Re: Help - Timing Logic

It is extremely important for the employees to receive the messages almost immediately after identifying (time val in table). So, threading ensures I can execute multiple processes at the same time... only problem is these threads read from the same queue table and have the potential to send the same message to the same guy x number of times. where x is the number of threads running.

"jeff" <jhersey at allnorth dottt com> wrote in message news:uQ7%23Y4wEHHA.4404@xxxxxxxxxxxxxxxxxxxxxxx

Still do not know why you need 4 threads ? Speed ? Does runnig the process in one thread not work fast enough for you since timing is import ? Why 4 threads why not 6,12 , 3, 36 ... ? How do you determine which row a 'thread retrieves' ??? does it simply select the next available row from the DATABASE ??? How does the thread determine 'the next row'???

If mutli-threading is duplicating 'send messages' it is not working properly...
If you are multi-threading to improve performance ... maybe look at the design...
If you need to implement a locking mechanism / or / logging mechanism / or / a checking mechanism to avoid duplicate messages caused by multi-threading ... these will all come at a cost ... performance cost!

What is bottle necking your process that you need 4 threads? Is it the READ from the database ... is it the SENDING the text message? Is it connecting to the database ??? do not know here ...

Look at what is causing problem ...

- you need better performance ... so, multiple thread it! However, mutli-threading the entire process is causing issues ... duplicate records ... now you either need to incorporate a locking procedure ... or a checking procedure to avoid duplicate messages... all this has a cost to the overall performance ... adding more threads may in fact negatively impact the overall performance....

Maybe implement a message broker ...

- message broker gets all the necessary messages or message id's from the database to be sent ... in-memory list...
- message broker loops through list of messageIDs ...
- message broker starts another thread for processing the sendtextmessage functionality for each messageID...
- message broker will include the messageID so the process knows which message to get and process...
- message broker will ensure each database row is only processed once...

It is very hard to help you without knowing exactly what you are doing...

Again, I will ask, what rational / reasoning did you use for using 4 threads ? Performance ? Speed ? ... Where is the bottle neck in the process that requires you to multi-thread it? Maybe just move the 'bottle neck section' to another thread ... ?

Grasping at straws ...

Code would be nice how to see what you are doing ...

Jeff















"Jay" <someone@xxxxxxxxxxxxx> wrote in message news:uBX7QewEHHA.1224@xxxxxxxxxxxxxxxxxxxxxxx
Thanks. I do need to select an individual row at a time and all 4 threads need to do this. What trasaction isolation level would you recommend? Perhaps a stored proc may be faster to execute and return the values as opposed to building the transaction in the code. What has to happen is every 5 seconds, and for each thread, a sub runs to get a single row then send the message to dtr("numbertosendto"). Because this app heavily relies on timing it is important that all threads run and only one distinct row can be returned at a time for each thread.



"jeff" <jhersey at allnorth dottt com> wrote in message news:%23qVFIYwEHHA.3576@xxxxxxxxxxxxxxxxxxxxxxx

are you selecting individual rows at a time from the database table ...

if so ... use a transaction ...

begin transaction ...

select a row from database ...

update the row in table set flag = 'Processed'

end transaction

this will lock the row until the end of transaction is issued ... as long as the isolation level is set accordingly...

Again, please let us know how you are getting the information from the database... then we can help!

If you are reading a bunch of rows in one statement, storing the rows in in-memory datasets on the workstation, looping through the rows one by one ... then you may need to either ...

implement as above locking only the records you retrieve / update - need to watch out here for table locking ... may impact performance,
implement using an update flag at sent and 'check before send method'...
...
lots of options here, just need to know how you are retrieving your data...and how you are processing the data.

Question, why do you need 4 threads running ... are they doing 'different processes', sending different 'types' of messages ... sending the same message ... just need 4 to make it faster ... what is the logic ...

Tieing a record to a thread my cause problems in the future ... what happens when a thread stops or hangs ... those messages will not be processed... When happens if somebody changes the ThreadID ... in the program ...

trying to help ..

Jeff

PS: you can lock individual rows ... look at how database transactions work and incorporate it in you program ...








"Jay" <someone@xxxxxxxxxxxxx> wrote in message news:%23f70LJwEHHA.2328@xxxxxxxxxxxxxxxxxxxxxxx
Thanks for the reply.

I do need all 4 threads running (maybe even more in the future). I can not delete the row from the table... it needs to be there for later update use. I am considering marking each row at insert (1 - 4) then having each of the 4 threads only select a row based on the mark. It's becoming a little tricky... it would be great if I could lock a single row when selecting it.


"jeff" <jhersey at allnorth dottt com> wrote in message news:OPCyvDwEHHA.4740@xxxxxxxxxxxxxxxxxxxxxxx

Question:

What is the reason for the 4 threads?

Do you need all these threads processing the same dataset?

Just asking, maybe you could change your approach to avoid this problem...

If not, investigate Record Locking / Transactions...

How are you retrieving your records in to memory for processing ...

ie,
- are you selecting on the entire contents of a 'message' table and looping through an in-memory dataset...

- do you retrieve 'chunks' or rows from the message table based on parameters...

You could use the internal locking mechanisms of the database to achieve this ... but procede with caution...

Start Transaction...
SELECT * FROM MESSAGE TABLE
...
stuff this into a data table on the desktop
...
DELETE * FROM MESSAGE TABLE
...
punt / clean the table.
...
End Transaction.

The transactions places locks on the table ... but selecting the entire tables contents ... you are essentially 'LOCKING' the entire table until the transaction is over...

So, each thread's 'retrieve' process will have to wait in line until the select and delete are processed. As well, anything triggering 'New Messages to be Added' will be delayed until the Transaction is completed...

You may impact the overall performance of you application ... need to investigate.

If locking does not work for you, you will need to implement some type of logging / checking approach...

ie... have a log table...once a process has sent the message, write to a log table ... message sent. Before each message is compiled and sent, check the log table to see if another process has sent the message...if not, send you message...


A snip of code would be great to figure out how you are retrieving a list of messages and how you are processing each record...

Jeff.


"Jay" <someone@xxxxxxxxxxxxx> wrote in message news:O2phzwvEHHA.4280@xxxxxxxxxxxxxxxxxxxxxxx
I have a multi threaded VB.NET application (4 threads) that I use to send text messages to many, many employees via system.timer at a 5 second interval. Basically, I look in a SQL table (queue) to determine who needs to receive the text message then send the message to the address. Only problem is, the employee may receive up to 4 of the same messages because each thread gets the recors then sends the message. I need somehow to prevent this... just can't think of how. Somehow I need the other threads to know that another thread is already using that record and move on to the next record. I thought of getting the record then marking it (column value) as in use and writing the code to look for records only where not in use... problem is all 4 run fast enough to all use/mark the row. Any thoughts?

Thanks a lot.

Jay










.

Relevant Pages

  • Re: Help - Timing Logic
    ... put a trigger on the database table. ... Transaction and Database Locking - look at isolation levels / settings ... Maybe implement a message broker ... ...
    (microsoft.public.dotnet.languages.vb)
  • Re: Help - Timing Logic
    ... build a trigger on the database .. ... Transaction and Database Locking - look at isolation levels / settings ... If you need to implement a locking mechanism / or / logging mechanism / or / a checking mechanism to avoid duplicate messages caused by multi-threading ... ... message broker gets all the necessary messages or message id's from the database to be sent ... ...
    (microsoft.public.dotnet.languages.vb)
  • Re: Help - Timing Logic
    ... put a trigger on the database table. ... Transaction and Database Locking - look at isolation levels / settings ... Maybe implement a message broker ... ...
    (microsoft.public.dotnet.languages.vb)
  • Re: Help - Timing Logic
    ... from the database ... ... now you either need to incorporate a locking procedure ... ... opposed to building the transaction in the code. ... implement as above locking only the records you retrieve / update - need ...
    (microsoft.public.dotnet.languages.vb)
  • Re: Help - Timing Logic
    ... If you DBA screams and yells and jumps up and down - this is a database server, we are not install you little program on it ... ... Retrieve list of messages from the database... ... will create a lot of wasted CHECKING and UPDATING and DATABASE LOCKING to ensure messages are not duplicated between recipients. ... Transaction and Database Locking - look at isolation levels / settings ...
    (microsoft.public.dotnet.languages.vb)