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.

multi versioning concurrency (MVC)

24 posts in Product Futures Discussion Last posting was on 2002-10-13 23:26:36.0Z
Elton Posted on 2002-06-03 02:12:08.0Z
From: "Elton" <elton@asia.bigfoot.com>
Subject: multi versioning concurrency (MVC)
Date: Mon, 3 Jun 2002 10:12:08 +0800
Lines: 10
X-Priority: 3
X-MSMail-Priority: Normal
X-Newsreader: Microsoft Outlook Express 6.00.2600.0000
X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2600.0000
Message-ID: <2axvuOqCCHA.255@forums.sybase.com>
Newsgroups: sybase.public.ase.product_futures_discussion
NNTP-Posting-Host: 203.177.73.45
Path: forums-1-dub!forums-master.sybase.com!forums.sybase.com
Xref: forums-1-dub sybase.public.ase.product_futures_discussion:472
Article PK: 93642

hello all,

does the engineering team has plans to implement MVC model for future ASE,
MVC is a good model for a true concurrent connections issues. I'm a
supporter for it.


Elton


Sethu M Posted on 2002-07-17 05:43:56.0Z
From: "Sethu M" <sethu@sybase.com>
References: <2axvuOqCCHA.255@forums.sybase.com>
Subject: Re: multi versioning concurrency (MVC)
Date: Tue, 16 Jul 2002 22:43:56 -0700
Lines: 7
MIME-Version: 1.0
Content-Type: text/plain; charset="Windows-1252"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-MSMail-Priority: Normal
X-Newsreader: Microsoft Outlook Express 6.00.2600.0000
X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2600.0000
Message-ID: <58ZKTaVLCHA.1004@forums.sybase.com>
Newsgroups: sybase.public.ase.product_futures_discussion
NNTP-Posting-Host: 10.22.120.60
Path: forums-1-dub!forums-master.sybase.com!forums.sybase.com
Xref: forums-1-dub sybase.public.ase.product_futures_discussion:389
Article PK: 93558

Does anyone know whether DB2 or Microsoft is planning to implement
this ?

I'm investigating the pros/cons of MVC in ASE

Sethu


Eugene Korolkov Posted on 2002-07-18 19:39:32.0Z
Message-ID: <3D371974.FDA1D0CD@dgny.com>
Date: Thu, 18 Jul 2002 15:39:32 -0400
From: Eugene Korolkov <ekorolkov@dgny.com>
Organization: The Davidsohn Group
X-Mailer: Mozilla 4.78 [en] (WinNT; U)
X-Accept-Language: ru,en
MIME-Version: 1.0
To: Sethu M <sethu@sybase.com>
Subject: Re: multi versioning concurrency (MVC)
References: <2axvuOqCCHA.255@forums.sybase.com> <58ZKTaVLCHA.1004@forums.sybase.com>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Newsgroups: sybase.public.ase.product_futures_discussion
Lines: 29
NNTP-Posting-Host: 12.3.91.242
Path: forums-1-dub!forums-master.sybase.com!forums.sybase.com
Xref: forums-1-dub sybase.public.ase.product_futures_discussion:382
Article PK: 93552

Sethu,

I think there is no pros/cons. Without that feature all your results
from simple select stmt in multiuser env. might be unreliable
and be dependable from access paths, indexes, CPU speed,
other users activity, query optimizer decisions and so on ,
shortly "situation on the road" or artificially consistent using
blocking other users.

With the absence of the snapshot concept (point of the time) you even
could end-up with result rows never being presented in the table at the
same moment. It seems to me that the whole idea of the implementing
isolation levels using locks is not a good idea and probably came to
Sybase
from OS theory where locks in general are keeping relatively short time

comparing to the possible query time in RDBMS.

DB2 and MSQL apparently will have this sooner or later, at least article

which I mention early has been written by Microsoft employees
(at that time)

Regards,
Eugene

Sethu M wrote:

> Does anyone know whether DB2 or Microsoft is planning to implement
> this ?
>
> I'm investigating the pros/cons of MVC in ASE
>
> Sethu


Sethu M Posted on 2002-07-19 06:03:19.0Z
From: "Sethu M" <sethu@sybase.com>
References: <2axvuOqCCHA.255@forums.sybase.com> <58ZKTaVLCHA.1004@forums.sybase.com> <3D371974.FDA1D0CD@dgny.com>
Subject: Re: multi versioning concurrency (MVC)
Date: Thu, 18 Jul 2002 23:03:19 -0700
Lines: 79
MIME-Version: 1.0
Content-Type: text/plain; charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Newsreader: Microsoft Outlook Express 6.00.2600.0000
X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2600.0000
Message-ID: <RicdguuLCHA.654@forums.sybase.com>
Newsgroups: sybase.public.ase.product_futures_discussion
NNTP-Posting-Host: 10.22.120.67
Path: forums-1-dub!forums-master.sybase.com!forums.sybase.com
Xref: forums-1-dub sybase.public.ase.product_futures_discussion:381
Article PK: 93551

Hi Eugene,

Isolation levels using locking technique have been there for 20+ years.
This is very useful for OLTP applications with several techniques
for locking mechanisms.
All the rdbms vendors that was there in 1980s did implement
isolation levels using the seminal papers published by
IBM system R group (notably Dr. Jim Gray).

I donot understand your statement about "absence of snapshot...".
The semantics of snapshot is to provide all the rows
at that point in time. One could theoretically achieve snapshot kind of
semantics with lock table exclusively and select data out
into a #tmp table. This comes at the expense of concurrency.
One could also use replication technology or hardware mirroring
to get a snapshot of the data.

( Don't get me wrong that I'm giving these reasons because Sybase is not
interested in doing the MVC. I'm just arguing at a technical level).

In terms of resource usage, both the built-in snapshot isolation model
and using replication or hardware mirroring model need same kind
of resources (disk, memory to keep copies of the snapshot).

It is not that MVC is next to sliced bread or some such thing.
I agree that snapshot isolation do resolve some anomalies in
existing isolation semantics and it is useful for certain types
of applications.

The paper you are referring to was written in 1995. We are also
investigating/researching the possibilities of providing snapshot isolation.

Thanks,
Sethu






"Eugene Korolkov" <ekorolkov@dgny.com> wrote in message
news:3D371974.FDA1D0CD@dgny.com...
Sethu,

I think there is no pros/cons. Without that feature all your results
from simple select stmt in multiuser env. might be unreliable
and be dependable from access paths, indexes, CPU speed,
other users activity, query optimizer decisions and so on ,
shortly "situation on the road" or artificially consistent using
blocking other users.

With the absence of the snapshot concept (point of the time) you even
could end-up with result rows never being presented in the table at the
same moment. It seems to me that the whole idea of the implementing
isolation levels using locks is not a good idea and probably came to
Sybase
from OS theory where locks in general are keeping relatively short time

comparing to the possible query time in RDBMS.

DB2 and MSQL apparently will have this sooner or later, at least article

which I mention early has been written by Microsoft employees
(at that time)

Regards,
Eugene

Sethu M wrote:

> Does anyone know whether DB2 or Microsoft is planning to implement
> this ?
>
> I'm investigating the pros/cons of MVC in ASE
>
> Sethu


Eugene Korolkov Posted on 2002-10-10 19:33:50.0Z
Message-ID: <3DA5D61E.87CE06BE@dgny.com>
Date: Thu, 10 Oct 2002 15:33:50 -0400
From: Eugene Korolkov <ekorolkov@dgny.com>
Organization: The Davidsohn Group
X-Mailer: Mozilla 4.77 [en] (WinNT; U)
X-Accept-Language: ru,en
MIME-Version: 1.0
To: Sethu M <sethu@sybase.com>
Subject: Re: multi versioning concurrency (MVC)
References: <2axvuOqCCHA.255@forums.sybase.com> <58ZKTaVLCHA.1004@forums.sybase.com> <3D371974.FDA1D0CD@dgny.com> <RicdguuLCHA.654@forums.sybase.com>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Newsgroups: sybase.public.ase.product_futures_discussion
Lines: 86
NNTP-Posting-Host: 12.3.91.168
Path: forums-1-dub!forums-master.sybase.com!forums.sybase.com
Xref: forums-1-dub sybase.public.ase.product_futures_discussion:276
Article PK: 93448

Sethu,

As you can see one of the author of that article is Dr. Jim Gray
Looks like he confessed that idea of implementing isolation levels
using locks was wrong.

Theoretically you can lock table exclusively, but
practically you cannot for different reasons including
even technically impossible in Sybase to lock table
exclusively when somebody constantly selecting/updating
that table (like table is queue in application actually).
For consistency in many tables for more or less complex query
it is impossible even "theoretically" I think.

Regards,
Eugene

Sethu M wrote:

> Hi Eugene,
>
> Isolation levels using locking technique have been there for 20+ years.
> This is very useful for OLTP applications with several techniques
> for locking mechanisms.
> All the rdbms vendors that was there in 1980s did implement
> isolation levels using the seminal papers published by
> IBM system R group (notably Dr. Jim Gray).
>
> I donot understand your statement about "absence of snapshot...".
> The semantics of snapshot is to provide all the rows
> at that point in time. One could theoretically achieve snapshot kind of
> semantics with lock table exclusively and select data out
> into a #tmp table. This comes at the expense of concurrency.
> One could also use replication technology or hardware mirroring
> to get a snapshot of the data.
>
> ( Don't get me wrong that I'm giving these reasons because Sybase is not
> interested in doing the MVC. I'm just arguing at a technical level).
>
> In terms of resource usage, both the built-in snapshot isolation model
> and using replication or hardware mirroring model need same kind
> of resources (disk, memory to keep copies of the snapshot).
>
> It is not that MVC is next to sliced bread or some such thing.
> I agree that snapshot isolation do resolve some anomalies in
> existing isolation semantics and it is useful for certain types
> of applications.
>
> The paper you are referring to was written in 1995. We are also
> investigating/researching the possibilities of providing snapshot isolation.
>
> Thanks,
> Sethu
>
> "Eugene Korolkov" <ekorolkov@dgny.com> wrote in message
> news:3D371974.FDA1D0CD@dgny.com...
> Sethu,
>
> I think there is no pros/cons. Without that feature all your results
> from simple select stmt in multiuser env. might be unreliable
> and be dependable from access paths, indexes, CPU speed,
> other users activity, query optimizer decisions and so on ,
> shortly "situation on the road" or artificially consistent using
> blocking other users.
>
> With the absence of the snapshot concept (point of the time) you even
> could end-up with result rows never being presented in the table at the
> same moment. It seems to me that the whole idea of the implementing
> isolation levels using locks is not a good idea and probably came to
> Sybase
> from OS theory where locks in general are keeping relatively short time
>
> comparing to the possible query time in RDBMS.
>
> DB2 and MSQL apparently will have this sooner or later, at least article
>
> which I mention early has been written by Microsoft employees
> (at that time)
>
> Regards,
> Eugene
>
> Sethu M wrote:
>
> > Does anyone know whether DB2 or Microsoft is planning to implement
> > this ?
> >
> > I'm investigating the pros/cons of MVC in ASE
> >
> > Sethu


Sethu M Posted on 2002-10-13 23:26:36.0Z
From: "Sethu M" <sethu@sybase.com>
References: <2axvuOqCCHA.255@forums.sybase.com> <58ZKTaVLCHA.1004@forums.sybase.com> <3D371974.FDA1D0CD@dgny.com> <RicdguuLCHA.654@forums.sybase.com> <3DA5D61E.87CE06BE@dgny.com>
Subject: Re: multi versioning concurrency (MVC)
Date: Sun, 13 Oct 2002 16:26:36 -0700
Lines: 107
MIME-Version: 1.0
Content-Type: text/plain; charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Newsreader: Microsoft Outlook Express 6.00.2600.0000
X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2600.0000
Message-ID: <hH3VmfzcCHA.297@forums.sybase.com>
Newsgroups: sybase.public.ase.product_futures_discussion
NNTP-Posting-Host: 10.22.120.32
Path: forums-1-dub!forums-master.sybase.com!forums.sybase.com
Xref: forums-1-dub sybase.public.ase.product_futures_discussion:274
Article PK: 93446

Hi Eugene,

don't want to start an argument here.

But locking table exclusively is possible in a busy system
with update/delete because, ASE has a fair-scheduling
algorithm for the lock manager. The requester of table lock
will be placed in the queue and will definitely be granted
the table lock when that request reaches the head of the queue.

Sethu

"Eugene Korolkov" <ekorolkov@dgny.com> wrote in message
news:3DA5D61E.87CE06BE@dgny.com...
Sethu,

As you can see one of the author of that article is Dr. Jim Gray
Looks like he confessed that idea of implementing isolation levels
using locks was wrong.

Theoretically you can lock table exclusively, but
practically you cannot for different reasons including
even technically impossible in Sybase to lock table
exclusively when somebody constantly selecting/updating
that table (like table is queue in application actually).
For consistency in many tables for more or less complex query
it is impossible even "theoretically" I think.

Regards,
Eugene

Sethu M wrote:

> Hi Eugene,
>
> Isolation levels using locking technique have been there for 20+ years.
> This is very useful for OLTP applications with several techniques
> for locking mechanisms.
> All the rdbms vendors that was there in 1980s did implement
> isolation levels using the seminal papers published by
> IBM system R group (notably Dr. Jim Gray).
>
> I donot understand your statement about "absence of snapshot...".
> The semantics of snapshot is to provide all the rows
> at that point in time. One could theoretically achieve snapshot kind of
> semantics with lock table exclusively and select data out
> into a #tmp table. This comes at the expense of concurrency.
> One could also use replication technology or hardware mirroring
> to get a snapshot of the data.
>
> ( Don't get me wrong that I'm giving these reasons because Sybase is not
> interested in doing the MVC. I'm just arguing at a technical level).
>
> In terms of resource usage, both the built-in snapshot isolation model
> and using replication or hardware mirroring model need same kind
> of resources (disk, memory to keep copies of the snapshot).
>
> It is not that MVC is next to sliced bread or some such thing.
> I agree that snapshot isolation do resolve some anomalies in
> existing isolation semantics and it is useful for certain types
> of applications.
>
> The paper you are referring to was written in 1995. We are also
> investigating/researching the possibilities of providing snapshot
isolation.
>
> Thanks,
> Sethu
>
> "Eugene Korolkov" <ekorolkov@dgny.com> wrote in message
> news:3D371974.FDA1D0CD@dgny.com...
> Sethu,
>
> I think there is no pros/cons. Without that feature all your results
> from simple select stmt in multiuser env. might be unreliable
> and be dependable from access paths, indexes, CPU speed,
> other users activity, query optimizer decisions and so on ,
> shortly "situation on the road" or artificially consistent using
> blocking other users.
>
> With the absence of the snapshot concept (point of the time) you even
> could end-up with result rows never being presented in the table at the
> same moment. It seems to me that the whole idea of the implementing
> isolation levels using locks is not a good idea and probably came to
> Sybase
> from OS theory where locks in general are keeping relatively short time
>
> comparing to the possible query time in RDBMS.
>
> DB2 and MSQL apparently will have this sooner or later, at least article
>
> which I mention early has been written by Microsoft employees
> (at that time)
>
> Regards,
> Eugene
>
> Sethu M wrote:
>
> > Does anyone know whether DB2 or Microsoft is planning to implement
> > this ?
> >
> > I'm investigating the pros/cons of MVC in ASE
> >
> > Sethu


Pablo Sanchez Posted on 2002-06-04 20:39:43.0Z
From: "Pablo Sanchez" <pablo@dev.null>
References: <2axvuOqCCHA.255@forums.sybase.com>
Subject: Re: multi versioning concurrency (MVC)
Date: Tue, 4 Jun 2002 14:39:43 -0600
Lines: 21
Organization: High-Performance Database Engineering
X-Priority: 3
X-MSMail-Priority: Normal
X-Newsreader: Microsoft Outlook Express 6.00.2600.0000
X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2600.0000
Message-ID: <nuV$glADCHA.81@forums.sybase.com>
Newsgroups: sybase.public.ase.product_futures_discussion
NNTP-Posting-Host: host-1-221.ionsky.com 12.151.58.221
Path: forums-1-dub!forums-master.sybase.com!forums.sybase.com
Xref: forums-1-dub sybase.public.ase.product_futures_discussion:471
Article PK: 93640


"Elton" <elton@asia.bigfoot.com> wrote in message
news:2axvuOqCCHA.255@forums.sybase.com...
> hello all,
>
> does the engineering team has plans to implement MVC model for
future ASE,
> MVC is a good model for a true concurrent connections issues. I'm a
> supporter for it.

Personally, I find that the existing locking scheme works very well.
In order to implement MVC, this is a resource intenstive activity,
which I believe is one of the reasons Oracle eats more resources the
ASE.
--
Pablo Sanchez, High-Performance Database Engineering
mailto:pablo@hpdbe.com
http://www.hpdbe.com
Available for short-term and long-term contracts


Sethu M Posted on 2002-07-21 17:49:47.0Z
From: "Sethu M" <sethu@sybase.com>
References: <2axvuOqCCHA.255@forums.sybase.com> <nuV$glADCHA.81@forums.sybase.com>
Subject: Re: multi versioning concurrency (MVC)
Date: Sun, 21 Jul 2002 10:49:47 -0700
Lines: 131
MIME-Version: 1.0
Content-Type: text/plain; charset="Windows-1252"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-MSMail-Priority: Normal
X-Newsreader: Microsoft Outlook Express 6.00.2600.0000
X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2600.0000
Message-ID: <ia5DqCOMCHA.195@forums.sybase.com>
Newsgroups: sybase.public.ase.product_futures_discussion
NNTP-Posting-Host: 10.22.120.59
Path: forums-1-dub!forums-master.sybase.com!forums.sybase.com
Xref: forums-1-dub sybase.public.ase.product_futures_discussion:380
Article PK: 93548

I'm attaching the email interaction I had with George. Thanks George
for helping me to clarify the statement i made.

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

I meant to say that donot take these reasons as if Sybase is not
interested in dong MVC. As I said later, we are investigating
this feature at this time.

Thanks,
Sethu

----- Original Message -----
From: "gmsayloriii" <gmsayloriii@email.msn.com>
To: "Sethu M" <sethu@sybase.com>
Sent: Sunday, July 21, 2002 10:37 AM
Subject: Re: multi versioning concurrency (MVC)

>
> ( Don't get me wrong that I'm giving these reasons because Sybase is not
> interested in doing the MVC. I'm just arguing at a technical level).

I may have misread, sorry if I did. Either way, it is good to know you guys
are responding to this favorably, I think it is a nice feature.
One question, if this is done, is Sybase considering getting rid of its
current scheme or just adding an optional feature?

thanks for your time.
George





----- Original Message -----
From: "Sethu M" <sethu@sybase.com>
To: "gmsayloriii" <gmsayloriii@email.msn.com>
Cc: "Sethu Meenakshisundara" <sethu@sybase.com>
Sent: Sunday, July 21, 2002 11:35 AM
Subject: Re: multi versioning concurrency (MVC)


> George,
>
> Where did I say that Sybase is not interested in doing MVC ?
>
> Sethu
>
> ----- Original Message -----
> From: "gmsayloriii" <gmsayloriii@email.msn.com>
> To: "Sethu M" <sethu@sybase.com>
> Sent: Sunday, July 21, 2002 6:50 AM
> Subject: Re: multi versioning concurrency (MVC)
>
>
>
>
> >
> > I donot understand your statement about "absence of snapshot...".
> > The semantics of snapshot is to provide all the rows
> > at that point in time.
>
> couldn't one do this with minimal overhead on RLL tables? Why not just
> snapshot the dirty rows in a snapshot segment/cache.
> There are other features such as readpast and reorg that are RLL specific,
> so why not this. With modern hardware, I can live with memory hogs. 64
> bit ASE can take care of scalability.
>
> One could theoretically achieve snapshot kind of
> > semantics with lock table exclusively and select data out
> > into a #tmp table. This comes at the expense of concurrency.
> > One could also use replication technology or hardware mirroring
> > to get a snapshot of the data.
>
> That is a lot of work sometimes, and requires special coding to send read
> queries to the snapshot, write queries to the primary. Honestly one
> couldn't achieve MVC for every transaction this way, only one particular
> image. We do this now, and would *love* to get rid of this.
>
> >
> > ( Don't get me wrong that I'm giving these reasons because Sybase is not
> > interested in doing the MVC. I'm just arguing at a technical level).
>
> ****?*****
> Sybase does MVC in ASIQ, it was one of two reasons we evaluated this
> product(the other being bitmap/join indexes). Ultimately we abandoned the
> project since IQ couldn't handle any OLTP.
>
> If Sybase had a product that supported both regular locking and MVC we'd
buy
> it right away, I think that long term Sybase has to implement this to
please
> many of its customers. It is only my own opinion, but this sounds like
the
> 11.5 decision that no RLL would be implemented (then came 11.9.2 =-))
>
> >
> > In terms of resource usage, both the built-in snapshot isolation model
> > and using replication or hardware mirroring model need same kind
> > of resources (disk, memory to keep copies of the snapshot).
>
> this difference being that it won't work off the shelf.
>
> >
> > It is not that MVC is next to sliced bread or some such thing.
>
> No it isn't, but some folks can't let go of the idea. As a long time
Sybase
> supporter, I find good value in the current transaction control methods.
> Oracle folks will not be so easiliy convinced however, since this
technology
> is integral to the product.
>
>
> We are also
> > investigating/researching the possibilities of providing snapshot
> isolation.
>
> Please clarify, above you've said Sybase isn't interested, this statement
> seems to contradict this
>
>
> George
>
>
>
>


Andrew Schonberger Posted on 2002-06-16 10:30:36.0Z
Message-ID: <3D0C68CB.D0317E40@idx.com.au>
Date: Sun, 16 Jun 2002 20:30:36 +1000
From: Andrew Schonberger <aschon@idx.com.au>
X-Mailer: Mozilla 4.77 [en] (WinNT; U)
X-Accept-Language: en
MIME-Version: 1.0
To: Pablo Sanchez <pablo@dev.null>
Subject: Re: multi versioning concurrency (MVC)
References: <2axvuOqCCHA.255@forums.sybase.com> <nuV$glADCHA.81@forums.sybase.com>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Newsgroups: sybase.public.ase.product_futures_discussion
Lines: 65
NNTP-Posting-Host: maxc97.idx.com.au 203.19.9.97
Path: forums-1-dub!forums-master.sybase.com!forums.sybase.com
Xref: forums-1-dub sybase.public.ase.product_futures_discussion:430
Article PK: 93597

Hi,

I've seen just too many applications with locking problems on Sybase.
I'm sure MVC eats up a lot of resources, but so does every other
construct in computer science above assemler level.

It is no fun watching powerfull servers doing mostly nothing, waiting for
exclusive locks to be lifted. True, those applications were not written
specifically for Sybase. Or if they were, their designers were focusing
on the client-side programming space, with Sybase relegated to the role
of a mere persistent storage.

The prevalent-populist OO paradigms (EJB for example) assume separate
objects cooperating in a transaction. The first object applies some
updates without committing, then the second objects starts its own
validations, and updates, and so on. All this in client code. Few people
realise that Syase will hold an exclusive lock on the updated records,
untill all objects have done their job and the transaction finally
commits. I'm not a supporter of this architecture, but I know it appeals
to both juniors and management, since it provides the illusion of modular
isolation and reuse.

Strange enough: the reason for Oracle gaining more market share is not in
their much-hyped Internet features. Rather, it is the decade old MVC
model, which allows each process to apply updates without hindering
others. The denial-mode attitude coming from Sybase may also have
something to do with the phenomenon...but let's be fare, I remember the
Oracle salesman telling me how much temp tables were bad for my health.
And there was the Sybase moderator back in '97 calling row-level locking
completely unnecessary and resource intensive - his postings are still
readable on these newsgroups.

Time will tell, and we hope to hear from Sybase Inc. as well.
My best regards, and thanks for takig up this issue,

Andrew Schonberger

Pablo Sanchez wrote:

> "Elton" <elton@asia.bigfoot.com> wrote in message
> news:2axvuOqCCHA.255@forums.sybase.com...
> > hello all,
> >
> > does the engineering team has plans to implement MVC model for
> future ASE,
> > MVC is a good model for a true concurrent connections issues. I'm a
> > supporter for it.
>
> Personally, I find that the existing locking scheme works very well.
> In order to implement MVC, this is a resource intenstive activity,
> which I believe is one of the reasons Oracle eats more resources the
> ASE.
> --
> Pablo Sanchez, High-Performance Database Engineering
> mailto:pablo@hpdbe.com
> http://www.hpdbe.com
> Available for short-term and long-term contracts


Eugene Korolkov Posted on 2002-06-17 15:10:25.0Z
Message-ID: <3D0DFBE0.A079F082@dgny.com>
Date: Mon, 17 Jun 2002 11:10:25 -0400
From: Eugene Korolkov <ekorolkov@dgny.com>
Organization: The Davidsohn Group
X-Mailer: Mozilla 4.78 [en] (WinNT; U)
X-Accept-Language: ru,en
MIME-Version: 1.0
To: Andrew Schonberger <aschon@idx.com.au>
Subject: Re: multi versioning concurrency (MVC)
References: <2axvuOqCCHA.255@forums.sybase.com> <nuV$glADCHA.81@forums.sybase.com> <3D0C68CB.D0317E40@idx.com.au>
Content-Type: text/plain; charset=us-ascii
Content-Transfer-Encoding: 7bit
Newsgroups: sybase.public.ase.product_futures_discussion
Lines: 80
NNTP-Posting-Host: 12.3.91.151
Path: forums-1-dub!forums-master.sybase.com!forums.sybase.com
Xref: forums-1-dub sybase.public.ase.product_futures_discussion:428
Article PK: 93596

On February 8, 2002 I posted the same question in this newsgroup
and even pointed to the very good article with ANSI Isolation level model
detailed critique and got no reply (!). I completely agree that Oracle
has current market share basically because of that multiversioning model
(which I think is the only true at all, maybe with some modifications)
everything else in Sybase is more or less the same/compatible.

Regards,
Eugene

Andrew Schonberger wrote:

> Hi,
>
> I've seen just too many applications with locking problems on Sybase.
> I'm sure MVC eats up a lot of resources, but so does every other
> construct in computer science above assemler level.
>
> It is no fun watching powerfull servers doing mostly nothing, waiting for
> exclusive locks to be lifted. True, those applications were not written
> specifically for Sybase. Or if they were, their designers were focusing
> on the client-side programming space, with Sybase relegated to the role
> of a mere persistent storage.
>
> The prevalent-populist OO paradigms (EJB for example) assume separate
> objects cooperating in a transaction. The first object applies some
> updates without committing, then the second objects starts its own
> validations, and updates, and so on. All this in client code. Few people
> realise that Syase will hold an exclusive lock on the updated records,
> untill all objects have done their job and the transaction finally
> commits. I'm not a supporter of this architecture, but I know it appeals
> to both juniors and management, since it provides the illusion of modular
> isolation and reuse.
>
> Strange enough: the reason for Oracle gaining more market share is not in
> their much-hyped Internet features. Rather, it is the decade old MVC
> model, which allows each process to apply updates without hindering
> others. The denial-mode attitude coming from Sybase may also have
> something to do with the phenomenon...but let's be fare, I remember the
> Oracle salesman telling me how much temp tables were bad for my health.
> And there was the Sybase moderator back in '97 calling row-level locking
> completely unnecessary and resource intensive - his postings are still
> readable on these newsgroups.
>
> Time will tell, and we hope to hear from Sybase Inc. as well.
> My best regards, and thanks for takig up this issue,
>
> Andrew Schonberger
>
> Pablo Sanchez wrote:
>
> > "Elton" <elton@asia.bigfoot.com> wrote in message
> > news:2axvuOqCCHA.255@forums.sybase.com...
> > > hello all,
> > >
> > > does the engineering team has plans to implement MVC model for
> > future ASE,
> > > MVC is a good model for a true concurrent connections issues. I'm a
> > > supporter for it.
> >
> > Personally, I find that the existing locking scheme works very well.
> > In order to implement MVC, this is a resource intenstive activity,
> > which I believe is one of the reasons Oracle eats more resources the
> > ASE.
> > --
> > Pablo Sanchez, High-Performance Database Engineering
> > mailto:pablo@hpdbe.com
> > http://www.hpdbe.com
> > Available for short-term and long-term contracts


andrew.schonberger Posted on 2002-06-18 07:47:51.0Z
From: andrew.schonberger@macquarie.com-nospam
Date: Tue, 18 Jun 2002 03:47:51 -0400
Newsgroups: sybase.public.ase.product_futures_discussion
Subject: Re: multi versioning concurrency (MVC)
Message-ID: <B31D92578E292847002AD4FD85256BDC.0059C6D285256BDB@webforums>
References: <2axvuOqCCHA.255@forums.sybase.com> <nuV$glADCHA.81@forums.sybase.com> <3D0C68CB.D0317E40@idx.com.au> <3D0DFBE0.A079F082@dgny.com>
Lines: 22
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
Path: forums-1-dub!forums-master.sybase.com!forums.sybase.com!webforums.sybase.com!news
Xref: forums-1-dub sybase.public.ase.product_futures_discussion:427
Article PK: 93599

Hi Eugene,

thanks for the pointer in your earlier posting, I'm going to read the
article.

I find it unacceptable for a database to sleep on a SELECT statement by
effectively saying: "I cannot tell you the answer right now, because
someone else is updating the data". The idea of a System State Number says
it all, and I vaguely remember learning similar theories at the Discrete
Systems course in 1974. It seems no-one at Sybase has found time to read
the first 20 pages of the Oracle Concepts manual. I'm not blaming them as
individuals: most large corporations are busy worshiping their own idols.

In your earlier posting, you mentioned the site of Microsoft, which is
still being despised as a beginner in database matters. But MS has some
smart and open-minded guys, who are not merely repeating yesterday's
mantras. They introduced row-level locking quite early, and I would not be
surprised if they come out soon with Multi Version Concurrency in one of
their rdbms products.

Meanwhile, back to Transact-SQL - it still pays my bills, but I'm not sure
for how long. I'm really worried.

Andrew


Rob Verschoor Posted on 2002-06-18 20:50:01.0Z
Reply-To: "Rob Verschoor" <rob@DO.NOT.SPAM.sypron.nl.REMOVE.THIS.DECOY>
From: "Rob Verschoor" <rob@DO.NOT.SPAM.sypron.nl.REMOVE.THIS.DECOY>
References: <2axvuOqCCHA.255@forums.sybase.com> <nuV$glADCHA.81@forums.sybase.com> <3D0C68CB.D0317E40@idx.com.au> <3D0DFBE0.A079F082@dgny.com> <B31D92578E292847002AD4FD85256BDC.0059C6D285256BDB@webforums>
Subject: Re: multi versioning concurrency (MVC)
Date: Tue, 18 Jun 2002 22:50:01 +0200
Lines: 54
Organization: Sypron B.V.
MIME-Version: 1.0
Content-Type: text/plain; charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-MSMail-Priority: Normal
X-Newsreader: Microsoft Outlook Express 5.00.2919.6600
X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2919.6600
Message-ID: <UntuauwFCHA.196@forums.sybase.com>
Newsgroups: sybase.public.ase.product_futures_discussion
NNTP-Posting-Host: a74246.upc-a.chello.nl 62.163.74.246
Path: forums-1-dub!forums-master.sybase.com!forums.sybase.com
Xref: forums-1-dub sybase.public.ase.product_futures_discussion:426
Article PK: 93595


<andrew.schonberger@macquarie.com-nospam> wrote in message
news:B31D92578E292847002AD4FD85256BDC.0059C6D285256BDB@webforums...
> Hi Eugene,
>
> thanks for the pointer in your earlier posting, I'm going to read
the
> article.
>
> I find it unacceptable for a database to sleep on a SELECT statement
by
> effectively saying: "I cannot tell you the answer right now, because
> someone else is updating the data". The idea of a System State
Number says
> it all, and I vaguely remember learning similar theories at the
Discrete
> Systems course in 1974. It seems no-one at Sybase has found time
to read
> the first 20 pages of the Oracle Concepts manual. I'm not blaming
them as
> individuals: most large corporations are busy worshiping their own
idols.
>
> In your earlier posting, you mentioned the site of Microsoft, which
is
> still being despised as a beginner in database matters. But MS has
some
> smart and open-minded guys, who are not merely repeating yesterday's
> mantras. They introduced row-level locking quite early, and I would
not be
> surprised if they come out soon with Multi Version Concurrency in
one of
> their rdbms products.
>
> Meanwhile, back to Transact-SQL - it still pays my bills, but I'm
not sure
> for how long. I'm really worried.
>
> Andrew

I think you're attaching too much value to the issue of MVC as if it
is the holy grail which will make all database-related worries go
away. At the 1998 European ISUG conference, Peter Thawley mentioned
(during one of his presentations) that they'd looked into implementing
versioning, but they decided not to do this because of the big impact
it would have on performance.
While I agree that versioning is a nice feature, I'm sure you're
familiar with the concept of a "tradeoff". Just holding on to the fact
that anyone who doesn't implement versioning is plain stupid (if I
may, this is the essence of your posts) seems rather unreasonable and
short-sighted to me.

Rob Verschoor


Matt Rogish Posted on 2002-06-19 00:39:13.0Z
From: "Matt Rogish" <matt@fanhome.com>
References: <2axvuOqCCHA.255@forums.sybase.com> <nuV$glADCHA.81@forums.sybase.com> <3D0C68CB.D0317E40@idx.com.au> <3D0DFBE0.A079F082@dgny.com> <B31D92578E292847002AD4FD85256BDC.0059C6D285256BDB@webforums> <UntuauwFCHA.196@forums.sybase.com>
Subject: Re: multi versioning concurrency (MVC)
Date: Tue, 18 Jun 2002 20:39:13 -0400
Lines: 41
X-Priority: 3
X-MSMail-Priority: Normal
X-Newsreader: Microsoft Outlook Express 6.00.2600.0000
X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2600.0000
Message-ID: <1I1$GlyFCHA.196@forums.sybase.com>
Newsgroups: sybase.public.ase.product_futures_discussion
NNTP-Posting-Host: user-0c932as.cable.mindspring.com 24.145.137.92
Path: forums-1-dub!forums-master.sybase.com!forums.sybase.com
Xref: forums-1-dub sybase.public.ase.product_futures_discussion:425
Article PK: 93594

Rob,

I don't think Andrew (and stop me if I'm wrong) is necessarily stating MVC
is the greatest thing since sliced bread, but *Joe Manager* who makes
decisions based upon the neat power point presentations from vendors has no
time, nor the inclination, to really delve into technical merits of software
and believes what his sales vendors tell him. Are there products which
would suit companies better than Oracle? Probably -- cheaper, faster, etc.
but for *some reason* they chose Oracle. Maybe, just maybe, if Sybase had
this feature then there'd be *no* reason to choose Oracle over Sybase ASE.
Just like RLL, it's not really a requirement nor needed for a lot of
applications (we use it for a couple tables, but that's due to some weird
'special cases') but it was a 'market need' (much like XML, MVC, etc.).

Are cursors the 'best' idea for relational dbs? Probably not (I hate
them!), but they're there. No one necessarily has to use MVC but at least
if it was there (maybe different bins could provide MVC so as not to taint
the current builds) then to people who don't know any better it may take
away a *perceived* negative from ASE and might make it a viable contender
for their project.
--
Matt

"Rob Verschoor" <rob@DO.NOT.SPAM.sypron.nl.REMOVE.THIS.DECOY> wrote in
message news:UntuauwFCHA.196@forums.sybase.com...
> I think you're attaching too much value to the issue of MVC as if it
> is the holy grail which will make all database-related worries go
> away. At the 1998 European ISUG conference, Peter Thawley mentioned
> (during one of his presentations) that they'd looked into implementing
> versioning, but they decided not to do this because of the big impact
> it would have on performance.
> While I agree that versioning is a nice feature, I'm sure you're
> familiar with the concept of a "tradeoff". Just holding on to the fact
> that anyone who doesn't implement versioning is plain stupid (if I
> may, this is the essence of your posts) seems rather unreasonable and
> short-sighted to me.
>
> Rob Verschoor


Anthony Mandic Posted on 2002-06-22 09:46:57.0Z
Message-ID: <3D144791.A5EA8284@start.com.au>
Date: Sat, 22 Jun 2002 19:46:57 +1000
From: Anthony Mandic <spam_block@start.com.au>
Organization: Mandic Consulting Pty. Ltd.
X-Mailer: Mozilla 4.61 [en] (WinNT; I)
MIME-Version: 1.0
Subject: Re: multi versioning concurrency (MVC)
References: <2axvuOqCCHA.255@forums.sybase.com> <nuV$glADCHA.81@forums.sybase.com> <3D0C68CB.D0317E40@idx.com.au> <3D0DFBE0.A079F082@dgny.com> <B31D92578E292847002AD4FD85256BDC.0059C6D285256BDB@webforums> <UntuauwFCHA.196@forums.sybase.com> <1I1$GlyFCHA.196@forums.sybase.com>
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: 8bit
Newsgroups: sybase.public.ase.product_futures_discussion
Lines: 80
NNTP-Posting-Host: 203.3.176.10
Path: forums-1-dub!forums-master.sybase.com!forums.sybase.com
Xref: forums-1-dub sybase.public.ase.product_futures_discussion:415
Article PK: 93588


Matt Rogish wrote:

> I don't think Andrew (and stop me if I'm wrong) is necessarily stating MVC
> is the greatest thing since sliced bread, but *Joe Manager* who makes
> decisions based upon the neat power point presentations from vendors has no
> time, nor the inclination, to really delve into technical merits of software
> and believes what his sales vendors tell him. Are there products which
> would suit companies better than Oracle? Probably -- cheaper, faster, etc.
> but for *some reason* they chose Oracle. Maybe, just maybe, if Sybase had
> this feature then there'd be *no* reason to choose Oracle over Sybase ASE.
> Just like RLL, it's not really a requirement nor needed for a lot of
> applications (we use it for a couple tables, but that's due to some weird
> 'special cases') but it was a 'market need' (much like XML, MVC, etc.).

I think you're oversimplifying the decision making process. Its
true that a lot of managers are idiots who are in over their
heads but there are also managers who are capable of independent
thought. However, the purchase of a database backend often comes
in as an afterthought. It usually happens that said managers are
buying third party software that they require to run their businesses.
Said software has to run on a database platform. This might come
supplied (usually some cobbled together C-ISAM hack) or it might
come pre-packaged with a database platform. Some third party
vendors only support some database platforms or they might make
recommendations. When there isn't one, a manager would be free
to choose what ever they like. There's nothing stopping them
buying the most expensive, most popular or the cheapest.

Some third party products were so poorly written that they really
needed RLL. A simple rewrite wasn't on the cards. Its simpler to
use a database backend that supports it. The end result is that
these poor applications still abound.

The other option is in-house developed software. A manager might
look around to see what's available but he would also look around
to see what resources he can get as well (at the moment there
appears to be a lot of Orable DBAs and developers about due to
the dot.com collapse - if they are a dime a dozen why not use
them and save? Of course, Orable's licensing compensates for
this). The smarter ones would evaluate the total cost of
ownership.

> Are cursors the 'best' idea for relational dbs? Probably not (I hate
> them!), but they're there.

Cursors are really an issue of data model design and code
development - tied in with leveraging a relational database in
a non-relational way.

> No one necessarily has to use MVC but at least
> if it was there (maybe different bins could provide MVC so as not to taint
> the current builds) then to people who don't know any better it may take
> away a *perceived* negative from ASE and might make it a viable contender
> for their project.

I would think it would be better to look at MVC in terms of
what it can provide and what the alternatives are and how
effective they are in comparison.

-am © 2002


Rob Verschoor Posted on 2002-06-19 07:33:20.0Z
Reply-To: "Rob Verschoor" <rob@DO.NOT.SPAM.sypron.nl.REMOVE.THIS.DECOY>
From: "Rob Verschoor" <rob@DO.NOT.SPAM.sypron.nl.REMOVE.THIS.DECOY>
References: <2axvuOqCCHA.255@forums.sybase.com> <nuV$glADCHA.81@forums.sybase.com> <3D0C68CB.D0317E40@idx.com.au> <3D0DFBE0.A079F082@dgny.com> <B31D92578E292847002AD4FD85256BDC.0059C6D285256BDB@webforums> <UntuauwFCHA.196@forums.sybase.com> <1I1$GlyFCHA.196@forums.sybase.com>
Subject: Re: multi versioning concurrency (MVC)
Date: Wed, 19 Jun 2002 09:33:20 +0200
Lines: 73
Organization: Sypron B.V.
MIME-Version: 1.0
Content-Type: text/plain; charset="Windows-1252"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-MSMail-Priority: Normal
X-Newsreader: Microsoft Outlook Express 5.00.2919.6600
X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2919.6600
Message-ID: <Gmyb5W2FCHA.111@forums.sybase.com>
Newsgroups: sybase.public.ase.product_futures_discussion
NNTP-Posting-Host: asd3-p0391.dial.euronet.nl 81.68.1.135
Path: forums-1-dub!forums-master.sybase.com!forums.sybase.com
Xref: forums-1-dub sybase.public.ase.product_futures_discussion:424
Article PK: 93593

I agree that Oracle generally masters the art of selling/marketing
better than Sybase. But as I understood it, Andrew's point was very
much about the technical merit of MVC just by itself (if I
misunderstood, apolgies).

Rob V.

"Matt Rogish" <matt@fanhome.com> wrote in message
news:1I1$GlyFCHA.196@forums.sybase.com...
> Rob,
>
> I don't think Andrew (and stop me if I'm wrong) is necessarily
stating MVC
> is the greatest thing since sliced bread, but *Joe Manager* who
makes
> decisions based upon the neat power point presentations from vendors
has no
> time, nor the inclination, to really delve into technical merits of
software
> and believes what his sales vendors tell him. Are there products
which
> would suit companies better than Oracle? Probably -- cheaper,
faster, etc.
> but for *some reason* they chose Oracle. Maybe, just maybe, if
Sybase had
> this feature then there'd be *no* reason to choose Oracle over
Sybase ASE.
> Just like RLL, it's not really a requirement nor needed for a lot of
> applications (we use it for a couple tables, but that's due to some
weird
> 'special cases') but it was a 'market need' (much like XML, MVC,
etc.).
>
> Are cursors the 'best' idea for relational dbs? Probably not (I
hate
> them!), but they're there. No one necessarily has to use MVC but at
least
> if it was there (maybe different bins could provide MVC so as not to
taint
> the current builds) then to people who don't know any better it may
take
> away a *perceived* negative from ASE and might make it a viable
contender
> for their project.
> --
> Matt
>
> "Rob Verschoor" <rob@DO.NOT.SPAM.sypron.nl.REMOVE.THIS.DECOY> wrote
in
> message news:UntuauwFCHA.196@forums.sybase.com...
> > I think you're attaching too much value to the issue of MVC as if
it
> > is the holy grail which will make all database-related worries go
> > away. At the 1998 European ISUG conference, Peter Thawley
mentioned
> > (during one of his presentations) that they'd looked into
implementing
> > versioning, but they decided not to do this because of the big
impact
> > it would have on performance.
> > While I agree that versioning is a nice feature, I'm sure you're
> > familiar with the concept of a "tradeoff". Just holding on to the
fact
> > that anyone who doesn't implement versioning is plain stupid (if I
> > may, this is the essence of your posts) seems rather unreasonable
and
> > short-sighted to me.
> >
> > Rob Verschoor
>
>
>


Matt Rogish Posted on 2002-06-19 11:26:53.0Z
From: "Matt Rogish" <matt@fanhome.com>
References: <2axvuOqCCHA.255@forums.sybase.com> <nuV$glADCHA.81@forums.sybase.com> <3D0C68CB.D0317E40@idx.com.au> <3D0DFBE0.A079F082@dgny.com> <B31D92578E292847002AD4FD85256BDC.0059C6D285256BDB@webforums> <UntuauwFCHA.196@forums.sybase.com> <1I1$GlyFCHA.196@forums.sybase.com> <Gmyb5W2FCHA.111@forums.sybase.com>
Subject: Re: multi versioning concurrency (MVC)
Date: Wed, 19 Jun 2002 07:26:53 -0400
Lines: 18
X-Priority: 3
X-MSMail-Priority: Normal
X-Newsreader: Microsoft Outlook Express 6.00.2600.0000
X-MimeOLE: Produced By Microsoft MimeOLE V6.00.2600.0000
Message-ID: <HifBCP4FCHA.196@forums.sybase.com>
Newsgroups: sybase.public.ase.product_futures_discussion
NNTP-Posting-Host: user-0c932as.cable.mindspring.com 24.145.137.92
Path: forums-1-dub!forums-master.sybase.com!forums.sybase.com
Xref: forums-1-dub sybase.public.ase.product_futures_discussion:423
Article PK: 93591

Rob,

I understand, I guess Andrew will have to clarify further! :)

--
Matt

"Rob Verschoor" <rob@DO.NOT.SPAM.sypron.nl.REMOVE.THIS.DECOY> wrote in
message news:Gmyb5W2FCHA.111@forums.sybase.com...
> I agree that Oracle generally masters the art of selling/marketing
> better than Sybase. But as I understood it, Andrew's point was very
> much about the technical merit of MVC just by itself (if I
> misunderstood, apolgies).
>
> Rob V.


andrew.schonberger Posted on 2002-06-21 05:50:25.0Z
From: andrew.schonberger@macquarie.com-no-spam-suffix
Date: Fri, 21 Jun 2002 01:50:25 -0400
Newsgroups: sybase.public.ase.product_futures_discussion
Subject: Re: multi versioning concurrency (MVC)
Message-ID: <0B9DBA059B0454B3002014AC85256BDF.0040EB4685256BDD@webforums>
References: <2axvuOqCCHA.255@forums.sybase.com> <nuV$glADCHA.81@forums.sybase.com> <3D0C68CB.D0317E40@idx.com.au> <3D0DFBE0.A079F082@dgny.com> <B31D92578E292847002AD4FD85256BDC.0059C6D285256BDB@webforums> <UntuauwFCHA.196@forums.sybase.com> <1I1$GlyFCHA.196@forums.sybase.com> <Gmyb5W2FCHA.111@forums.sybase.com> <HifBCP4FCHA.196@forums.sybase.com>
Lines: 35
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
Path: forums-1-dub!forums-master.sybase.com!forums.sybase.com!webforums.sybase.com!news
Xref: forums-1-dub sybase.public.ase.product_futures_discussion:420
Article PK: 93587

Rob,

please don't shoot the messenger. The lack of MVC is not affecting
Transact-SQL developers (I include myself into this category). Rather, it
affects people who write code in client-side procedural languages, in one
of the fashionable OO-modular-beans architectures. It's a sizeable market,
and they won't go away.

We cannot convince Sun and 200,000 Java programmers that their EJB
architecture is all wrong. Any the many others writing code in VB, Cgi-Perl
or whatever. They will still develop (unwittingly) huge transactions which
bring together separate modules. Those systems will run reasonably well on
Oracle. But any such system quickly reduces Sybase to a single-threaded
engine.

I repeat the scenario, which I've seen all too frequently:

- a transaction is started
- module 1 reads from table A, then updates a record in A
- module 2 reads from table B, then updates a record in B
- module 3 same pattern on C, but the read takes 90% of the total time for
the entire transaction
- Commit

As you can see, I simplified the scenario, by assuming modules do not
access each others tables. Even so, multiple instances of the process above
will sooner or later lock one another. The lengthy validation in the last
module prevents locks on A and B from being released. In most practical
cases, the records being read and updated are the in the "active portion"
of historical tables, e.g. trades done today.

As Sybase "gurus", we could argue (for this example) that module C should
be placed first. But this defeats the purpose of modular isolation.
Besides, it works on Oracle - I find it hard to argue against that.

Andrew


Rob Verschoor Posted on 2002-06-24 10:56:18.0Z
Reply-To: "Rob Verschoor" <rob@DO.NOT.SPAM.sypron.nl.REMOVE.THIS.DECOY>
From: "Rob Verschoor" <rob@DO.NOT.SPAM.sypron.nl.REMOVE.THIS.DECOY>
References: <2axvuOqCCHA.255@forums.sybase.com> <nuV$glADCHA.81@forums.sybase.com> <3D0C68CB.D0317E40@idx.com.au> <3D0DFBE0.A079F082@dgny.com> <B31D92578E292847002AD4FD85256BDC.0059C6D285256BDB@webforums> <UntuauwFCHA.196@forums.sybase.com> <1I1$GlyFCHA.196@forums.sybase.com> <Gmyb5W2FCHA.111@forums.sybase.com> <HifBCP4FCHA.196@forums.sybase.com> <0B9DBA059B0454B3002014AC85256BDF.0040EB4685256BDD@webforums>
Subject: Re: multi versioning concurrency (MVC)
Date: Mon, 24 Jun 2002 12:56:18 +0200
Lines: 111
Organization: Sypron B.V.
MIME-Version: 1.0
Content-Type: text/plain; charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-MSMail-Priority: Normal
X-Newsreader: Microsoft Outlook Express 5.00.2919.6600
X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2919.6600
Message-ID: <yf$YR92GCHA.1010@forums.sybase.com>
Newsgroups: sybase.public.ase.product_futures_discussion
NNTP-Posting-Host: a74246.upc-a.chello.nl 62.163.74.246
Path: forums-1-dub!forums-master.sybase.com!forums.sybase.com
Xref: forums-1-dub sybase.public.ase.product_futures_discussion:412
Article PK: 93578


<andrew.schonberger@macquarie.com-no-spam-suffix> wrote in message
news:0B9DBA059B0454B3002014AC85256BDF.0040EB4685256BDD@webforums...
> Rob,
>
> please don't shoot the messenger. The lack of MVC is not affecting
> Transact-SQL developers (I include myself into this category).
Rather, it
> affects people who write code in client-side procedural languages,
in one
> of the fashionable OO-modular-beans architectures. It's a sizeable
market,
> and they won't go away.
>
> We cannot convince Sun and 200,000 Java programmers that their EJB
> architecture is all wrong. Any the many others writing code in VB,
Cgi-Perl
> or whatever. They will still develop (unwittingly) huge transactions
which
> bring together separate modules. Those systems will run reasonably
well on
> Oracle. But any such system quickly reduces Sybase to a
single-threaded
> engine.
>
> I repeat the scenario, which I've seen all too frequently:
>
> - a transaction is started
> - module 1 reads from table A, then updates a record in A
> - module 2 reads from table B, then updates a record in B
> - module 3 same pattern on C, but the read takes 90% of the total
time for
> the entire transaction
> - Commit
>
> As you can see, I simplified the scenario, by assuming modules do
not
> access each others tables. Even so, multiple instances of the
process above
> will sooner or later lock one another. The lengthy validation in the
last
> module prevents locks on A and B from being released. In most
practical
> cases, the records being read and updated are the in the "active
portion"
> of historical tables, e.g. trades done today.
>
> As Sybase "gurus", we could argue (for this example) that module C
should
> be placed first. But this defeats the purpose of modular isolation.
> Besides, it works on Oracle - I find it hard to argue against that.
>
> Andrew

Andrew,

IMO, there are two issues here.

First, there's the functionality of versioning itself. As a feature by
itself, I like it, and I would be happy if Sybase would implement this
in ASE. If Sybase could position itself like "if it runs on Oracle, it
runs faster on Sybase" then this would be a good thing. I don't
disagree with you here.

Second, there's the opinion as expressed in earlier postings that
versioning is an essential must-have; especially because the merits
have been described in a research paper from years ago, so there would
be no excuse for not implementing it (my summary of your words). My
criticism was against this reasoning, because I think there's just not
much of a point here: research is years if not decades ahead of
commercial implementation. For example, just take the date of birth of
Codd's relational model and the moment when RDBMSs were mature enough
to replace IMS/CICS-based mainframe applications -- there were more
than 15 years in between. There are many interesting ideas which
haven't (yet) made it into commercial products as a result of
practical, real-world tradeoffs.

I recognise the type of problem your describe in your example -- this
type of no-brain approach occurs a lot in practice. While versioning
would indeed ease the pain a bit here, I think we're fighting syptoms
rather than causes here: I've also seen cases where versioning would
not help at all (for example, there was this application where every
new order would add its amount to a column in a 1-row table which
contained the total outstanding order amount -- so all transactions
will serialise on this point). Also, the very concept of a
"transaction" doesn't always seem to be very clear to some
designers/developers. What I'm saying here is that developing database
apps bigger than Mickey-Mouse size *does* require some common sense
and understanding of what you're doing. Still I agree that versioning
would be nice to have -- but only if it would not cost performance
(but it will, so it's a rather expensive solution).
Summarising, I'm not against versioning per se, but I don't agree with
the reasoning behind it as worded in previous posts. Hence my response
to those posts.

HTH,

Rob
----------------------------------------------------------------------
Rob Verschoor

Certified Sybase Professional DBA for ASE 12.0/11.5/11.0

Author of "The Complete Sybase ASE Quick Reference Guide"
Online orders accepted at http://www.sypron.nl/qr

email mailto:rob@DO.NOT.SPAM.sypron.nl.REMOVE.THIS.DECOY
WWW http://www.sypron.nl
snail Sypron B.V., P.O.Box 10695, 2501HR Den Haag, The Netherlands
----------------------------------------------------------------------


andrew.schonberger Posted on 2002-06-25 02:16:47.0Z
From: andrew.schonberger@macquarie.com.no-spam
Date: Mon, 24 Jun 2002 22:16:47 -0400
Newsgroups: sybase.public.ase.product_futures_discussion
Subject: Re: multi versioning concurrency (MVC)
Message-ID: <396C890EBAF4D031000C85FC85256BE3.003FAE2885256BE2@webforums>
References: <2axvuOqCCHA.255@forums.sybase.com> <nuV$glADCHA.81@forums.sybase.com> <3D0C68CB.D0317E40@idx.com.au> <3D0DFBE0.A079F082@dgny.com> <B31D92578E292847002AD4FD85256BDC.0059C6D285256BDB@webforums> <UntuauwFCHA.196@forums.sybase.com> <1I1$GlyFCHA.196@forums.sybase.com> <Gmyb5W2FCHA.111@forums.sybase.com> <HifBCP4FCHA.196@forums.sybase.com> <0B9DBA059B0454B3002014AC85256BDF.0040EB4685256BDD@webforums> <yf$YR92GCHA.1010@forums.sybase.com>
Lines: 40
MIME-Version: 1.0
Content-Type: text/plain; charset="us-ascii"
Content-Transfer-Encoding: 7bit
Path: forums-1-dub!forums-master.sybase.com!forums.sybase.com!webforums.sybase.com!news
Xref: forums-1-dub sybase.public.ase.product_futures_discussion:411
Article PK: 93581

Hi,

this was indeed an interesting thread. The main thing I've learned was that
"multiversion concurrency" isn't just a commercial term used by Oracle -
I've found heaps of research papers on the subject.

I apologise if emotional language has found it's way to an earlier posting.
And I'm glad we've are back to rational analysis, thanks for your patience.

Rob, I agree whith what you wrote:

>> If Sybase could position itself like
>> if it runs on Oracle,
>> it runs faster on Sybase
>> then this would be a good thing

and:

>> developing database apps [...] does require some common
>> sense and understanding of what you're doing.

On this later issue, help is under way from unexpected quarters. In his
latest book, the Oracle evangelist Thomas Kyte makes the same point, almost
word by word( chapter One available for download from Amazon ). Apparently,
the big Oh is also feeling the pressure from large masses of client-side
programmers.

>> versioning would be nice to have
>> but only if it would not cost performance

My experience (may or may not be typical), is that with well-tuned query
plans we seldom had pure performace problems. Most of the problems we did
have were attributable to locking and concurrency. The likely reason is
that I had to support some highly analytical systems written (risk
management, yield curves, live P&L calculation). These systems, written
elsewhere, applied complex calculations on a relatively small amount of
data. I also had to develop and support "integration jobs", essentially
polling queues accross servers.

Regards,

Andrew


Anthony Mandic Posted on 2002-06-26 09:43:48.0Z
Message-ID: <3D198CD4.EACB828F@start.com.au>
Date: Wed, 26 Jun 2002 19:43:48 +1000
From: Anthony Mandic <spam_block@start.com.au>
Organization: Mandic Consulting Pty. Ltd.
X-Mailer: Mozilla 4.61 [en] (WinNT; I)
MIME-Version: 1.0
Subject: Re: multi versioning concurrency (MVC)
References: <2axvuOqCCHA.255@forums.sybase.com> <nuV$glADCHA.81@forums.sybase.com> <3D0C68CB.D0317E40@idx.com.au> <3D0DFBE0.A079F082@dgny.com> <B31D92578E292847002AD4FD85256BDC.0059C6D285256BDB@webforums> <UntuauwFCHA.196@forums.sybase.com> <1I1$GlyFCHA.196@forums.sybase.com> <Gmyb5W2FCHA.111@forums.sybase.com> <HifBCP4FCHA.196@forums.sybase.com> <0B9DBA059B0454B3002014AC85256BDF.0040EB4685256BDD@webforums> <yf$YR92GCHA.1010@forums.sybase.com> <396C890EBAF4D031000C85FC85256BE3.003FAE2885256BE2@webforums>
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: 8bit
Newsgroups: sybase.public.ase.product_futures_discussion
Lines: 7
NNTP-Posting-Host: 203.3.176.10
Path: forums-1-dub!forums-master.sybase.com!forums.sybase.com
Xref: forums-1-dub sybase.public.ase.product_futures_discussion:407
Article PK: 93575


andrew.schonberger@macquarie.com.no-spam wrote:

> In his latest book, the Oracle evangelist Thomas Kyte

[You'll have to excuse me while I laugh uncontrollably.]
Long term readers of comp.databases.sybase will be very familiar
this name.

-am © 2002


Rob Verschoor Posted on 2002-06-25 05:53:44.0Z
Reply-To: "Rob Verschoor" <rob@DO.NOT.SPAM.sypron.nl.REMOVE.THIS.DECOY>
From: "Rob Verschoor" <rob@DO.NOT.SPAM.sypron.nl.REMOVE.THIS.DECOY>
References: <2axvuOqCCHA.255@forums.sybase.com> <nuV$glADCHA.81@forums.sybase.com> <3D0C68CB.D0317E40@idx.com.au> <3D0DFBE0.A079F082@dgny.com> <B31D92578E292847002AD4FD85256BDC.0059C6D285256BDB@webforums> <UntuauwFCHA.196@forums.sybase.com> <1I1$GlyFCHA.196@forums.sybase.com> <Gmyb5W2FCHA.111@forums.sybase.com> <HifBCP4FCHA.196@forums.sybase.com> <0B9DBA059B0454B3002014AC85256BDF.0040EB4685256BDD@webforums> <yf$YR92GCHA.1010@forums.sybase.com> <396C890EBAF4D031000C85FC85256BE3.003FAE2885256BE2@webforums>
Subject: Re: multi versioning concurrency (MVC)
Date: Tue, 25 Jun 2002 07:53:44 +0200
Lines: 14
Organization: Sypron B.V.
MIME-Version: 1.0
Content-Type: text/plain; charset="iso-8859-1"
Content-Transfer-Encoding: 7bit
X-Priority: 3
X-MSMail-Priority: Normal
X-Newsreader: Microsoft Outlook Express 5.00.2919.6600
X-MimeOLE: Produced By Microsoft MimeOLE V5.00.2919.6600
Message-ID: <UjI3L6AHCHA.654@forums.sybase.com>
Newsgroups: sybase.public.ase.product_futures_discussion
NNTP-Posting-Host: a74246.upc-a.chello.nl 62.163.74.246
Path: forums-1-dub!forums-master.sybase.com!forums.sybase.com
Xref: forums-1-dub sybase.public.ase.product_futures_discussion:410
Article PK: 93580

For this type of 'polling', look at the "readpast" locking option in
ASE; it will probably solve many of your concurrency problems,
especially in combination with datarows locking.

Rob V.

> data. I also had to develop and support "integration jobs",
essentially
> polling queues accross servers.
>
> Regards,
>
> Andrew


Anthony Mandic Posted on 2002-06-22 09:53:18.0Z
Message-ID: <3D14490E.2A6D5D23@start.com.au>
Date: Sat, 22 Jun 2002 19:53:18 +1000
From: Anthony Mandic <spam_block@start.com.au>
Organization: Mandic Consulting Pty. Ltd.
X-Mailer: Mozilla 4.61 [en] (WinNT; I)
MIME-Version: 1.0
Subject: Re: multi versioning concurrency (MVC)
References: <2axvuOqCCHA.255@forums.sybase.com> <nuV$glADCHA.81@forums.sybase.com> <3D0C68CB.D0317E40@idx.com.au> <3D0DFBE0.A079F082@dgny.com> <B31D92578E292847002AD4FD85256BDC.0059C6D285256BDB@webforums> <UntuauwFCHA.196@forums.sybase.com> <1I1$GlyFCHA.196@forums.sybase.com> <Gmyb5W2FCHA.111@forums.sybase.com> <HifBCP4FCHA.196@forums.sybase.com> <0B9DBA059B0454B3002014AC85256BDF.0040EB4685256BDD@webforums>
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: 8bit
Newsgroups: sybase.public.ase.product_futures_discussion
Lines: 31
NNTP-Posting-Host: 203.3.176.10
Path: forums-1-dub!forums-master.sybase.com!forums.sybase.com
Xref: forums-1-dub sybase.public.ase.product_futures_discussion:414
Article PK: 93585


andrew.schonberger@macquarie.com-no-spam-suffix wrote:

> I repeat the scenario, which I've seen all too frequently:
>
> - a transaction is started
> - module 1 reads from table A, then updates a record in A
> - module 2 reads from table B, then updates a record in B
> - module 3 same pattern on C, but the read takes 90% of the total time for
> the entire transaction
> - Commit
>
> As you can see, I simplified the scenario, by assuming modules do not
> access each others tables. Even so, multiple instances of the process above
> will sooner or later lock one another. The lengthy validation in the last
> module prevents locks on A and B from being released. In most practical
> cases, the records being read and updated are the in the "active portion"
> of historical tables, e.g. trades done today.
>
> As Sybase "gurus", we could argue (for this example) that module C should
> be placed first. But this defeats the purpose of modular isolation.
> Besides, it works on Oracle - I find it hard to argue against that.

I would first ask if an appropriate isolation level and RLL
would suffice. Its not really the same as MVC but it should
be good enough in the majority of cases.

-am © 2002


Mike Harrold Posted on 2002-06-25 14:21:35.0Z
Subject: Re: multi versioning concurrency (MVC)
References: <2axvuOqCCHA.255@forums.sybase.com> <HifBCP4FCHA.196@forums.sybase.com> <0B9DBA059B0454B3002014AC85256BDF.0040EB4685256BDD@webforums> <3D14490E.2A6D5D23@start.com.au>
X-Newsreader: trn 4.0-test75 (Feb 13, 2001)
From: ao@shell.core.com (Mike Harrold)
Originator: ao@shell.core.com (Mike Harrold)
Message-ID: <1efpcOFHCHA.290@forums.sybase.com>
Newsgroups: sybase.public.ase.product_futures_discussion
Date: Tue, 25 Jun 2002 10:21:35 -0400
Lines: 37
NNTP-Posting-Host: shell.core.com 169.207.1.89
Path: forums-1-dub!forums-master.sybase.com!forums.sybase.com!not-for-mail
Xref: forums-1-dub sybase.public.ase.product_futures_discussion:409
Article PK: 93579

In article <3D14490E.2A6D5D23@start.com.au>,

Anthony Mandic <spam_block@start.com.au> wrote:
>
>andrew.schonberger@macquarie.com-no-spam-suffix wrote:
>
>> I repeat the scenario, which I've seen all too frequently:
>>
>> - a transaction is started
>> - module 1 reads from table A, then updates a record in A
>> - module 2 reads from table B, then updates a record in B
>> - module 3 same pattern on C, but the read takes 90% of the total time for
>> the entire transaction
>> - Commit
>>
>> As you can see, I simplified the scenario, by assuming modules do not
>> access each others tables. Even so, multiple instances of the process above
>> will sooner or later lock one another. The lengthy validation in the last
>> module prevents locks on A and B from being released. In most practical
>> cases, the records being read and updated are the in the "active portion"
>> of historical tables, e.g. trades done today.
>>
>> As Sybase "gurus", we could argue (for this example) that module C should
>> be placed first. But this defeats the purpose of modular isolation.
>> Besides, it works on Oracle - I find it hard to argue against that.
>
> I would first ask if an appropriate isolation level and RLL
> would suffice. Its not really the same as MVC but it should
> be good enough in the majority of cases.

That was my reaction; I am curious to hear of a situation where this
would not work?

/Mike


Anthony Mandic Posted on 2002-06-26 09:46:31.0Z
Message-ID: <3D198D77.1104BDD5@start.com.au>
Date: Wed, 26 Jun 2002 19:46:31 +1000
From: Anthony Mandic <spam_block@start.com.au>
Organization: Mandic Consulting Pty. Ltd.
X-Mailer: Mozilla 4.61 [en] (WinNT; I)
MIME-Version: 1.0
Subject: Re: multi versioning concurrency (MVC)
References: <2axvuOqCCHA.255@forums.sybase.com> <HifBCP4FCHA.196@forums.sybase.com> <0B9DBA059B0454B3002014AC85256BDF.0040EB4685256BDD@webforums> <3D14490E.2A6D5D23@start.com.au> <1efpcOFHCHA.290@forums.sybase.com>
Content-Type: text/plain; charset=iso-8859-1
Content-Transfer-Encoding: 8bit
Newsgroups: sybase.public.ase.product_futures_discussion
Lines: 19
NNTP-Posting-Host: 203.3.176.10
Path: forums-1-dub!forums-master.sybase.com!forums.sybase.com
Xref: forums-1-dub sybase.public.ase.product_futures_discussion:406
Article PK: 93574


Mike Harrold wrote:

> >> As Sybase "gurus", we could argue (for this example) that module C should
> >> be placed first. But this defeats the purpose of modular isolation.
> >> Besides, it works on Oracle - I find it hard to argue against that.
> >
> > I would first ask if an appropriate isolation level and RLL
> > would suffice. Its not really the same as MVC but it should
> > be good enough in the majority of cases.
>
> That was my reaction; I am curious to hear of a situation where this
> would not work?

Based on my experiences with shonky applications developed in
Orable and ported to Sybase I can't think of one. If they can be
made to work, better ones would work even better.

-am © 2002