Thursday
Nov222007
Why not Cache from Intersystems?
Thursday, November 22, 2007 at 9:05AM
I have some experience with a very large OLTP system that is 7+ TB in size and performs very well for 30K+ concurrent users. It is built using Intersystems Cache based on the very old but very scalable MUMPS platform. Why don't I see more discussions about archiectures such as these in this forum? I am curious why this platform scales so much better then the typical RDBMS.
Reader Comments (3)
> Why don't I see more discussions about archiectures such as these in this forum?
I just haven't yet run across any websites that use this solution. Please email me and then maybe we can do a profile?
thanx
Because it doesn't line up with LAMP architecture? It's a proprietary solution that isn't supported by the skill sets of most practitioners. One of the factors in any architecture decision has to be: do we have the skills to use this? If not, is the value of it sufficiently high to pay to learn it? The latter question is a hard barrier to get over, particularly if you have mainstream skills and the technology is fairly inaccessible or proprietary. And how about people who know how to troubleshoot or support it?
The difference between having a deep skill set and a shallow skill set can be the difference between success and failure. Any project can be done in any pretty much any technology (Turing) but doing it well and quickly is life and death. Also, the inability to find people who know an unusual skill or who are willing to sidetrack their careers, and resumes, can kill a project in even the greatest of new technologies. Sometimes it's just better to go with what you know.
Then there's cost. What can beat MySQL for cost?
I've been wondering why ODBMSs haven't been discussed more as well. It's funny. You have all these different frameworks (Active Record, Hibernate, etc) that fake an ODBMS through a persistance layer. Why not just use an ODBMS?
Having been working in an IT department driven by Accounting (CFO/CTO combined into one position), Reporting is what drives all of this. Specifically, user generated reports through Crystal reports, Excell, etc. There's a misconception that you have an ODBMS or an RDBMS and that you'll experience a heavy hit when using the RDBMS facade of an ODBMS (Perception, no idea if it's true). That, and users still think of data as flat and, when you think of how data is pulled from an RDBMS, it is flat.
ODBMSs add a true heirarchy to the representation.