Hi Martin,
I didn't say you could not spend less than USD5M in 5 years on TD HW/SW etc...I said that if you don't have USD5M to spend don't buy TD.
The message was that if the project is not large enough to warrant that kind of spend you can almost certainly achieve what you want on other platforms for a lower price than doing a very similar thing on TD.
TD is great when you really do have very unstructured queries to run on very large volumes of data which do not lend themselves to dimensional or summarized styles of data models......there is a place in the world for TD but it is shrinking as the other vendors scale up and as other newer technologies eat away at the lower end of the DW size....which is getting past 20TB disk reasonably easily now....and not many companies have more than 20TB disk EDWs....
And no...people here can talk about what they want....freedom of speech and all.....
I just have a personal opinion that if more people in the BI or Data Warehousing world thought about business benefit more and feeds and speeds less we might actually deliver more business benefit.. ;-)
The vast majority of EDWs I see deliver no top line revenue at all and barely reduce costs if at all......
In my home country, Australia, we used to only begin sales cycles with prospects who had failed twice yet the business still believed there was value to be had and they were not about to let their 'feeds and speeds'
IT group stuff it up again.....then we could sell some real consulting!
And, sad to say, we had a target rich environment because there were plenty of failures....
Then again...in many ways I am pleased that so many people are so bad at this....makes it easy to 'look good'.. ;-)
Eg...you talk about subject areas and how many systems you brought you're your EDW...I talk about gaining a client 15% gross margin and 5% market share over two years......and I don't care what I did it on and how many systems I got data from....or how much it cost.... ;-)
I care the MD of the company was promoted to CEO of the group for the stellar business performance he was able to turn in year after year after we put in their BI Infrastructure... ;-)
I just find it strange that there is no public debate anywhere about business value of BI.....
Peter
-----Original Message-----
From: martin.willcox via dw-select
[
Sent: 14 August 2006 11:23
To: pete...
Subject: RE:[dw-select] RE:Re: RE: Teradata Vs Oracle
Pete,
To your point: "If you don't have USD5M to spend on hardware and
software over the next
2-3 years don't buy TD."
I was the Warehouse Manager at a major UK grocery retailer for
several years prior to joining Teradata; our Data Warhouse was
built on Teradata. In my time there we (1) bought a Teradata
system, (2) acquired data or had to re-engineer non-Y2K
compliant interfaces (an issue bequeathed to me by my
predecessor, must remember to thank him some time!) from about
10 source systems into 4 subject areas and (3) delivered
half-a-dozen major BI projects (and yes, realized a lot of
business value). We had significant change from USD5M - despite
the fact that we used Teradata Professional Services to do the
majority of our development work, i.e. we bought much more than
hardware and software from Teradata in that period.
To your "wasteful of time point": do we only get to talk about
the things things that are important to you, then, Pete...
:-)
MW
---------------Original Message---------------
>Hi All,
>I see discussions like this and I despair of our industry....
>
>In all this discussion nothing about delivering more sustainable
profit
>growth the the organisation....sigh...
>Anyway, feeds and speeds....Sybase IQ...
>I've used it....it is nothing short of amazing....
>That Sybase are not selling it like hotcakes is a sad indictment
of our
>industry and how so many people think the only answer to any
database
>question is 'Oracle' having never experienced anything else.
>If only a small percentage of people buying EDW databases showed
a small
>modicum of common sense Sybase would have a hit on their
hands....
>alas,
>they don't.
>I can personally assure and guarantee the following in comparing
IQ and
>Oracle having used both....
>1. IQ will use 20% of the disk of oracle for the SAME input data
for the
>SAME target model.
>2. IQ will run between 10 and 100 times faster than the SAME
QUERY on th
>e
>SAME database design as Oracle on the SAME hardware. IQ will
NEVER run
>slower.....
>3. Total system cost for hardware and software for similar
performance i
>s
>about 50% when using IQ vs Oracle. That is an 8CPU oracle
system will
>easily be outperformed by a 4 CPU IQ system but probably not by
a 2 CPU IQ
>system......
>4. If you spend the SAME amount of money on hardware (CPUs
because you o
>nly
>need to buy 20% of the disk) and database license....IQ is
priced
>in parity
>with Oracle on a CPU bases...at least at list price with means
nothi
>ng any
>more....you can reasonably expect performance to be 10-100 times
f
>aster for
>every query you run.
>Like I said...at half the price for similar performance and
10-100 t
>imes
>faster for the same money you might wonder how IT people keep
buying Oracl
>as an EDW database....
>I don't wonder...it's because people who buy oracle as an EDW
databa
>se seem
>to have extremely limited experience of other database
technologies...
>they
>have an 'Oracle hammer' and all the world is a nail.....sigh..
>.
>Having said that...of course all databases have their 'features'
and
> IQ has
>it's share...but once you know what you are doing it is
something of
> a joy
>to work with.....
>Best Regards
>Peter
>Original Message:
>-----------------
>From: martin.willcox via dw-select
dw-select@Groups.ITtoolbox.com
>Date: Sat, 12 Aug 2006 08:59:43 GMT
>To: peter@peternolan.com
>Subject: RE:[dw-select] Re: RE: Teradata Vs Oracle
>Don't feel too bad. When you're not fighting fires and
managing
>tablespace, you get to hang out with users and design cool new
BI
>applications instead... :-)
>MW
>---------------Original Message---------------
>>> When working on a replacement of Oracle with Teradata on a
>>>multi-terabyte implementation we were able to cut the 29 hour
daily load
>>>to 4 hours,
>>
>>That is impressive!
>>>cut the DBA staff from 5 FTEs to 1,
>>That is depressive!
>>and improved the performance of all queries that were being run
- without
>>exception
>>(including the data that remained in a Star Schema).
>>That is too strong. Maybe the ORACLE configuration was
obsolete?
>>Kalman Toth, Database, Data Warehouse & BI Architect
>>URL:
Best Pra
>ctices
>>----- Original Message -----
>>From: "jeffpankow via dw-select"
<dw-select@Groups.ITtoolbox.com>
>>To: "SQLUSA" <technicalsqlusa@earthlink.net>
>>Sent: Friday, August 11, 2006 11:30 AM
>>Subject: RE: [dw-select] RE: Teradata Vs Oracle
>>>
>>> Some comments:
>>> Teradata could not have been designed for Data Warehousing,
because the
>>> concept of Data Warehousing grew up after the creation of the
first
>>> Teradata system (oringinally called the DBC or Data Base
Computer).
>>> Teradata was originally designed for and connected only to
mainframe
>>> systems and had a size limit of 1 Terabyte. A great deal has
changed
>in
>>> the RDMBS since its early days, and most of it has been done
to allow
>>> the system to support the Data Warehouse workload.
>>> Many things have not changed - the base table structure and
basic data
>>> access, the YNET intelligent interconnect concept (now known
as the
>>> BYNET), the ability to change query resource utilization
on-the-fly; al
>l
>>> built upon a RDBMS that pretty much manages itself (often a 3
to 1 or
>>> greater Oracle DBA vs Teradata DBA requirement).
>>> But many things have changed - the continual evolution of the
optimizer
>,
>>> unique index structures and access strategies to support the
data
>>> warehouse workload (including star schemas), new connectivity
and load
>>> tools to support all popular operating systems. For example,
an new a
>nd
>>> unique on-the-fly hash join consruct has been implemented in
the 6.1
>>> release of Teradata specifically to assist in the
optimization of
>>> StarSchema database queries.
>>> To your comment: Oracle is clearly the leader over the other
three in
>>> terms of functionality and features relating to data
warehousing. Thi
>>> might be the case, but Oracle falls VERY short when it comes
to
>>> performance. When working on a replacement of Oracle with
Teradata on
> a
>>> multi-terabyte implementation we were able to cut the 29 hour
daily loa
>d
>>> to 4 hours, cut the DBA staff from 5 FTEs to 1, and improved
the
>>> performance of all queries that were being run - without
exception
>>> (including the data that remained in a Star Schema).
>>> As far as Sybase IQ is concerned, I do not have any
experience with
>>> them, so I cannot speak to their product. The white paper
being
>>> discussed in other threads can be aquired at:
>>> FYI - ALL Teradata documentation can be aquired at:
>>> Thanks for your time,
>>> Jeff Pankow
>>> -----Original Message-----
>>> From: ngalemm... via dw-select
>ox.com]
>>> Sent: Thursday, August 10, 2006 1:27 PM
>>> To: Pankow, Jeff
>>> Subject: RE:[dw-select] RE: Teradata Vs Oracle
>>> Just some comments... Sybase IQ, Teradata & Oracle are very
diff
>erent
>>> animals. Of the three, the ONLY one designed solely for
BI/Data
>>> Warehousing applications is Sybase IQ.
>>> Oracle is middle of the road and works well in both BI and
OLTP
>>> applications.
>>> Teradata was originally intended and sold as a solution for
very high
>>> throughput OLTP applications. That is what its architecture
does be
>st.
>>> However, in the early 90's they pushed it as a data
warehouse engine
>>> because of its ability to handle very large databases (and
>y
>>> could sell more
>>> hardware). The main problem with Teradata is that, because
of
>>> its architecture, the only way you can realize the levels
>>> performance it's hardware promises is to implement using a
3NF
>>> relational model. Queries against dimensional models simply
do not
>>> work very well in a MPP shared-nothing hardware
architecure.
>>> So, it gets down to what kind of data warehouse is being
created.
>>> If it is a dimensional design, do not use Teradata... you
will
>not be
>>> able to make effective use of the hardware. You would be
far better
>>> off using Oracle RAC on commodity Linux servers.
>>> Comparing Oracle and Sybase IQ is another matter. Sybase IQ
is true
>ly
>>> unique. Data is stored as inverted bitmap indexes allowing
for a hi
>gh
>>> degree of data compression and very fast queries. The
downside is t
>hat
>>> loading and updating data is relatively slow because of the
need to
>>> maintain the bitmaps.
>>> But it is not slow enough that it precludes its use in 80%
of data
>>> warehouses.
>>> Oracle is a traditional relational database comparable to
DB2,
>>> SQLServer and Sybase Adaptive Server. However, Oracle is
clearly th
>>> leader over the other three in terms of functionality and
features
>>> relating to data warehousing. Its partitioning and
parallism featur
>es
>>> allow for very large databases with heavy query loads.
Support for
>>> bitmap indexes and star joins provides very good query
performance
>>> against dimensional schemas. They have continually expanded
>>> features of its engine to include embedded statistical
capabilities,
>>> XML, market basket analysis and other features.
>>> It may not be as good as IQ in some areas, but overall is a
better
>>> well-rounded choice.
>>> ---------------Original Message---------------
>>> >In the past two years I've developed a data warehouse in
Oracle
>>> then one
>>> >in Sybase IQ (which is similar to Teradata in that it is
>>> designed for
>>> >data warehousing). The major differences are that in Sybase
IQ I
>>> didn't
>>> >need to create measure table or a star schema to get the
benefit
>>> of all
>>> >the data. Due to IQ's indexing and storage being different
from
>>> the
>>> >table structure that Oracle demands what took hours in
oracle
>>> took
>>> >seconds in Sybase IQ.
>>> >
>>> >Oracle is designed for inserting, updating and deleting.
Sybase
>>> IQ is
>>> >designed for bulk loading and querying.
>>> >The only current difference between Teradata and Sybase IQ
is
>>> the cost.
>>> >Good luck.
>>> >-----Original Message-----
>>> >From: manjiri.kulkarni via dw-select
>>> >[
>>> >Sent: August 8, 2006 7:47 AM
>>> >To: Petrie, Susan
>>> >Subject: [dw-select] Teradata Vs Oracle
>>> >In our company, we are trying to establish parameters for
short
>>> listing
>>> >Data
>>> >Warehouse platform and trying to evaluate Teradata Vs
Oracle.
>>> Would like
>>> >to
>>> >hear some experiences from people who are already working
on
>>> these
>>> >platforms.