mainframe DW conversion to SQL svr

[sweetleaf]

Hi,

My company has a data warehouse thats about .5 terabytes in size with over 300 tables. Its currently in db2 on the mainframe. We're thinking of converting it to MS SQL Server. But I'm not really sold on what benefits this brings. Does anyone have have ideas or specific examples in support of converting?

Thanks

 

[johnherman]

I suspect someone in your company has done a ROI or Cost-Benefit analysis as to why to spend the time and money to move to SQL Server. If not, your company must be making loads of money since it is spending without sound financial reasoning.

Some reasons why it might be considered:
performance, access to better analysis tools, cost, ease of use and maintenance, corporate strategy, mainframe obsolescence, etc.

-------------------------
The trouble with doing something right the first time is that nobody appreciates how difficult it was - Steven Wright

 

[Hans63]

John has identified several reasons why a migration could be considered. Scalablity might be another one.

I have encountered articles about several occasions where a DWH was migrated from a mainframe to mini computers (linux/unix). This was done because of the cost benefit + scalability.
I have not yet heard of a migration of a DWH to SQL server.

Oh yes. It might of course also be that the sales people of MS did a good job

 

[sweetleaf]

Thanks. theres two types of cost associated with this type of study. Unit cost is about charging clients back for such things as utilization, consumption and processing. And in a server environment its always harder to connect the dots between processes and to get an accurate read of such things as who is consuming and utilizing what processes (eg, in order to track for chargeback purposes).

The move would, in this case, be helped by things additional to a cba based on hard dollars. say, ease of sharing data with other server based databases throughout the company etc. I'm wondering if anyone here has experience with or has attempted any kind of formula to help quanitify these less than tangible reasons.

John raises good points:
1)Performance - I'm not sure I understand how?
2)Access to better analysis tools - agreed
3)cost - per above
4)Ease of use and maintenance - depends on in house skills mix (the job market seems to have richness in both)
5)Corporate strategy - would like to know what some of the chief drivers for this would be?
6)Mainframe obsolescence - ibm keeps coming out with new processing engines for the mainframe and facilities to run linux and oracle among other dbms' on it

I wonder how banks who are riddled with mainframes tend to resolve this?

Thanks

 

[johnherman]

Performance - a server and database such as SQL Server might be able to be better tuned to perform with Data Warehouse structures such as cubes and star schemas, thus getting better performance that a mainframe. Or, SQL SErver on Windows/Intel might be flat out faster than Oracle or DB2 on whatever mainframe OS.

Corporate strategy - Examples: 1. We plan to grow by acquisition so plan on having many disparate data sources over the next few years. 2. We plan to consolidate all computer operations in one location over the next 5 years. We plan to outsource ATM and ATM reconciliation. 3. We plan to offer Data Warehouse/BI services to our customers and help them cross-sell their services to other customers of ours who are not familiar with or are not using their services. 4. We plan to grow overseas and thus prefer to stay with open source databases and operating systems. 5. We expect the government to apply new restrictions on data privacy and security and believe we can save money by going with a Microsoft solution (or mainframe solution). You get the idea.

With cost, remember that there are initial costs (purchase, start-up, conversion, training) and then ongoing/recurring costs (license renewals/support contracts, staff salaries, etc).

-------------------------
The trouble with doing something right the first time is that nobody appreciates how difficult it was - Steven Wright

 

[sweetleaf]

Good points to think about, thanks John.

I am particularly interested in and will see what information I can get on your point that SQL server may be more tuned to star schemas, cubes etc...