OLAP

  • Hi,

    it seems to me that OLAP/DTS is probably the most interesting technical area for SQL DBAs to get involved in - if you can! Unfortunately in my company they have no current plans to implement it, although there is a constant need for standard database design and stored proc scripting. Is this a common state of affairs - or are all you other DBAs having much more technically challenging fun than me?

    Paul Ibison PhD, MCSD

    email: Paul.Ibison@btinternet.com


    Paul Ibison
    Paul.Ibison@replicationanswers.com

  • Paul - Im not sold on DTS, even though I use it a lot for moving data around. I think at some point you hit the wall and have to write code - which accounts for the majority of our DTS questions - and once you write code, I'm not sure it's the right environment. So far I don't use it for much besides scheduled moving of data, that kind of thing. Have been experimenting with writing COM add in's to see how that would work.

    OLAP, Data mining, so far almost no use. I count it as my weakest area. Only way to improve is to find a project, haven't had the time to find a project or do it either!

    Right now I'm just finishing an application for another part of our IS team, developers were overbooked and I had time, so why not? Other than that, I'm replicating about 230 dbs, which makes for great fun when you have to apply schema changes to all 230. But no, nothing REALLY exciting.

    Andy

  • Thanks for the interesting feedback - it sounds as though SQL Server OLAP isn't as widespread as I feared. I have come to the same conclusion regarding learning OLAP, and have installed my work databases at home to implement a data warehouse, investigate scenarios etc.

    Paul Ibison PhD, MCSD

    email: Paul.Ibison@btinternet.com


    Paul Ibison
    Paul.Ibison@replicationanswers.com

  • Agree about DTS (twice in one day - I must be mellowing).

    People seem to see transfer and automatically jump into it and get into trouble.

    I think of a transfer as move the data onto your server into staging tables (by any means) then run stored procedures to deal with it. The transfer could be DTS but I would keep it simple (prefer bulk insert if it needs to be fast - VB if not or you want to validate). In this way your processing is not bound to the source data - you can change the data source or transfer method without affecting anything else (maybe).

    OLAP - again gets used where it's not really needed (was on a project where a large management consultancy A... and a huge software company M....S... tried to suggest an OLAP solution for a maximum 10,000 transaction database). I've never used it apart from investigation and helping people out.


    Cursors never.
    DTS - only when needed and never to control.

  • In my environment we have only had a couple of projects employ OLAP. We have hundreds of projects that do not. I haven't had the pleasure of working on either OLAP project.

    A new opportunity for myself appears to be making sense out of NT/W2K/SQL peformance metrics. The S2K resource kit in the performance monitoring section suggests that you may need to use OLAP to make sense of it all.

    Now if I can just get some results before the next batch of crises descends on me! Part of handling the "my app is slow today" battle is having good baseline data so I will get there eventually.

  • DTS - wasn't sold, but use it more and more. I like the chaining features of different tasks, though it's mainly for simple tasks, like auto generating excel, my push ftp, doing filesystemobject type tasks, etc. It's quick and dirty, and in 2000, very reliable, imho. Of course, for mission critical stuff, I'd agree with Andy and probabl write code. Not sure if I'd use VB/DTS or .NET, probably depend on what it was.

    OLAP - Took an online course (interesting) and learned quite a bit, but working for a strtup, there isn't enough data to use it for. I think it really would help in places where there is lots of data and you are looking for patterns. I started a project here, but it's an on my time deal, so not much support. Two places where I think it would have worked:

    1. two jobs ago I worked for an importer. They sold hundreds of different items (raw materials) each day and had shipping considerations all over the US as well as planning for the imports from all over the world. I think OLAP would have helped the purchasing people as well as the shipping people in avraging the load they needed and smoothing out the bumps in the supply chain. When I left, they were still "winging" it.

    2. Last job - Financial company that managed assets for different people. There was lots of data on trading activity and the "types" of things people did with their accounts (look up balances, check prices, etc.) We could have used this to "push" products (marketing) as well as anticipate better the load for different areas of the system.

    Just my 2 cents. I think there are places it will work, but most of us are still learning and aren't sure where it helps.

    Steve Jones

    steve@dkranch.net

  • hi,

    I used OLAP some time ago, but not in SQLServer. the product was bought by Oracle a few years back. I worked in a Direct Marketing Co, and we used it to identify buying trends, customer retentions etc. Also was useful to identify liklihood of bad debters based on demographics, product mix etc. It was used together with a stats package - but as I say, this was about 5 years ago.

    I enjoyed using it, but alas haven't had the opportunity since.

  • Hello to all, I am not a big fan of DTS to me it is useful for small tasks that don't require much effort, however if you are trying to move millions of records i rather have a DBlink and develop a stored procedure that allows me to have complete control over it.

    I recently became in contact with OLAP, in fact i am still learning about it , so far, i find it to be a use full tool helps improve the reporting time and in a call center environment where you get a record fort every call you receive, I can say it is quite useful to improve processing time, how ever it still depends on a relational database to full fill its purpose and as a data lover i am, i am not very fond of having to generate a second database (the OLAP database) to build a cube, to me is replicating information, and when you are talking about a 5,000,000 records and that is just on one table, this is madness, sure, you don't put al records on the cube you summarize and all but still you are still taking lots of space on the server (our relational database is over 200 gig), that in some cases you have to share with the relational db to save expenses, it sounds to me like a big investment in time, money and effort just to make you're reports faster i agree the functionality's that it offers are great, however there might be room for improvement on the data side, why not generate beta index technology or tridimensional tables, any ways that is just my opinion.

    regards

    Charly

    P.S NO'p my name is not badly written.

    "Everything in life happens for a reason and helps you become what you are today"

Viewing 8 posts - 1 through 7 (of 7 total)

You must be logged in to reply to this topic. Login to reply