 queue)
Why do I have all of these active consumers/sessions ?
(and I haven't yet run my jms client)
Thanks in advance
Best Regards
Mark
Hi Mark,
Before 6.1SP5, each MDB pool initializes
Math.min(default-execute-thread-pool-size,
max-beans-in-free-pool)
consumers and sessions.
For 6.1SP5 and later, each MDB pool initializes
Math.min(default-execute-thread-pool-size/2 + 1,
max-beans-in-free-pool)
consumers and sessions.
WL 7.0SP? (SP3 maybe?), and 8.1
allow MDBs to use the "dispatch-policy" descriptor field
and so assign themselves to a different
thread-pool. In which case initialized max concurrent
instances becomes:
Math.min(custom-thread-pool-size,
max-beans-in-free-pool)
I think that the max-beans-in-free-pool default is
some very high number. (At one point it was 1500).
The consumer instances will not necessarily all use
a thread at the same time, as they do not use
up a thread when they are not operating on a
message. In essence, thread usage depends on how
slow the receivers are, how fast the producers are, and
what other resources are also using
the thread pool.
I recommend tuning max-beans-in-free-pool so that total
instance count on a server is somewhat less than
the the thread count. This helps prevent dead-locks
where all threads are running app code, and none
are left to do other work.
In case you are unaware, I also recommend
using less than 100 threads unless the the
app typically blocks a thread for long periods of time.
The reason is that lots of threads may negatively
impact performance, the operating system ends up
burning CPU switching itself between threads
rather than doing actual work.
Each MDB deployment uses a single connection.
Hope this helps,
Tom, BEA
P.S. The above information will be in the next
iteration of the JMS Performance Guide white-paper,
which I've currently started working on again.
"Mark Cordobal" <[email protected]> wrote:
>
>Hello,
>
>I've implemented on a wl cluster (10 6.1sp4 instances, patched with CR091195)
>the approximated distributed destination pattern.
>In my case the destination is a queue.
>
>Every wl instance has thread count = 100.
>
>I have 1 mdb, non transactional, and I haven't configured the max-beans-in-free-pool.
>
>It happens, when I deploy the MDB, targetting it to the cluster, that
>I can see
>on every wl instance (I have a monitoring program):
>
>JMSRuntime.connectionsCurrentCount = 1
>JMSConnectionRuntime.sessionsCurrentCount = 100 (I have only a "connection4116")
>JMSDestinationRuntime.consumersCurrentCount = 100 (I have only 1 queue)
>
>Why do I have all of these active consumers/sessions ?
>(and I haven't yet run my jms client)
>
>
>Thanks in advance
>
>Best Regards
>
>
>Mark
-
High thread count on store.exe
I understand that the mdb store process utilizes as much memory as is available. My question is regarding the thread count on the process. One of the three mailbox servers tends to have a higher thread count than the other two. It
usually ranges from 150 to 200 (while the other two are usually 130 to 150) but today it is a little over 300. The rpc thread count on all the mailbox servers stays pretty low (60 or under) and the rpc thread count on the CAS servers usually
hang out around 55-65, so that's all good. All three mailbox servers also use VERY close to the same amount of memory all the time, so that doesn't seem to be making a difference. I am just trying to determine why the thread count is noticably
higher. This server and one of other mailbox servers replicate the public folder database, and I can understand why that would account for some additional activity, but you would think the thread count would be higher on the other server with the public
folder database as well, but it stays pretty low. Hardware-wise these servers are all identical. So is there something I can monitor in perf mon for instance? I don't want to use experfwiz because that monitors everything under the sun and
I need it a bit more centralized. I have googled all over the place and the only thing I can find is that with newer versions of exchange this thread count ranges from 200-300 on heavy usage, and exch07 can handle up to 500 threads, so I can't even find
which specific things to monitor that directly correlate with this issue. So while I understand that it's not really too high, I would like to find out why it deviates from the norm if the other two servers are equal for all intents and purposes (same
hardware, balanced DAG. etc). I would like to mitigate any potential issues. Thank you.Hi pchw,
Thank you for your question.
I am sorry that we could not find the specific things to monitor this issuer directly, because there are many reason which caused it. For example: virus, big size attachment in queue, hardware performance and so on. We could refer to the following steps
to troubleshoot:
Restart the service of “Microsoft Exchange transport”;
Create a new database and move mailbox to new mailbox;
We could do some disk health check if the load test is normal;
Collect related application log and system log to
[email protected] for our troubleshooting.
We could run SDP on Exchange server and send it to
[email protected] for our troubleshooting.
http://blogs.msdn.com/b/askie/archive/2012/05/22/introduction-to-support-diagnostics-platform-sdp.aspx
We could use tools and logs to analysis if there are anything wrong or abnormal on Exchange server 2013. our bottle neck is disk IO latency, store.exe do require some CPU usage, but if our disk IO speed can keep in a normal range, the
CPU usage will be smaller. Exchange is highly rely on disk IO performance because it need lots of IO operation regarding to transaction log and database file. If Disk IO has a high latency, I will definitely impact store.exe, if it`s getting seriously sometimes
it may cause store.exe no response.
We could refer to the following link:
https://technet.microsoft.com/en-us/library/ee832791(v=exchg.141).aspx
If there are any questions regarding this issue, please be free to let me know.
Best Regard,
Jim -
Execute queue thread count set back?
I have a script which automatically sets up execute queues (and their thread counts) for either an admin server or two managed servers in a cluster (depending on whether it's a standalone deployment). Upon building a domain, the script connects to a temporary server and sets everything up, at this point it appears all the queues and thread counts are set correctly. However once the build has completed and the generated config.xml is checked, three queues don't have any thread counts associated with them at all; these queues are:
ThreadQueue (Count)
wli.internal.ProcessInstanceInfo (15)
wli.internal.ProcessTracking (15)
wli.process.event.thread.pool (15)
The other queues have all their thread counts set correctly. Does anybody know how these queues are used and if this will have an impact on the above problem?No worries about this post, the reason has been solved.
Reason
The default thread count for execute queues in development mode is 15 threads, therefore if a queue is specified with this value it won't be set in the config.xml (when an execute queue doesn't have a ThreadCount attribute, the default is used instead). The trouble is that when the server is started we enter production mode where the default thread count is 25 threads instead. -
I'm trying to write up advice for Storage node thread count.
Obviously, this is dependent of the machines - we have 48G 8processor machines, running 20 x2G Storage nodes and 2 x 2G Proxy nodes.
My understand is that when you don't have a cache-store you would set the thread-count to 0 (run on service thread).
I'm wondering how one might calculate this if you have a cache store ( and whether its different if you have write through / behind ).
I can see that in write-through it will spend most of the time with the database and a higher thread count would be good (its a pity its not a variable size pool with min and max).
I'm not sure about write-behind - I can imagine that a low thread count might be reasonable here.
I'm sure Rob (or others ) have a good formula for this.
Best, Andrew.I have a brief recollection (may be wrong about this) that there is some info (via JMX) about average thread pool usage that may be of assistance (if the average is close to your specified max it may pay off to increase it furthe and see if throughput goes up or down, if the average is close to zero it may be just as well to use zero thread count rather than what is used at the moment).
One must also decide what is most important - to max out throughput or to limit median response time. With a larger thread pool more queries can be in progress at the same time (hopefully reducing median responce time since long runing queries dont lock out other queries) but resources will be split between them (at least if the thread count is larger than the number of cores) making the total throughput lower...
/Magnus -
I understand that the thread-count determines the number of service threads for a particular service. So, if the thread-count for the ordersCache is 10, then a max of 10 OrderProcessor agents would be able to process orders at the same time.
Questions:
1. What happens if an 11th agent is invoked? Does it wait in a queue, waiting for a service thread to be freed? If so, how long does it wait?
2. For pure reads, does the thread-count matter? For instance, if I have a "public static NamedCache testCache" shared across 15 threads that are reading from the cache and if the thread-count defined is 10, then does that mean that only 10 threads can do "cache.get" at the same time? If so, what happens when the 11th thread calls "cache.get"? Does it wait?
Thanks
GhanshyamHi Ghanshyam,
I understand that the thread-count determines the
number of service threads for a particular service.
So, if the thread-count for the ordersCache is 10,
then a max of 10 OrderProcessor agents would be able
to process orders at the same time.Yes.
Questions:
1. What happens if an 11th agent is invoked? Does it
wait in a queue, waiting for a service thread to be
freed? If so, how long does it wait?Yes, it waits until a thread is available.
Starting in release 3.3, it will be possible to specify a time-out and/or a priority for your operations.
2. For pure reads, does the thread-count matter? For
instance, if I have a "public static NamedCache
testCache" shared across 15 threads that are reading
from the cache and if the thread-count defined is 10,
then does that mean that only 10 threads can do
"cache.get" at the same time? If so, what happens
when the 11th thread calls "cache.get"? Does it
wait?Yes, it waits. An in-memory read is typically so quick that it would not be an issue, but it could be if the gets were going to a CacheLoader and the database was slow or blocked. In the upcoming 3.3 release, it will be possible to specify time-outs for those operations.
Peace,
Cameron Purdy
Tangosol Coherence: The Java Data Grid -
OnAlarm/wait thread count configuration in BPEL
BPEL 10g, We have a com.oracle.bpel.expirationAgent.threadCount property to set the thread count for OnAlarm/wait. ORACLE_HOME/bpel/domain/DOMAIN_NAME/config/resources-quartz.properties.
BPEL 11g, Can you please point me the configuration file to set the thread count for OnAlarm/wait. I could not find the resources-quartz.properties in SOA 11g.Have you been able to find the file or a similar properties location in 11g? I am also looking for it.
-
Query Regarding Execute Thread Count
Hi,
My understanding of Execute Thread Count is the threads which are assigned to service
requests coming to Weblogic Server.
In our current architecture a request for generating report is directed to EJB method
which makes a call to another Server (Report Server for executing reports), the report
Server in turn calls the EJB residing on Weblogic Server for getting the data.
So, is my assumption correct that with our current architecture we are limited to
concurrency of Execute Thread Count -1. (Every request for report will consume 2
Excute threads, and others will have to wait till the first request gets completed
and 2 threads are freed).
I also read from the postings that Weblogic takes some of the threads, so it actually
will be limited to (Execute thread count - Weblogic Held- 1).
Please corect me if I am wrong.
Thanks and Regards
RashmiHi,
Thanks very much for the suggestion. I tried, and it is using 2 Execute Threads.
Thanks
Rashmi
"Dave Martin" <[email protected]> wrote:
>
Rashmi:
If you are interested in answering the question rigorously, why not just
throw a
fake exception along both the report generation path and the runReport method
of
the ReportServer, then record their stack traces to a file for comparison?
From the sounds of it, your app wants to make the actual report generation
asynchronous
from the post of the reporting request. If they're really asynchronous,
then your
ReportServer must have some kind of blocking queue that will "wake up" when
it has
work to be done. Depending on how you implemented this, your runReport
method may
not be running a WebLogic execute thread at all.
Seems to me that you should be thinking of this as consuming one execute
thread regardless.
Even if the two pieces of work are asynchronous, the first thread finished
its work
at the point that it posts to the second thread (at least, per your description).
So at any one time, at max one execute thread is being consumed per request.
But capture the stack traces and have a look for yourself.
Dave Martin
"Rashmi S" <[email protected]> wrote:
Hi,
Thanks for your reply.The reason why I say 2 threads will be consumed is
as follows
1. First execute thread will be used for the request to Weblogic Server
to run the
report originating from a client.
2. Now, within the ReportServices EJB there is a call to Report Serverto
run the
report. The running of report is done on the Report Server which is ona
separate
m/c. Within the runReport method of the Report Server ,the data is fetched
for the
report by making a context look-up of another session EJB which fetches
the data.
So, the second Execute thread will be consumed for executing the EJB that
fetches
the data.
Am I correct?
Thanks and Regards
Rashmi
"Cameron Purdy" <[email protected]> wrote:
Incoming requests (from outside the server) use one thread for the duration
of their processing. Previous to their processing they are queued until
a
thread is available. So you would use one thead per concurrent request,
not
two. Otherwise, it sounds right.
Peace,
Cameron Purdy
Tangosol, Inc.
Clustering Weblogic? You're either using Coherence, or you should be!
Download a Tangosol Coherence eval today at http://www.tangosol.com/
"Rashmi S" <[email protected]> wrote in message
news:[email protected]...
Hi,
My understanding of Execute Thread Count is the threads which are assignedto service
requests coming to Weblogic Server.
In our current architecture a request for generating report is directedto EJB method
which makes a call to another Server (Report Server for executingreports), the report
Server in turn calls the EJB residing on Weblogic Server for getting
the
data.
So, is my assumption correct that with our current architecture we arelimited to
concurrency of Execute Thread Count -1. (Every request for report willconsume 2
Excute threads, and others will have to wait till the first request
gets
completed
and 2 threads are freed).
I also read from the postings that Weblogic takes some of the threads,so
it actually
will be limited to (Execute thread count - Weblogic Held- 1).
Please corect me if I am wrong.
Thanks and Regards
Rashmi -
Configuring Execute Thread Count
Hi,
I am working on a webapp thats using Weblogic 8.1 (SP2) as the application server.
The admin server for Weblogic currently provides 2 execute threads as default.
However I was wondering if I can configure the admin server for a higher number
of execute threads.
I have the following questions:
1.> Does configuring the number of threads for the admin sever provide any additional
value? has any one come across a big application where more than 2 execute threads
are required for the admin server?
2.> If yes, is it possible to configure the number of execute threads for the
admin server? I did some search on the bea website and didnt find anything that
explicitly says that the property is configurable. Is this true?
3.> If it is configurable, then how do we go about setting it up?
If anyone has worked around this and can provide me with any information or references,
that would be great.
ThanksSriram,
Take a look at the following URL for setting the thread count in WLS 6.1
http://edocs.bea.com/wls/docs61/perform/WLSTuning.html#1112343
Chuck Nelson
Developer Relations Engineer
BEA Technical Support
-
Thread count in a distributed service
Hi,
i needed some information on the thread-count attribute in a distributed service. i would like to know how do we decide on the thread count. would increasing the thread count in the tangosol-coherence.xml increase the performance of tangosol as well.
looking forward to your help
Thanks
JigsTo go into this a bit further...
Each service instance has its own primary thread. This thread has the option of using its own isolated thread pool if the thread-count is greater than zero. If the thread-count is zero, then all work will be performed by the primary service thread. If the thread-count is greater than zero, then all work will be performed by the thread pool (the primary thread acts as a task coordinator).
Note that the thread-count is per-service and per-cluster-member. Each cache service has a unique name. The CacheFactory class uses a single cache service instance for each cache type (Replicated/Distributed/etc). If you manually create additional cache services, they will each have their own isolated thread pools.
Additionally, each Invocation service has its own thread pool. This setting is very important as the Invocation service is quite often used to execute long-running user tasks. If the thread pool is saturated with user tasks, then further calls to the Invocation service will block until user tasks complete and threads become available. Using named Invocation service instances will allow you to isolate critical tasks for better application availability.
Note that this information is current as of Coherence 2.5, but may change in future releases as threading models have a huge impact on performance and scalability. Having said that, there are no significant changes planned at this point in time.
Jon Purdy
Tangosol, Inc.
=================
Added 2004-12-02 by JP:
Also, as a very specific instance, the backup copy(ies) of a distributed cache are created on the service thread ... this means that while backup copies are being created on a node, no new tasks will be dispatched to that service's thread pool. For in-memory backup copies, this is actually optimal as adding an item to a HashMap/HashTable is roughly as fast as handing it off to another thread. However, if the backup-copy is configured to disk-based storage (or other high-latency resource), an asynchronous backup implementation should be used to eliminate latency. See AsyncBinaryStore for more information. -
Can I get the thread count back in ios7
After upgrading to ios7, I realized that when inside my inbox the total of unread emails is no longer at the top of the screen. Also in my threaded email conversations, it does not give me the number of emails in the thread, which it did prior to ios7. Is there a way to get this back or has it been removed. This is an aweful change, Why would apple get rid of the thread count.
No, the Disk Usage and Memory Usage pie charts are gone. Also gone is the Memory Usage Dock icon. (New graphs are much more useful, imo.)
-
Thread Count Vs no of simultaneous users
Hi,
I have been testing my customer made adapter on WLI2.1 and WLS6.1SP1 for
simulataneous users access.
I found that as long as the no of simultaneous users is less than the Thread
Count defined for WLI, it works fine.
But if the no of users exceed the thread count, the server hangs.....idle
threads become 0.
Does my Thread Count always have to be greater than the no of simulatneous
users am expecting ?
Thanks
AmitIn my openion, yes, number of threads configured should be greater than simultaneous users.
Thanks
Niranjan
Maybe you are looking for
-
Satellite U405 - What is the best way to increases battery life
Hi everyone! I've been looking for this quite a bit and all I could find are conflicting answers and some confusing instructions... :) Even the PDF file for my U405 has some conflicting tips (is it recommended to fully discharge it or not?) Let's say
-
How I can connect MacBook air to apple tv
-
Where does one find the product serial number?
Where does one find the product serial number to enter and get a "Generate a response code" to finalize the license if the computer has no internet access? I went to "Help>System Info" entered that number and it came up short of numbers.
-
I have a Visual Basic console application that runs other Visual Basic back-end applications on a scheduled and triggered basis. I have a list of scheduled and triggered entries in a database that this console application will read upon starting up.
-
How would I go about removing the required code from a form field?
I have this contact form that I downloaded and its fine. I just need to know how to remove the required field code from the contact no field. Thanks in advance Here is a link to the form and all of its assets http://www.formget.com/jquery-contact-for