.Net project failures

Having recently moved on to a BA role (After studying .Net for a while and
passing couple of ms.net exams )
I have witnessed couple of failed projects in .Net . Thought I might share
my sadness
Let me outline the project failed (Brief Discription)
Client: (namewithheld)
Supplier: (namewithheld from India(They are learning)) (But word is that MS
brains got involved in last minute)
Requirement: Handle 600 - 800 transations /Sec on Peak Hours
30-40 tables per Transaction / Numerous updates and inserts/ Little CPU more
IO due to DB
After shelling out few million dollars the project died since it
cannot match (not even close) legacy Performance offered by IBM
Personally I am quite sad at this outcome
Beacause
a. A Initial Development opportunity in .Net lost so other .net initiatives
are shelved
b.Missed chance to Kick IBM on butt !
c.>net developers had to shelf their desire to be on a good organisation
for long
d.Afraid that it sent wrong signals to other IT managers especially during
friday drinking time ....

And this is not the only project that has been failed
in .Net due to peformance implications
I personally know at least 1 other project .
I can't help wondering how many more are there

I wonder does anybody has been involved in project with
such performance implications
Any comments appreciated
------------------

Disillusioned_01

>Requirement: Handle 600 - 800 transations /Sec on Peak Hours
>30-40 tables per Transaction / Numerous updates and inserts/ Little CPU more
>IO due to DB

Simply down to poor infrastructure planning/understanding - don't tell
me, single database server!!! please don't say yes....

How application servers in the farm?

How many of those transactions come through the wan? How many web
servers in the farm?

I was involved with a banking system back in 1998 which was expected
to be able to handle 100,000 secure 'business transactions' an hour
(peek time = 1 to 1.5 hours per day), which is somewhat less than your
scenario, however this involved :-

32 web servers dual PII 450 512 ram.

Initially 16 application servers (same spec as the web server's I
think)

There were four Oracle database servers running on AIX (big spec
machines)

The whole system, as a result of performance testing was determined to
have a [theoretical] functional limit of about one million business
transaction per hour. In reality, the system never received more than
10,000 business transaction an hour, even at peek time, so the real
winner was idle time!

The main reason for the success of the system was the excellent
communication between the ms, ibm and oracle consultants, there was no
inter company bickering whatsoever - we would not have tolerated that
anyway!

Kline Sphere (Chalk) MCNGP #3

The Poster Formerly Known as Kline Sphere

>And this is not the only project that has been failed
>in .Net due to peformance implications

The development tools have nothing to do with why projects fail.
Projects fail because of incompetence brought about by stupidity and
ignorance. This project would just have likely have failed if it were
being developed on the j2ee platform.

Kline Sphere (Chalk) MCNGP #3

The Poster Formerly Known as Kline Sphere

well as far as concurrent transactions are concerned (without considering
the 30-40 table inserts) DB2 is the best thing. Read an article on
concurrent transactions databases..
SQL Server was almost as good as Oracle but DB2 is way ahead (more than
twice as many as oracle and still going strong).

As my bro once put it a few years back... DB2 had to be scaled down to make
it work on *nix / NT. Where as Oracle / Sql Server and the rest of them had
to be scaled up from uni-processor arch to use more than 1..

--

Regards,

Hermit Dave
(http://hdave.blogspot.com)
"Disillusioned_01" <> wrote in message
news:...
> Having recently moved on to a BA role (After studying .Net for a while and
> passing couple of ms.net exams )
> I have witnessed couple of failed projects in .Net . Thought I might
> share
> my sadness
> Let me outline the project failed (Brief Discription)
> Client: (namewithheld)
> Supplier: (namewithheld from India(They are learning)) (But word is that
> MS
> brains got involved in last minute)
> Requirement: Handle 600 - 800 transations /Sec on Peak Hours
> 30-40 tables per Transaction / Numerous updates and inserts/ Little CPU
> more
> IO due to DB
> After shelling out few million dollars the project died since it
> cannot match (not even close) legacy Performance offered by IBM
> Personally I am quite sad at this outcome
> Beacause
> a. A Initial Development opportunity in .Net lost so other .net
> initiatives
> are shelved
> b.Missed chance to Kick IBM on butt !
> c.>net developers had to shelf their desire to be on a good organisation
> for long
> d.Afraid that it sent wrong signals to other IT managers especially during
> friday drinking time ....
>
> And this is not the only project that has been failed
> in .Net due to peformance implications
> I personally know at least 1 other project .
> I can't help wondering how many more are there
>
> I wonder does anybody has been involved in project with
> such performance implications
> Any comments appreciated
> ------------------
>
>

Hermit Dave

>As my bro once put it a few years back... DB2 had to be scaled down to make
>it work on *nix / NT. Where as Oracle / Sql Server and the rest of them had
>to be scaled up from uni-processor arch to use more than 1..

nice.

Kline Sphere (Chalk) MCNGP #3

The Poster Formerly Known as Kline Sphere

Is inserting into 40 tables per transaction really a business requirement or
just the result of poor database design? I bet they didn't even start with a
proven industry specific meta model or hire professional database developers
to do the work. They probably just gave each programmer a DBO login and let
them do their own thang.
WKidd

"Disillusioned_01" <> wrote in message
news:...
> Having recently moved on to a BA role (After studying .Net for a while and
> passing couple of ms.net exams )
> I have witnessed couple of failed projects in .Net . Thought I might
share
> my sadness
> Let me outline the project failed (Brief Discription)
> Client: (namewithheld)
> Supplier: (namewithheld from India(They are learning)) (But word is that
MS
> brains got involved in last minute)
> Requirement: Handle 600 - 800 transations /Sec on Peak Hours
> 30-40 tables per Transaction / Numerous updates and inserts/ Little CPU
more
> IO due to DB
> After shelling out few million dollars the project died since it
> cannot match (not even close) legacy Performance offered by IBM
> Personally I am quite sad at this outcome
> Beacause
> a. A Initial Development opportunity in .Net lost so other .net
initiatives
> are shelved
> b.Missed chance to Kick IBM on butt !
> c.>net developers had to shelf their desire to be on a good organisation
> for long
> d.Afraid that it sent wrong signals to other IT managers especially during
> friday drinking time ....
>
> And this is not the only project that has been failed
> in .Net due to peformance implications
> I personally know at least 1 other project .
> I can't help wondering how many more are there
>
> I wonder does anybody has been involved in project with
> such performance implications
> Any comments appreciated
> ------------------
>
>

The Poster Formerly Known as Kline Sphere wrote:

>>And this is not the only project that has been failed
>>in .Net due to peformance implications
>
>
> The development tools have nothing to do with why projects fail.
> Projects fail because of incompetence brought about by stupidity and
> ignorance. This project would just have likely have failed if it were
> being developed on the j2ee platform.
>
> Kline Sphere (Chalk) MCNGP #3

here here.



--
------------------------
Think your smart?
Prove your programming power @ the OSI Geek Challenges
http://www.osix.net
------------------------

bb

Well please consider this a reply for every kind folk(and not so kind 'I
knew it all' teens)
We had an investment on existing application which took 7 years to develop(I
don't want hear some ignorant yelling
then some thing wrong with it at this point we did not have off shelf
applications @ that point in time ) on a db2 /MVS(Recent upgrade dont pick
on it)
all we done was to replace existing application using IBM prop hardware with
..Net /mq front end.
Well we thought MS guys would know something about app performance

I know(in theory) if you put an oracle 10 g Federated Server and some
machines around it It would work
well it is just that this little customised applications having around 900+
tables can't be ported
that easily

To kline it was only one DB (But I am not ashamed yet because it still
works )


"Hermit Dave" <> wrote in message
news:...
> well as far as concurrent transactions are concerned (without considering
> the 30-40 table inserts) DB2 is the best thing. Read an article on
> concurrent transactions databases..
> SQL Server was almost as good as Oracle but DB2 is way ahead (more than
> twice as many as oracle and still going strong).
>
> As my bro once put it a few years back... DB2 had to be scaled down to
make
> it work on *nix / NT. Where as Oracle / Sql Server and the rest of them
had
> to be scaled up from uni-processor arch to use more than 1..
>
> --
>
> Regards,
>
> Hermit Dave
> (http://hdave.blogspot.com)
> "Disillusioned_01" <> wrote in message
> news:...
> > Having recently moved on to a BA role (After studying .Net for a while
and
> > passing couple of ms.net exams )
> > I have witnessed couple of failed projects in .Net . Thought I might
> > share
> > my sadness
> > Let me outline the project failed (Brief Discription)
> > Client: (namewithheld)
> > Supplier: (namewithheld from India(They are learning)) (But word is that
> > MS
> > brains got involved in last minute)
> > Requirement: Handle 600 - 800 transations /Sec on Peak Hours
> > 30-40 tables per Transaction / Numerous updates and inserts/ Little CPU
> > more
> > IO due to DB
> > After shelling out few million dollars the project died since it
> > cannot match (not even close) legacy Performance offered by IBM
> > Personally I am quite sad at this outcome
> > Beacause
> > a. A Initial Development opportunity in .Net lost so other .net
> > initiatives
> > are shelved
> > b.Missed chance to Kick IBM on butt !
> > c.>net developers had to shelf their desire to be on a good
organisation
> > for long
> > d.Afraid that it sent wrong signals to other IT managers especially
during
> > friday drinking time ....
> >
> > And this is not the only project that has been failed
> > in .Net due to peformance implications
> > I personally know at least 1 other project .
> > I can't help wondering how many more are there
> >
> > I wonder does anybody has been involved in project with
> > such performance implications
> > Any comments appreciated
> > ------------------
> >
> >
>
>

Disillusioned_01

Guess so. But thanx for your insight!!
"The Poster Formerly Known as Kline Sphere" <.> wrote in message
news:...
> >And this is not the only project that has been failed
> >in .Net due to peformance implications
>
> The development tools have nothing to do with why projects fail.
> Projects fail because of incompetence brought about by stupidity and
> ignorance. This project would just have likely have failed if it were
> being developed on the j2ee platform.
>
> Kline Sphere (Chalk) MCNGP #3

Disillusioned_01

thanx . In don't have any practial experience in scaling up/out
web farms. Will try to read more
"The Poster Formerly Known as Kline Sphere" <.> wrote in message
news:...
> >Requirement: Handle 600 - 800 transations /Sec on Peak Hours
> >30-40 tables per Transaction / Numerous updates and inserts/ Little CPU
more
> >IO due to DB
>
> Simply down to poor infrastructure planning/understanding - don't tell
> me, single database server!!! please don't say yes....
>
> How application servers in the farm?
>
> How many of those transactions come through the wan? How many web
> servers in the farm?
>
> I was involved with a banking system back in 1998 which was expected
> to be able to handle 100,000 secure 'business transactions' an hour
> (peek time = 1 to 1.5 hours per day), which is somewhat less than your
> scenario, however this involved :-
>
> 32 web servers dual PII 450 512 ram.
>
> Initially 16 application servers (same spec as the web server's I
> think)
>
> There were four Oracle database servers running on AIX (big spec
> machines)
>
> The whole system, as a result of performance testing was determined to
> have a [theoretical] functional limit of about one million business
> transaction per hour. In reality, the system never received more than
> 10,000 business transaction an hour, even at peek time, so the real
> winner was idle time!
>
> The main reason for the success of the system was the excellent
> communication between the ms, ibm and oracle consultants, there was no
> inter company bickering whatsoever - we would not have tolerated that
> anyway!
>
> Kline Sphere (Chalk) MCNGP #3

Disillusioned_01