Sybase NNTP forums - End Of Life (EOL)

The NNTP forums from Sybase - forums.sybase.com - are now closed.

All new questions should be directed to the appropriate forum at the SAP Community Network (SCN).

Individual products have links to the respective forums on SCN, or you can go to SCN and search for your product in the search box (upper right corner) to find your specific developer center.

Performance problem (Linux newsgroup is not available again and again and again...)

7 posts in General Discussion Last posting was on 2007-05-09 14:29:19.0Z
Alex Posted on 2007-05-08 11:30:52.0Z
Sender: 3d86.46363f95.1804289383@sybase.com
From: Alex
Newsgroups: ianywhere.public.general
Subject: Performance problem (Linux newsgroup is not available again and again and again...)
X-Mailer: WebNews to Mail Gateway v1.1t
Message-ID: <46405f6c.5970.1681692777@sybase.com>
NNTP-Posting-Host: 10.22.241.41
X-Original-NNTP-Posting-Host: 10.22.241.41
Date: 8 May 2007 04:30:52 -0700
X-Trace: forums-1-dub 1178623852 10.22.241.41 (8 May 2007 04:30:52 -0700)
X-Original-Trace: 8 May 2007 04:30:52 -0700, 10.22.241.41
Lines: 13
Path: forums-1-dub!not-for-mail
Xref: forums-1-dub ianywhere.public.general:5980
Article PK: 2438

Just replaced Linux x32 9.0.2.3219 with Linux x64 9.0.2.3397
and got performance problem. After re-creating statistics
it's started to work better but still not enough.

on x32 load average: 4.32, 5.12, 4.56
on x64 load average: 12.81, 11.23, 11.01

Parameters of server identical, chashe size 3gb, cache used
x32: 2.1gb , x64: 2.8gb.

Is it necessary to do any additional actions for x64 3397?
or it's just unsuccessful EBF, cause I've run x64 without
problem but due to migration setup x32.


Mark Culp Posted on 2007-05-08 11:48:17.0Z
Message-ID: <46406380.F36218E7@iAnywhere.com>
From: Mark Culp <reply_to_newsgroups_only_please_nospam_mark.culp@iAnywhere.com>
X-Mailer: Mozilla 4.75 [en] (Windows NT 5.0; U)
X-Accept-Language: en
MIME-Version: 1.0
Newsgroups: ianywhere.public.general
Subject: Re: Performance problem (Linux newsgroup is not available again and again and again...)
References: <46405f6c.5970.1681692777@sybase.com>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
NNTP-Posting-Host: cpe0004e23b1c57-cm00111a59be64.cpe.net.cable.rogers.com
X-Original-NNTP-Posting-Host: cpe0004e23b1c57-cm00111a59be64.cpe.net.cable.rogers.com
Date: 8 May 2007 04:48:17 -0700
X-Trace: forums-1-dub 1178624897 72.143.143.187 (8 May 2007 04:48:17 -0700)
X-Original-Trace: 8 May 2007 04:48:17 -0700, cpe0004e23b1c57-cm00111a59be64.cpe.net.cable.rogers.com
Lines: 31
Path: forums-1-dub!not-for-mail
Xref: forums-1-dub ianywhere.public.general:5982
Article PK: 2440

Some questions:

What OS (distro) are you using?
What are your computer configurations?
- i.e. processor speed of the x86 and x64 computers, disk systems, etc.

What query are you running?

Did you compare the plans between the x32 and x64 runs? (hint: you should :-)
--
Mark Culp
SQLAnywhere Research and Development
-------------------------------------------------------------------------
** Whitepapers, TechDocs, bug fixes are all available through the **
** iAnywhere Developer Community at http://www.ianywhere.com/developer **
-------------------------------------------------------------------------

Alex wrote:
>
> Just replaced Linux x32 9.0.2.3219 with Linux x64 9.0.2.3397
> and got performance problem. After re-creating statistics
> it's started to work better but still not enough.
>
> on x32 load average: 4.32, 5.12, 4.56
> on x64 load average: 12.81, 11.23, 11.01
>
> Parameters of server identical, chashe size 3gb, cache used
> x32: 2.1gb , x64: 2.8gb.
>
> Is it necessary to do any additional actions for x64 3397?
> or it's just unsuccessful EBF, cause I've run x64 without
> problem but due to migration setup x32.


Alex Posted on 2007-05-08 12:22:46.0Z
Sender: 59db.46406254.1804289383@sybase.com
From: Alex
Newsgroups: ianywhere.public.general
Subject: Re: Performance problem (Linux newsgroup is not available again and again and again...)
X-Mailer: WebNews to Mail Gateway v1.1t
Message-ID: <46406b96.5b0e.1681692777@sybase.com>
References: <46406380.F36218E7@iAnywhere.com>
NNTP-Posting-Host: 10.22.241.41
X-Original-NNTP-Posting-Host: 10.22.241.41
Date: 8 May 2007 05:22:46 -0700
X-Trace: forums-1-dub 1178626966 10.22.241.41 (8 May 2007 05:22:46 -0700)
X-Original-Trace: 8 May 2007 05:22:46 -0700, 10.22.241.41
Lines: 49
Path: forums-1-dub!not-for-mail
Xref: forums-1-dub ianywhere.public.general:5983
Article PK: 2445


> Some questions:
>
> What OS (distro) are you using?
rhel4
> What are your computer configurations?
4gb, 2cpu
> - i.e. processor speed of the x86 and x64 computers, disk
> systems, etc.

2.4mgz xeon, scsi serveraid

>
> What query are you running?
it's not a single particular query. It's working system.
Thus, it's list of queries.

>
> Did you compare the plans between the x32 and x64 runs?
> (hint: you should :-) --
Some of them yes, no diff.


But why should it be ? It's the same hardware, same db
structure, same page size, same indexes. Or x32 team is
better
then x64 team? ;)


> Mark Culp
> SQLAnywhere Research and Development
> ----------------------------------------------------------
> --------------- ** Whitepapers, TechDocs, bug fixes are
> all available through the ** ** iAnywhere Developer
> Community at http://www.ianywhere.com/developer **
> ----------------------------------------------------------
> --------------- Alex wrote:
> >
> > Just replaced Linux x32 9.0.2.3219 with Linux x64
> > 9.0.2.3397 and got performance problem. After
> > re-creating statistics it's started to work better but
> > still not enough.
> > on x32 load average: 4.32, 5.12, 4.56
> > on x64 load average: 12.81, 11.23, 11.01
> >
> > Parameters of server identical, chashe size 3gb, cache
> > used x32: 2.1gb , x64: 2.8gb.
> >
> > Is it necessary to do any additional actions for x64
> > 3397? or it's just unsuccessful EBF, cause I've run x64
> > without problem but due to migration setup x32.


Mark Culp Posted on 2007-05-08 13:16:51.0Z
Message-ID: <46407842.86DA9270@iAnywhere.com>
From: Mark Culp <reply_to_newsgroups_only_please_nospam_mark.culp@iAnywhere.com>
X-Mailer: Mozilla 4.75 [en] (Windows NT 5.0; U)
X-Accept-Language: en
MIME-Version: 1.0
Newsgroups: ianywhere.public.general
Subject: Re: Performance problem (Linux newsgroup is not available again and again and again...)
References: <46406380.F36218E7@iAnywhere.com> <46406b96.5b0e.1681692777@sybase.com>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
NNTP-Posting-Host: mculp-pc.sybase.com
X-Original-NNTP-Posting-Host: mculp-pc.sybase.com
Date: 8 May 2007 06:16:51 -0700
X-Trace: forums-1-dub 1178630211 10.25.107.117 (8 May 2007 06:16:51 -0700)
X-Original-Trace: 8 May 2007 06:16:51 -0700, mculp-pc.sybase.com
Lines: 84
Path: forums-1-dub!not-for-mail
Xref: forums-1-dub ianywhere.public.general:5984
Article PK: 2442

Please note that the x86 and x64 Linux releases, as are all of
the SQLAnywhere release, all come from exactly the same codeline,
so the code is exactly the same. Therefore you would not expect
any major difference between one platform and another.

Note however, there can be some very subtle differences in the
environment that can show up in getting drastically different
behaviour. One thing to consider is that the word size of the
executable can have an effect on the amount of memory used to
hold data structures in memory (i.e. 64bit pointers vs 32bit
pointers), and hence the amount of cache used (for example)
can change.

So if you are using the same cache size for both the x64 and x86
engines, the x86/32bit engine will in fact use less amount of the
cache for holding meta data and other structures than the x64/64bit
engine, and therefore more cache will, on average, be available to
hold db pages in the cache in the 32 bit process. Depending on
your exact circumstances (db size, page size, table sizes and
distribution, and the queries that you run, workload, etc), the
optimizer in the two engines could be choosing different plans
and/or taking different amounts of time to execute, ... even though
the two engines are compiled from exactly the same code.

In order to understand exactly why you are seeing a difference, you
need to look at a number of factors, some of which I have already
mentioned, but I would focus on analyzing your workload.
One way to do this would be to get a request-level-log for each
of the two engines and look for differences in how long equivalent
queries execute.

Others reading this thread may have other suggestions?

- Mark

Alex wrote:
>
> > Some questions:
> >
> > What OS (distro) are you using?
> rhel4
> > What are your computer configurations?
> 4gb, 2cpu
> > - i.e. processor speed of the x86 and x64 computers, disk
> > systems, etc.
> 2.4mgz xeon, scsi serveraid
>
> >
> > What query are you running?
> it's not a single particular query. It's working system.
> Thus, it's list of queries.
>
> >
> > Did you compare the plans between the x32 and x64 runs?
> > (hint: you should :-) --
> Some of them yes, no diff.
>
> But why should it be ? It's the same hardware, same db
> structure, same page size, same indexes. Or x32 team is
> better
> then x64 team? ;)
>
> > Mark Culp
> > SQLAnywhere Research and Development
> > ----------------------------------------------------------
> > --------------- ** Whitepapers, TechDocs, bug fixes are
> > all available through the ** ** iAnywhere Developer
> > Community at http://www.ianywhere.com/developer **
> > ----------------------------------------------------------
> > --------------- Alex wrote:
> > >
> > > Just replaced Linux x32 9.0.2.3219 with Linux x64
> > > 9.0.2.3397 and got performance problem. After
> > > re-creating statistics it's started to work better but
> > > still not enough.
> > > on x32 load average: 4.32, 5.12, 4.56
> > > on x64 load average: 12.81, 11.23, 11.01
> > >
> > > Parameters of server identical, chashe size 3gb, cache
> > > used x32: 2.1gb , x64: 2.8gb.
> > >
> > > Is it necessary to do any additional actions for x64
> > > 3397? or it's just unsuccessful EBF, cause I've run x64
> > > without problem but due to migration setup x32.


Alex Posted on 2007-05-08 13:43:45.0Z
Sender: 5caa.464078c8.1804289383@sybase.com
From: Alex
Newsgroups: ianywhere.public.general
Subject: Re: Performance problem (Linux newsgroup is not available again and again and again...)
X-Mailer: WebNews to Mail Gateway v1.1t
Message-ID: <46407e91.5d59.1681692777@sybase.com>
References: <46407842.86DA9270@iAnywhere.com>
NNTP-Posting-Host: 10.22.241.41
X-Original-NNTP-Posting-Host: 10.22.241.41
Date: 8 May 2007 06:43:45 -0700
X-Trace: forums-1-dub 1178631825 10.22.241.41 (8 May 2007 06:43:45 -0700)
X-Original-Trace: 8 May 2007 06:43:45 -0700, 10.22.241.41
Lines: 110
Path: forums-1-dub!not-for-mail
Xref: forums-1-dub ianywhere.public.general:5986
Article PK: 2444


> Note however, there can be some very subtle differences in
> the environment that can show up in getting drastically
> different behaviour. One thing to consider is that the
> word size of the executable can have an effect on the
> amount of memory used to hold data structures in memory
> (i.e. 64bit pointers vs 32bit pointers), and hence the
> amount of cache used (for example) can change.

agree, it's the reason why i reported a used cache size.
2.1gb vs 2.8gb. but i guess cache size is not reason,
cause
both engine didn't use all allocate cache 3gb.

>
> So if you are using the same cache size for both the x64
> and x86 engines, the x86/32bit engine will in fact use
> less amount of the cache for holding meta data and other
> structures than the x64/64bit engine, and therefore more
> cache will, on average, be available to hold db pages in
> the cache in the 32 bit process. Depending on your exact
> circumstances (db size, page size, table sizes and
> distribution, and the queries that you run, workload, etc)
> , the optimizer in the two engines could be choosing
> different plans and/or taking different amounts of time to
> execute, ... even though the two engines are compiled from
> exactly the same code.

yeap, but if it's no cache what is reason for it?


>
> In order to understand exactly why you are seeing a
> difference, you need to look at a number of factors, some
> of which I have already mentioned, but I would focus on
> analyzing your workload. One way to do this would be to
> get a request-level-log for each of the two engines and
> look for differences in how long equivalent queries
> execute.

it's huge work. there are a lot of queries. i've took few
most expensive, plan was the same.

current x64 has tendency to downgrade performance with
increasing cpu utilization and almost no activity on disk.

my point is, as far as no any valuable advice ,x64 build
3397
shouldn't be used until reason of such behaviour is not
found.


and pls fix linux newsgroup as well :)

>
> Others reading this thread may have other suggestions?
>
> - Mark
>
> Alex wrote:
> >
> > > Some questions:
> > >
> > > What OS (distro) are you using?
> > rhel4
> > > What are your computer configurations?
> > 4gb, 2cpu
> > > - i.e. processor speed of the x86 and x64 computers,
> > > disk systems, etc.
> > 2.4mgz xeon, scsi serveraid
> >
> > >
> > > What query are you running?
> > it's not a single particular query. It's working
> > system. Thus, it's list of queries.
> >
> > >
> > > Did you compare the plans between the x32 and x64
> > > runs? (hint: you should :-) --
> > Some of them yes, no diff.
> >
> > But why should it be ? It's the same hardware, same db
> > structure, same page size, same indexes. Or x32 team
> > is better
> > then x64 team? ;)
> >
> > > Mark Culp
> > > SQLAnywhere Research and Development
> > >
> > >
> ----------------------------------------------------------
> > > --------------- ** Whitepapers, TechDocs, bug fixes
> are all available through the ** ** iAnywhere
> > > Developer Community at
> > > http://www.ianywhere.com/developer **
> > >
> ----------------------------------------------------------
> > > --------------- Alex wrote: >
> > > > Just replaced Linux x32 9.0.2.3219 with Linux x64
> > > > 9.0.2.3397 and got performance problem. After
> > > > re-creating statistics it's started to work better
> > > > but still not enough.
> > > > on x32 load average: 4.32, 5.12, 4.56
> > > > on x64 load average: 12.81, 11.23, 11.01
> > > >
> > > > Parameters of server identical, chashe size 3gb,
> > > > cache used x32: 2.1gb , x64: 2.8gb.
> > > >
> > > > Is it necessary to do any additional actions for x64
> > > > 3397? or it's just unsuccessful EBF, cause I've run
> > > > x64 without problem but due to migration setup x32.


Rob Waywell Posted on 2007-05-09 14:29:19.0Z
From: "Rob Waywell" <rwaywell_no_spam_please@ianywhere.com>
Newsgroups: ianywhere.public.general
References: <46407842.86DA9270@iAnywhere.com> <46407e91.5d59.1681692777@sybase.com>
Subject: Re: Performance problem (Linux newsgroup is not available again and again and again...)
Lines: 136
X-Priority: 3
X-MSMail-Priority: Normal
X-Newsreader: Microsoft Outlook Express 6.00.2900.3028
X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2900.3028
X-RFC2646: Format=Flowed; Original
NNTP-Posting-Host: rwaywell-xp2.sybase.com
X-Original-NNTP-Posting-Host: rwaywell-xp2.sybase.com
Message-ID: <4641dabf$1@forums-1-dub>
Date: 9 May 2007 07:29:19 -0700
X-Trace: forums-1-dub 1178720959 10.25.98.235 (9 May 2007 07:29:19 -0700)
X-Original-Trace: 9 May 2007 07:29:19 -0700, rwaywell-xp2.sybase.com
X-Authenticated-User: techsupp
Path: forums-1-dub!not-for-mail
Xref: forums-1-dub ianywhere.public.general:5989
Article PK: 2452

Sounds like you have a repro to submit to Tech Support showing the
performance degradation on specific queries when moving from an X86 to X64
version of Linux.

--
-----------------------------------------------
Robert Waywell
Sybase Adaptive Server Anywhere Developer - Version 8
Sybase Certified Professional

Sybase's iAnywhere Solutions

Please respond ONLY to newsgroup

EBF's and Patches: http://downloads.sybase.com
choose SQL Anywhere Studio >> change 'time frame' to all

To Submit Bug Reports:
http://case-express.sybase.com/cx/cx.stm?starturl=casemessage.ssc?CASETYPE=Bug

SQL Anywhere Studio Supported Platforms and Support Status
http://my.sybase.com/detail?id=1002288

<Alex> wrote in message news:46407e91.5d59.1681692777@sybase.com...
>> Note however, there can be some very subtle differences in
>> the environment that can show up in getting drastically
>> different behaviour. One thing to consider is that the
>> word size of the executable can have an effect on the
>> amount of memory used to hold data structures in memory
>> (i.e. 64bit pointers vs 32bit pointers), and hence the
>> amount of cache used (for example) can change.
>
> agree, it's the reason why i reported a used cache size.
> 2.1gb vs 2.8gb. but i guess cache size is not reason,
> cause
> both engine didn't use all allocate cache 3gb.
>
>>
>> So if you are using the same cache size for both the x64
>> and x86 engines, the x86/32bit engine will in fact use
>> less amount of the cache for holding meta data and other
>> structures than the x64/64bit engine, and therefore more
>> cache will, on average, be available to hold db pages in
>> the cache in the 32 bit process. Depending on your exact
>> circumstances (db size, page size, table sizes and
>> distribution, and the queries that you run, workload, etc)
>> , the optimizer in the two engines could be choosing
>> different plans and/or taking different amounts of time to
>> execute, ... even though the two engines are compiled from
>> exactly the same code.
>
> yeap, but if it's no cache what is reason for it?
>
>
>>
>> In order to understand exactly why you are seeing a
>> difference, you need to look at a number of factors, some
>> of which I have already mentioned, but I would focus on
>> analyzing your workload. One way to do this would be to
>> get a request-level-log for each of the two engines and
>> look for differences in how long equivalent queries
>> execute.
>
> it's huge work. there are a lot of queries. i've took few
> most expensive, plan was the same.
>
> current x64 has tendency to downgrade performance with
> increasing cpu utilization and almost no activity on disk.
>
> my point is, as far as no any valuable advice ,x64 build
> 3397
> shouldn't be used until reason of such behaviour is not
> found.
>
>
> and pls fix linux newsgroup as well :)
>
>>
>> Others reading this thread may have other suggestions?
>>
>> - Mark
>>
>> Alex wrote:
>> >
>> > > Some questions:
>> > >
>> > > What OS (distro) are you using?
>> > rhel4
>> > > What are your computer configurations?
>> > 4gb, 2cpu
>> > > - i.e. processor speed of the x86 and x64 computers,
>> > > disk systems, etc.
>> > 2.4mgz xeon, scsi serveraid
>> >
>> > >
>> > > What query are you running?
>> > it's not a single particular query. It's working
>> > system. Thus, it's list of queries.
>> >
>> > >
>> > > Did you compare the plans between the x32 and x64
>> > > runs? (hint: you should :-) --
>> > Some of them yes, no diff.
>> >
>> > But why should it be ? It's the same hardware, same db
>> > structure, same page size, same indexes. Or x32 team
>> > is better
>> > then x64 team? ;)
>> >
>> > > Mark Culp
>> > > SQLAnywhere Research and Development
>> > >
>> > >
>> ----------------------------------------------------------
>> > > --------------- ** Whitepapers, TechDocs, bug fixes
>> are all available through the ** ** iAnywhere
>> > > Developer Community at
>> > > http://www.ianywhere.com/developer **
>> > >
>> ----------------------------------------------------------
>> > > --------------- Alex wrote: >
>> > > > Just replaced Linux x32 9.0.2.3219 with Linux x64
>> > > > 9.0.2.3397 and got performance problem. After
>> > > > re-creating statistics it's started to work better
>> > > > but still not enough.
>> > > > on x32 load average: 4.32, 5.12, 4.56
>> > > > on x64 load average: 12.81, 11.23, 11.01
>> > > >
>> > > > Parameters of server identical, chashe size 3gb,
>> > > > cache used x32: 2.1gb , x64: 2.8gb.
>> > > >
>> > > > Is it necessary to do any additional actions for x64
>> > > > 3397? or it's just unsuccessful EBF, cause I've run
>> > > > x64 without problem but due to migration setup x32.


Alex Posted on 2007-05-08 11:47:32.0Z
Sender: 59db.46406254.1804289383@sybase.com
From: Alex
Newsgroups: ianywhere.public.general
Subject: Re: Performance problem (Linux newsgroup is not available again and again and again...)
X-Mailer: WebNews to Mail Gateway v1.1t
Message-ID: <46406353.5a01.1681692777@sybase.com>
References: <46405f6c.5970.1681692777@sybase.com>
NNTP-Posting-Host: 10.22.241.41
X-Original-NNTP-Posting-Host: 10.22.241.41
Date: 8 May 2007 04:47:32 -0700
X-Trace: forums-1-dub 1178624852 10.22.241.41 (8 May 2007 04:47:32 -0700)
X-Original-Trace: 8 May 2007 04:47:32 -0700, 10.22.241.41
Lines: 3
Path: forums-1-dub!not-for-mail
Xref: forums-1-dub ianywhere.public.general:5981
Article PK: 2441

Of course, work load is approximately same. Avg load
distribution had been measured during 5 hours. Thus, it's
Avg(Avg load).