17.txt

This complete matlab for neural network

发信人: fervvac (高远), 信区: DataMining
标  题: Re: 国内有没有人做过Olap analysis engine?
发信站: 南京大学小百合站 (Wed Mar 20 12:46:37 2002), 站内信件

I've got your motivation. Here are some rough thoughts that came to my 
mind immediately:
1) Why not leverage on existing OLAP engines? You mentioned the customers
   need to buy a RDBMS. If they use Windows and buy MS SQL Server, the OLAP
   part is already integrated inside.
   Of course, this argument can be easily defeated, :p

2) ROLAP engines usually employs special technologies to accelerate query 
   processing, such as bitmap index, star-join, etc., apart from data cube and
    materialized view. Efficiency will be compromised if those parts are
   missing or done exterior to the DBMS kernel.
   Of course, if the data volume is small, that's not a big deal.

3) So I guess it might be practical only if the data is not big and customers
   are poor (, :p). My suggestion is that you might need to do some
   proessing on the client side to avoid too much data transfer over the
   network.

【 在 sn (小龙) 的大作中提到: 】
: I am in a corporation, my aim is to build a system that can be used in my 
: project implementtation, you know that Olap server is expensive, and I often 
: do some small project, the customer don't want to spend that much, but they 
: want to own the olap analysis. 
: So I want to do an olap analysis engine , which I will store  meta data in re
: lational
: database, and access data over OLE DB interface, and I don't want to organize
:  data
: myself, it 's the matter of relational db, and of course customer should buy 
: a relational database at least.
: Do you think it feasible?
: 【 在 fervvac (高远) 的大作中提到: 】
: : His two sentences covers two architectural choices for an OLAP engine: the
: : ROLAP and MOLAP approaches.
: : By storing the fact table or the cube in MOLAP format, it means data are 
: : treated as multidimensional points and the table (relation) is treated as
: : multidimensional array.
: : Both ROLAP and MOLAP implementation have some trickly places and not that 
: : straightforward, :p. For example, I am still wondering if there are any
: : tricks to enable the MOLAP approach to scale up, for example, 10 dim with..
: (以下引言省略 ... ...)

--
※ 来源:．南京大学小百合站 bbs.nju.edu.cn．[FROM: 饮水思源BBS]