SAS's Goodnight won't say goodbye yet
Published: 23 May 2006 16:05 BST
A lot of people don't see the difference between business process management (BPM) and business intelligence (BI) software. What's your take on the difference?
In the past, BI has represented just the creation of reports. BPM has been a dashboard of KPIs (key performance indicators).We like to use the BPM as a sort of overview of all the KPIs and then we let you drill down and go into the BI reports. They really belong together. They are still two different pieces of software but that ability to start with the performance monitoring and then being able to drill down is the way to do it.
I share my dashboard with most of the managers in the company, so when we are doing really well and the revenues are really good they know when to come over and ask. The company knows what profitability and revenue levels we are shooting for and they can try and do their share.
Can you talk about the next release of SAS?
The next release is System 9.2, set for some time around March next year. We set development targets but we can't ship the product until it is finished. It has to be ready. We don't ship until the bugs are out.
We are doing a couple of things. One is "cascading prompts". That requires a lot of changes in a lot of software. The major releases are when we try and get all of the features that we can into the basic system. Multi-threading is part of that too. This is becoming more prevalent with dual-core and later this year, quad-core processors.
I was at a meeting when Scott McNealy pulled a chip right out of his pocket and said, "Jim, guess how many cores are in there?" Not four, not eight -- 16 CPUs, right there on that one chip. So we have that kind of power, but we won't be able to take advantage of it.
SAS, like all other programs, is written so that only one processor is working. Like all others they are written sequentially, so the software does one block and then moves to the next block.
How do you take an existing program that is written sequentially like that and spread it over multiple processors? Well, one way is you can break up the input into four different input streams and let each processor do exactly the same thing to each one of those input streams. Another way is pipelining. Take those same sequential blocks of code and get them to one thread. We can distribute the blocks but they still have to execute sequentially.
Full Talkback thread
3 comments
-
Having several years experience in SAS, been invol... Pascal Guignard -
I fully share your opinion Pascal, we are exa... Etienne Baert -
Forget to add this in my post:
It becom... Etienne Baert








