Re: successful migration from client/server to 3 tier systems ?

Casey Green wrote:

> Microsoft has been posting some phenomenal numbers
> (like 5676 tpm-c under a simulated load of 5000 users) without
> the aid of a TP monitor. It seems to me that the claim that
> 2-tier doesn't scale is starting to look awfully thin.

Then Bruce Tobin said:

> Nowadays consultants and other assorted gurus are routinely specifying
> 3-tier architectures for projects that will never get within shouting
> distance of these numbers. The last four projects I've worked on have all
> been three tier, and all but one are never expected to be deployed to more
> than 150 users. Nevertheless, the most commonly cited reason for going
> three-tier is that "two-tier doesn't scale." Well, I think it's pretty
> clear from results like these that 2-tier scales pretty well.

Lemme see if I can say this without starting The Great 2-Tier vs 3-Tier Definitional Flamefest...

From a performance point of view, TPC-C configurations share an important 3-tier characteristic: the distribution of client load across "front end" processors. This is true even though TPC-C doesn't define a full 3T logic. For that matter, even some 1-tier (host/mainframe) environments share this offloading characteristic. While they may not be your cup of tea, MVS mainframes scale darn well, thank you very much!

There is no necessary or general correlation between the number of architectual tiers and delivered performance. IMHO, blanket devotion to any of 1T, 2T, or 3T designs not only misses the point, it misses all important points. It's no more sensible than "Blue circles rule! Red triangles stink!"

-- 
Jonathan Eunice
Analyst, client/SERVER Companion