IDS Forum

Re: Data warehouse: RedBrick vs IDS

Posted By: FREDERICK HO
Date: Friday, 25 July 2014, at 10:43 a.m.

In Response To: Data warehouse: RedBrick vs IDS (Kern Doe)

Hi,

Prior to IDS 11.7, the answer would be Red Brick, hands down. That's
simply because IDS was not designed for data warehousing, like the fact
that Red Brick was never designed for OLTP.

Red Brick was the first commercially available data warehouse product
on the market. The Red Brick company was originally founded by Ralph
Kimball, who invented the concept of dimensional modeling, which remains
today the de facto standard for designing data warehouses. The key
features that Red Brick has are:

- Star schema support throughout, i.e. Star Index, Star Join (using the
zigzag method)

- Target and bitmap indexes

- One pass loader, i.e. tables and indexes are ready to use after data
loading - no need for deferred index build

- Pre-computed view (i.e. MQT in DB2) aka. summary tables with advisor
based on run-time stats (vs. schema based advisor)

- Versioning, i.e. trickle feed, for real-time warehousing

- Rule-based Optimizer (vs. cost-based optimizer in IDS)

- Built-in OLAP functions (prior to SQL-OLAP becoming a standard)

All these features provide superb performance "out of the box" with
very simple setup, e.g. a single Star index on the fact table would
provide very adequate performance.

Starting with IDS 11.7, a number of the key features listed above were
implemented in IDS (using some code from XPS, a MPP-based data warehousing
product developed mostly in Portland, as part of Informix). Since then,
IDS has comparable performance for queries requiring star-join, bit-map
indexing (in the form of multi-index scan, though indexes are not
materialized). However, IDS still does not have summary tables, versioning
or a one-pass loader.

Starting with 11.7xC3, you may be aware of a product called Informix
Warehouse Accelerator (IWA). I won't repeat all the features that come
with this product, as there are plenty of articles that describe this
product (or read the blog listed below). Suffice to say that IWA is built
using "modern" technology in that it is an entirely in-memory database
that will outperform standard disk-based systems by over 100x.

The fundamental difference between the two approaches is that Red Brick
was built during a period where "smart indexing" with Star Schema was key
to its simplicity and performance. (Red Brick sales folks would routinely
tell an Oracle shop that Red Brick is 10x over Oracle in performance and
proceeds to prove it to them). Technologies like IWA use technologies
(i.e. compression, multi-core cpus and SIMD instructions) to scan data and
do predicate evaluation very efficiently. The combination of IDS for OLTP
data and transparent acceleration of warehouse/analytic queries makes for
a very compelling combination (not to mention other IDS features like HA,
etc).

If you need more info, feel free to contact me or the IDS team
directly.

Fred Ho

Program Director

Chief Technologist - Informix Warehouse

IBM Corporation

Blog:
https://www.ibm.com/developerworks/mydeveloperworks/blogs/fredho66/?lang=en

Tel. 408-956-2422

Tie Line - 475-2422

email: hof@us.ibm.com

Messages In This Thread

• Data warehouse: RedBrick vs IDS
  Kern Doe -- Thursday, 24 July 2014, at 1:45 p.m.
  • Re: Data warehouse: RedBrick vs IDS
    FREDERICK HO -- Friday, 25 July 2014, at 10:43 a.m.
    • Re: Data warehouse: RedBrick vs IDS
      Kern Do -- Friday, 25 July 2014, at 10:55 a.m.

IDS Forum is maintained by Administrator with WebBBS 5.12.