Performance Issue

Performance Issue

Post by Heath Norto » Fri, 22 Feb 2002 21:40:05



 Hi all
This is a comparison between these two machines. Could anyone help me on what config parameters I could tune to improve the Sun E550.

      SUN E5500
     Fujitsu Li Server

      6 x 400MHz processors
     2 x 200MHz processors

      2 Gig Memory
     512 MB Memory

      Spark Technology
     Intel Based

A test was carried out on both machines to load 480,000 transactions from an asci file.The exercise took the following times on the two machines:

SUN E5500                 -            6 minutes

Fujitsu Li Server  -            4 minutes

Besides, the Li Server was under heavy usage at the time of the exercise while the E5500 was not being used at all.Clearly this is not the expected performance of the SUN E5500.The Informix running on these servers is as follows:

      SUN E5500
     Fujitsu Li Server

      Informix Dynamic Server Version: 7.30.UC2
     Informix Dynamic Server Version: 7.31.UC5

INTEL

-------------------------

MULTIPROCESSOR 1    #  0 for single-processor, 1 for multi-processor

NUMCPUVPS 2             # Number of user (cpu) vps

SINGLE_CPU_VP 0         # If non-zero, limit number of cpu vps to one

NOAGE 0                        # Process aging

AFF_SPROC 0               # Affinity start processor

AFF_NPROCS 0             # Affinity number of processors

SUN

-----------

MULTIPROCESSOR 1 # 0 for single-processor, 1 for multi-processor

NUMCPUVPS 2          # Number of user (cpu) vps

SINGLE_CPU_VP 0    # If non-zero, limit number of cpu vps to one

NOAGE 0                   # Process aging

AFF_SPROC 0            # Affinity start processor

AFF_NPROCS 0         # Affinity number of processors

Any Help would be Appreciated

Heath

 
 
 

Performance Issue

Post by Obnoxio The Clo » Fri, 22 Feb 2002 23:36:35


Are you saying that the Sun was running a different version of IDS
from the Intel box?

On Thu, 21 Feb 2002 14:40:05 +0200, "Heath Norton"


> Hi all
>This is a comparison between these two machines. Could anyone help me on what config parameters I could tune to improve the Sun E550.

>      SUN E5500
>     Fujitsu Li Server

>      6 x 400MHz processors
>     2 x 200MHz processors

>      2 Gig Memory
>     512 MB Memory

>      Spark Technology
>     Intel Based

>A test was carried out on both machines to load 480,000 transactions from an asci file.The exercise took the following times on the two machines:

>SUN E5500                 -            6 minutes

>Fujitsu Li Server  -            4 minutes

>Besides, the Li Server was under heavy usage at the time of the exercise while the E5500 was not being used at all.Clearly this is not the expected performance of the SUN E5500.The Informix running on these servers is as follows:

>      SUN E5500
>     Fujitsu Li Server

>      Informix Dynamic Server Version: 7.30.UC2
>     Informix Dynamic Server Version: 7.31.UC5

>INTEL

>-------------------------

>MULTIPROCESSOR 1    #  0 for single-processor, 1 for multi-processor

>NUMCPUVPS 2             # Number of user (cpu) vps

>SINGLE_CPU_VP 0         # If non-zero, limit number of cpu vps to one

>NOAGE 0                        # Process aging

>AFF_SPROC 0               # Affinity start processor

>AFF_NPROCS 0             # Affinity number of processors

>SUN

>-----------

>MULTIPROCESSOR 1 # 0 for single-processor, 1 for multi-processor

>NUMCPUVPS 2          # Number of user (cpu) vps

>SINGLE_CPU_VP 0    # If non-zero, limit number of cpu vps to one

>NOAGE 0                   # Process aging

>AFF_SPROC 0            # Affinity start processor

>AFF_NPROCS 0         # Affinity number of processors

>Any Help would be Appreciated

>Heath


 
 
 

Performance Issue

Post by Bill Dar » Fri, 22 Feb 2002 23:32:09


All of the information you give below has only to do with CPU resources.  If
you are loading 480,000 rows in a few minutes without any massaging of the
data, your bottleneck is undoubtedly going to be I/O bandwidth.  So, what
kind of disks you got?  How are they configured, RAID what?  What does sar
-d look like when the load is running?  

Regards,
Bill

> -----Original Message-----

> Sent:      Thursday, February 21, 2002 7:40 AM

> Subject:   Performance Issue

>  Hi all
> This is a comparison between these two machines. Could anyone help me on
> what config parameters I could tune to improve the Sun E550.

>       SUN E5500
>      Fujitsu Li Server

>       6 x 400MHz processors
>      2 x 200MHz processors

>       2 Gig Memory
>      512 MB Memory

>       Spark Technology
>      Intel Based

> A test was carried out on both machines to load 480,000 transactions from
> an asci file.The exercise took the following times on the two machines:

> SUN E5500                 -            6 minutes

> Fujitsu Li Server  -            4 minutes

> Besides, the Li Server was under heavy usage at the time of the exercise
> while the E5500 was not being used at all.Clearly this is not the expected
> performance of the SUN E5500.The Informix running on these servers is as
> follows:

>       SUN E5500
>      Fujitsu Li Server

>       Informix Dynamic Server Version: 7.30.UC2
>      Informix Dynamic Server Version: 7.31.UC5

> INTEL

> -------------------------

> MULTIPROCESSOR 1    #  0 for single-processor, 1 for multi-processor

> NUMCPUVPS 2             # Number of user (cpu) vps

> SINGLE_CPU_VP 0         # If non-zero, limit number of cpu vps to one

> NOAGE 0                        # Process aging

> AFF_SPROC 0               # Affinity start processor

> AFF_NPROCS 0             # Affinity number of processors

> SUN

> -----------

> MULTIPROCESSOR 1 # 0 for single-processor, 1 for multi-processor

> NUMCPUVPS 2          # Number of user (cpu) vps

> SINGLE_CPU_VP 0    # If non-zero, limit number of cpu vps to one

> NOAGE 0                   # Process aging

> AFF_SPROC 0            # Affinity start processor

> AFF_NPROCS 0         # Affinity number of processors

> Any Help would be Appreciated

> Heath

 
 
 

Performance Issue

Post by Jack Parke » Sat, 23 Feb 2002 00:14:47


Oh dear.

Start with the basics.  How fast can you read/write to the raw disks on each
machine?  Pick a dummy disk somewhere and dd 5MB to and from it.  From that,
figure out how big your table is (or is going to be) and figure out how long
it SHOULD take to read/write that table.  Now you have a target to shoot
for.

You don't tell us a lot about your $ONCONFIG, but it looks like your only
giving 2 of 6 processors to the engine?  As a general rule, don't bother
with multiple CPUs unless you're going to give it 4.  (how many people do
you think will scream at that one?)

How are you loading?  LOAD?  dbload?  HPL?
Is the table logged on both machines?
Are both tables sized the same?
Are tables fragmented across disks?
Check your log activity between them both.

So many questions...

cheers
j.

----- Original Message -----


Sent: Thursday, February 21, 2002 7:40 AM
Subject: Performance Issue

> Hi all
> This is a comparison between these two machines. Could anyone help me on

what config parameters I could tune to improve the Sun E550.
Quote:

>       SUN E5500
>      Fujitsu Li Server

>       6 x 400MHz processors
>      2 x 200MHz processors

>       2 Gig Memory
>      512 MB Memory

>       Spark Technology
>      Intel Based

> A test was carried out on both machines to load 480,000 transactions from

an asci file.The exercise took the following times on the two machines:
Quote:

> SUN E5500                 -            6 minutes

> Fujitsu Li Server  -            4 minutes

> Besides, the Li Server was under heavy usage at the time of the exercise

while the E5500 was not being used at all.Clearly this is not the expected
performance of the SUN E5500.The Informix running on these servers is as
follows:
Quote:

>       SUN E5500
>      Fujitsu Li Server

>       Informix Dynamic Server Version: 7.30.UC2
>      Informix Dynamic Server Version: 7.31.UC5

> INTEL

> -------------------------

> MULTIPROCESSOR 1    #  0 for single-processor, 1 for multi-processor

> NUMCPUVPS 2             # Number of user (cpu) vps

> SINGLE_CPU_VP 0         # If non-zero, limit number of cpu vps to one

> NOAGE 0                        # Process aging

> AFF_SPROC 0               # Affinity start processor

> AFF_NPROCS 0             # Affinity number of processors

> SUN

> -----------

> MULTIPROCESSOR 1 # 0 for single-processor, 1 for multi-processor

> NUMCPUVPS 2          # Number of user (cpu) vps

> SINGLE_CPU_VP 0    # If non-zero, limit number of cpu vps to one

> NOAGE 0                   # Process aging

> AFF_SPROC 0            # Affinity start processor

> AFF_NPROCS 0         # Affinity number of processors

> Any Help would be Appreciated

> Heath

 
 
 

Performance Issue

Post by Bill Dar » Sat, 23 Feb 2002 00:54:14


> -----Original Message-----

> Sent:      Thursday, February 21, 2002 10:15 AM

> Subject:   Re: Performance Issue

> Oh dear.

> Start with the basics.  How fast can you read/write to the raw disks on
> each
> machine?  Pick a dummy disk somewhere and dd 5MB to and from it.  From
> that,
> figure out how big your table is (or is going to be) and figure out how
> long
> it SHOULD take to read/write that table.  Now you have a target to shoot
> for.

> You don't tell us a lot about your $ONCONFIG, but it looks like your only
> giving 2 of 6 processors to the engine?  As a general rule, don't bother
> with multiple CPUs unless you're going to give it 4.  (how many people do
> you think will scream at that one?)

        I wouldn't scream, but I would like to know why.

        Bill

- Show quoted text -

> How are you loading?  LOAD?  dbload?  HPL?
> Is the table logged on both machines?
> Are both tables sized the same?
> Are tables fragmented across disks?
> Check your log activity between them both.

> So many questions...

> cheers
> j.

> ----- Original Message -----


> Sent: Thursday, February 21, 2002 7:40 AM
> Subject: Performance Issue

> > Hi all
> > This is a comparison between these two machines. Could anyone help me on
> what config parameters I could tune to improve the Sun E550.

> >       SUN E5500
> >      Fujitsu Li Server

> >       6 x 400MHz processors
> >      2 x 200MHz processors

> >       2 Gig Memory
> >      512 MB Memory

> >       Spark Technology
> >      Intel Based

> > A test was carried out on both machines to load 480,000 transactions
> from
> an asci file.The exercise took the following times on the two machines:

> > SUN E5500                 -            6 minutes

> > Fujitsu Li Server  -            4 minutes

> > Besides, the Li Server was under heavy usage at the time of the exercise
> while the E5500 was not being used at all.Clearly this is not the expected
> performance of the SUN E5500.The Informix running on these servers is as
> follows:

> >       SUN E5500
> >      Fujitsu Li Server

> >       Informix Dynamic Server Version: 7.30.UC2
> >      Informix Dynamic Server Version: 7.31.UC5

> > INTEL

> > -------------------------

> > MULTIPROCESSOR 1    #  0 for single-processor, 1 for multi-processor

> > NUMCPUVPS 2             # Number of user (cpu) vps

> > SINGLE_CPU_VP 0         # If non-zero, limit number of cpu vps to one

> > NOAGE 0                        # Process aging

> > AFF_SPROC 0               # Affinity start processor

> > AFF_NPROCS 0             # Affinity number of processors

> > SUN

> > -----------

> > MULTIPROCESSOR 1 # 0 for single-processor, 1 for multi-processor

> > NUMCPUVPS 2          # Number of user (cpu) vps

> > SINGLE_CPU_VP 0    # If non-zero, limit number of cpu vps to one

> > NOAGE 0                   # Process aging

> > AFF_SPROC 0            # Affinity start processor

> > AFF_NPROCS 0         # Affinity number of processors

> > Any Help would be Appreciated

> > Heath

 
 
 

Performance Issue

Post by Jack Parke » Sat, 23 Feb 2002 01:16:56


----- Original Message -----



Sent: Thursday, February 21, 2002 10:54 AM
Subject: RE: Performance Issue

> > You don't tell us a lot about your $ONCONFIG, but it looks like your
only
> > giving 2 of 6 processors to the engine?  As a general rule, don't bother
> > with multiple CPUs unless you're going to give it 4.  (how many people
do
> > you think will scream at that one?)

> I wouldn't scream, but I would like to know why.

> Bill

<Squirm>

It just isn't worth it.

Running an engine with one CPU has historically been faster than running
with 2CPUs and multiple cpus turned on.

I feel that you don't get the synchronicity benefit until you get up to 4.
It's a base point for multi-cpu.

cheers
j.

 
 
 

Performance Issue

Post by Bill Dar » Sat, 23 Feb 2002 01:43:25


> -----Original Message-----

> Sent:      Thursday, February 21, 2002 11:17 AM

> Subject:   Re: Performance Issue

> ----- Original Message -----



> Sent: Thursday, February 21, 2002 10:54 AM
> Subject: RE: Performance Issue

> > > You don't tell us a lot about your $ONCONFIG, but it looks like your
> only
> > > giving 2 of 6 processors to the engine?  As a general rule, don't
> bother
> > > with multiple CPUs unless you're going to give it 4.  (how many people
> do
> > > you think will scream at that one?)

> > I wouldn't scream, but I would like to know why.

> > Bill

> <Squirm>

> It just isn't worth it.

> Running an engine with one CPU has historically been faster than running
> with 2CPUs and multiple cpus turned on.

> I feel that you don't get the synchronicity benefit until you get up to 4.
> It's a base point for multi-cpu.

> cheers
> j.

        I have to disagree.  I've worked quite a bit with HP and Sun boxes,
OLTP applications in the cellular phone business, and I've seen considerable
difference between a multiple CPU box with 2 CPUVPs and 4 CPUVPs.  When
benchmarking the apps the method I used for simulating the size of the
system (1 CPU, 2 CPU, 4 CPU, etc.) was by adding CPUVPs and the performance
increase in adding the CPUVPs was close to linear
        These particular applications  did almost all their processing in
the DB server.  No users, just application processes banging away at the
database.  Biggest systems were doing fraud analysis on call detail records.
        Seems likely that the characteristics of the application might have
something to do with this.  

        Regards,
        Bill

 
 
 

Performance Issue

Post by Bill Dar » Sat, 23 Feb 2002 01:58:10


> -----Original Message-----

> Sent:      Thursday, February 21, 2002 11:51 AM

> Subject:   Re: Performance Issue

> > > I feel that you don't get the synchronicity benefit until you get up
> to
> 4.
> > > It's a base point for multi-cpu.

> > > cheers
> > > j.

> > I have to disagree.  I've worked quite a bit with HP and Sun boxes,
> > OLTP applications in the cellular phone business, and I've seen
> considerable
> > difference between a multiple CPU box with 2 CPUVPs and 4 CPUVPs.  When
> > benchmarking the apps the method I used for simulating the size of the
> > system (1 CPU, 2 CPU, 4 CPU, etc.) was by adding CPUVPs and the
> performance
> > increase in adding the CPUVPs was close to linear
> > These particular applications  did almost all their processing in
> > the DB server.  No users, just application processes banging away at the
> > database.  Biggest systems were doing fraud analysis on call detail
> records.
> > Seems likely that the characteristics of the application might have
> > something to do with this.

> > Regards,
> > Bill

> Interesting.  So from 1 to 4 the performance improvement was linear?  And

        Not linear, but pretty  close.  Double the number of CPUVPs and I'd
get maybe 85% improvement in throughput.

Quote:> we're just talking CPUVPs?  I'd be interested in reading your findings if
> you kept them around a bit.  Always good to be debunked (me that is, not
> you).  I have no benchmark data to back what I claim - hence my
> squirminess.
> Just something I picked up back when in this group.

        Not unless I go back to my previous employer and see if they still
have my old PC around.  This is from memory. But I did spend 2 weeks each at
Sun and HP doing the benchmarks so I don't believe I could be too far off.

        Regards,
        Bill

- Show quoted text -

Quote:> cheers
> j.

 
 
 

Performance Issue

Post by Jack Parke » Sat, 23 Feb 2002 02:10:34


Quote:

> > Interesting.  So from 1 to 4 the performance improvement was linear?
And

> Not linear, but pretty  close.  Double the number of CPUVPs and I'd
> get maybe 85% improvement in throughput.

85% is close enough.  I remember what I forgot to ask.  Did you have
MULTIPROCESSOR and NUMCPUVPS set to a single proc when you were running the
1 CPU version?  There was an issue with 7.0 being slower than 5.0, hence why
those settings were put into ONCONFIG.

Quote:

> > we're just talking CPUVPs?  I'd be interested in reading your findings
if
> > you kept them around a bit.  Always good to be debunked (me that is, not
> > you).  I have no benchmark data to back what I claim - hence my
> > squirminess.
> > Just something I picked up back when in this group.

> Not unless I go back to my previous employer and see if they still
> have my old PC around.  This is from memory. But I did spend 2 weeks each
at
> Sun and HP doing the benchmarks so I don't believe I could be too far off.

Bummer.  If I had a multi-cpu box lying around I'd give it a whirl right
now.  Hmmm, wonder if....

cheers
j.

 
 
 

Performance Issue

Post by Bill Dar » Sat, 23 Feb 2002 03:14:53


> -----Original Message-----

> Sent:      Thursday, February 21, 2002 12:11 PM

> Subject:   Re: Performance Issue

> > > Interesting.  So from 1 to 4 the performance improvement was linear?
> And

> > Not linear, but pretty  close.  Double the number of CPUVPs and I'd
> > get maybe 85% improvement in throughput.

> 85% is close enough.  I remember what I forgot to ask.  Did you have
> MULTIPROCESSOR and NUMCPUVPS set to a single proc when you were running
> the
> 1 CPU version?  There was an issue with 7.0 being slower than 5.0, hence
> why
> those settings were put into ONCONFIG.

        Yes, they were set correctly.  I think the performance hit came if
you had one CPUVP and did not have SINGLE_CPU_VP set to something other than
0.

- Show quoted text -

Quote:

> > > we're just talking CPUVPs?  I'd be interested in reading your findings
> if
> > > you kept them around a bit.  Always good to be debunked (me that is,
> not
> > > you).  I have no benchmark data to back what I claim - hence my
> > > squirminess.
> > > Just something I picked up back when in this group.

> > Not unless I go back to my previous employer and see if they still
> > have my old PC around.  This is from memory. But I did spend 2 weeks
> each
> at
> > Sun and HP doing the benchmarks so I don't believe I could be too far
> off.

> Bummer.  If I had a multi-cpu box lying around I'd give it a whirl right
> now.  Hmmm, wonder if....

> cheers
> j.

 
 
 

Performance Issue

Post by Jack Parke » Sat, 23 Feb 2002 01:51:20


Quote:> > I feel that you don't get the synchronicity benefit until you get up to
4.
> > It's a base point for multi-cpu.

> > cheers
> > j.

> I have to disagree.  I've worked quite a bit with HP and Sun boxes,
> OLTP applications in the cellular phone business, and I've seen
considerable
> difference between a multiple CPU box with 2 CPUVPs and 4 CPUVPs.  When
> benchmarking the apps the method I used for simulating the size of the
> system (1 CPU, 2 CPU, 4 CPU, etc.) was by adding CPUVPs and the
performance
> increase in adding the CPUVPs was close to linear
> These particular applications  did almost all their processing in
> the DB server.  No users, just application processes banging away at the
> database.  Biggest systems were doing fraud analysis on call detail
records.
> Seems likely that the characteristics of the application might have
> something to do with this.

> Regards,
> Bill

Interesting.  So from 1 to 4 the performance improvement was linear?  And
we're just talking CPUVPs?  I'd be interested in reading your findings if
you kept them around a bit.  Always good to be debunked (me that is, not
you).  I have no benchmark data to back what I claim - hence my squirminess.
Just something I picked up back when in this group.

cheers
j.

 
 
 

Performance Issue

Post by Jack Parke » Sat, 23 Feb 2002 03:20:58


I need to find myself a box to recreate.

cheers
j.

----- Original Message -----



Sent: Thursday, February 21, 2002 1:14 PM
Subject: RE: Performance Issue

> > -----Original Message-----

> > Sent: Thursday, February 21, 2002 12:11 PM

> > Subject: Re: Performance Issue

> > > > Interesting.  So from 1 to 4 the performance improvement was linear?
> > And

> > > Not linear, but pretty  close.  Double the number of CPUVPs and I'd
> > > get maybe 85% improvement in throughput.

> > 85% is close enough.  I remember what I forgot to ask.  Did you have
> > MULTIPROCESSOR and NUMCPUVPS set to a single proc when you were running
> > the
> > 1 CPU version?  There was an issue with 7.0 being slower than 5.0, hence
> > why
> > those settings were put into ONCONFIG.

> Yes, they were set correctly.  I think the performance hit came if
> you had one CPUVP and did not have SINGLE_CPU_VP set to something other
than
> 0.

> > > > we're just talking CPUVPs?  I'd be interested in reading your
findings
> > if
> > > > you kept them around a bit.  Always good to be debunked (me that is,
> > not
> > > > you).  I have no benchmark data to back what I claim - hence my
> > > > squirminess.
> > > > Just something I picked up back when in this group.

> > > Not unless I go back to my previous employer and see if they still
> > > have my old PC around.  This is from memory. But I did spend 2 weeks
> > each
> > at
> > > Sun and HP doing the benchmarks so I don't believe I could be too far
> > off.

> > Bummer.  If I had a multi-cpu box lying around I'd give it a whirl right
> > now.  Hmmm, wonder if....

> > cheers
> > j.

 
 
 

Performance Issue

Post by Neil Trub » Sat, 23 Feb 2002 06:55:57


did you have the same logging status for each db?
and were they using identical disk sub-systems?


Quote:

>  Hi all
> This is a comparison between these two machines. Could anyone help me on

what config parameters I could tune to improve the Sun E550.
Quote:

>       SUN E5500
>      Fujitsu Li Server

>       6 x 400MHz processors
>      2 x 200MHz processors

>       2 Gig Memory
>      512 MB Memory

>       Spark Technology
>      Intel Based

> A test was carried out on both machines to load 480,000 transactions from

an asci file.The exercise took the following times on the two machines:
Quote:

> SUN E5500                 -            6 minutes

> Fujitsu Li Server  -            4 minutes

> Besides, the Li Server was under heavy usage at the time of the exercise

while the E5500 was not being used at all.Clearly this is not the expected
performance of the SUN E5500.The Informix running on these servers is as
follows:
Quote:

>       SUN E5500
>      Fujitsu Li Server

>       Informix Dynamic Server Version: 7.30.UC2
>      Informix Dynamic Server Version: 7.31.UC5

> INTEL

> -------------------------

> MULTIPROCESSOR 1    #  0 for single-processor, 1 for multi-processor

> NUMCPUVPS 2             # Number of user (cpu) vps

> SINGLE_CPU_VP 0         # If non-zero, limit number of cpu vps to one

> NOAGE 0                        # Process aging

> AFF_SPROC 0               # Affinity start processor

> AFF_NPROCS 0             # Affinity number of processors

> SUN

> -----------

> MULTIPROCESSOR 1 # 0 for single-processor, 1 for multi-processor

> NUMCPUVPS 2          # Number of user (cpu) vps

> SINGLE_CPU_VP 0    # If non-zero, limit number of cpu vps to one

> NOAGE 0                   # Process aging

> AFF_SPROC 0            # Affinity start processor

> AFF_NPROCS 0         # Affinity number of processors

> Any Help would be Appreciated

> Heath

 
 
 

1. OPENROWSET performance issue or SET FMTONLY OFF issue

I apologize if someone has already got this answer posted and answered
here, but I could not find it with the search I was using.  So here it
is (if it is redundant):

We are using SELECT * INTO #temp OPENROWSET (SET FMTONLY OFF EXEC
sp_name) in a short-hand version but you get the idea.  From time to
time, and for specific user, this is extremely slow and it actually
hangs for 4 mins.  Then for some period of time, for the same user, it
executes without problems.  The stored procedure called 'sp_name' has
#temp tables created within so we had to use SET FMTONLY OFF in order
not to get dreaded 'Invalid object name ...' error.  Now what is going
on behind the scenes to make OPENROWSET slow?  Does SET FMTONLY OFF
causes perf issues or are there issues with using this option and ADO
- SQLOLEDB?  Is there any other solution for #temp tables and
OPENROWSET issue?  Can you share any info that can help me?

Thanks,
Tigger

2. Q: Creating a copy of a database

3. Performance issues with IN (.....)

4. Bit-Mapped Indexing.....(Beta Testers Wanted)

5. SQL Server 7.0 performance issues

6. linking a table form another SQL Server iso importing

7. Query problem - serious performance issues. PLEASE HELP

8. Ingres 6.4 to Oracle 7 database migration tools

9. Performance issue

10. Performance issues with linked Oracle database

11. Database Design and Performance Issue

12. performance issues in SQL Server

13. 6.5 7.0 Upgrade SQL Server Performance Issues Stored Procedure