Threads

I have on package which has run 3 sets of truncated all rows in the table and
then loads a new version of the table. The loading of the data into the
table run consecutively rather concurrently. I there a way that I can make
then run concurrently?

Likewise, if I have a package which executes 2 packages (I have not done
this yet), can I somehow make each thread run concurrently?

Hello Jim,


Things execute consecutively when you join them by workflow. They execute
concurrently when you don't. That said I would not have more steps going
at the same time than I have processors in the box and you must also specify
that you want more than one thing to go at the same time in the package properties
(default == 4)




Allan Mitchell
www.SQLDTS.com
www.SQLIS.com
www.Konesans.com

I have on package which has run 3 sets of truncated all rows in the
table and then loads a new version of the table. The loading of the
data into the table run consecutively rather concurrently. I there a
way that I can make then run concurrently?

Likewise, if I have a package which executes 2 packages (I have not
done this yet), can I somehow make each thread run concurrently?

Thanks, A couple of follow up questions.

When you say "when you join them by workflow" what exactly does this mean.
They are all using the same connection object...is this "joined by workflow"?
So if I create seperate connection objects, then they have the potential of
being "seperate"?

When I look out and see how much CPU is being utilized when they are run
consecutively, it appears to be about 20 to 30%. So why would I only run one
process per CPU? I am guessing that they are more IO bound then CPU bound
and perhaps I would get no more real thru-put by running concurrently.

Again, thanks for your thoughts/recommendations

For workflow look here

Introduction to Workflow
(http://www.sqldts.com/default.aspx?287)


If everything is joined to the self same connection manager then only one of
them is really going at any one time. Create seperate connectionn objects
for each pump.

Processes need to scheduled on and off a processor. If you have too many
things going at the same time then you will see a lot of context switching
on the CPUs and this will cost you.

Sure Disk speed will hurt you as well. Have a slow disk then it cannot be
written to quickly and things will take time.



--


Allan Mitchell
www.SQLDTS.com
www.SQLIS.com
www.Konesans.com


"Jim Heavey" <JimHeavey (AT) discussions (DOT) microsoft.com> wrote

Thanks, A couple of follow up questions.

When you say "when you join them by workflow" what exactly does this mean.
They are all using the same connection object...is this "joined by
workflow"?
So if I create seperate connection objects, then they have the potential
of
being "seperate"?

When I look out and see how much CPU is being utilized when they are run
consecutively, it appears to be about 20 to 30%. So why would I only run
one
process per CPU? I am guessing that they are more IO bound then CPU bound
and perhaps I would get no more real thru-put by running concurrently.

Again, thanks for your thoughts/recommendations