no performance improvement seen after costly hardware update

In my company GIS applications are very important. To improve performance some 6000 Euro have been spent on a new PC  (2x4 core XEONS with 12 MB L2 Cache, Raid Controller, fast memory, etc.). Theoretically the new PC should be at least twice as powerful as a 2 years old PC which it replaces as the major workhorse. Comparative tests have shown ZERO improvement in execution times. The tests done so far involve extensive floating point calculations and DB I/O (calculating the area of parcels, creating shapes for some 4000 polygons and doing the I/O for ORACLE). DB connection is done via ORACLE 11g2 (with latest patches). OS is XP 32 prof. SP3. Microstaton XM in the latest download version is used.

Performace has always been an issue and on older PCs of the pentium class the programms have run for hours on end (10 and more hours). Upgrading hardware on an almost yearly basis has always brought down execution times but not this time. This  last and most costly hardware update done right now has shown no improvement at all which is very, very frustrating.

The programms in question have been developed over almost 10 years and the company who is doing this says it would be difficult to isolate the relevant code so that it could be testet meaningfully in another environment.

Any ideas, suggestions or explanations are welcome. It must be also of concern to Bentley if Microstation does not seem to make use of advancements in hardware. I had great hopes to cut down the many 3 Minutes plus waiting times which cannot be used otherwise. These hopes has been crushed so far.

Regards

Erwin

Parents
  • One Process: Multiple Processors

    Erwin:

    In my company GIS applications are very important. To improve performance some 6000 Euro have been spent on a new PC (2x4 core XEONS with 12 MB L2 Cache, Raid Controller, fast memory, etc.). Theoretically the new PC should be at least twice as powerful as a 2 years old PC which it replaces. Comparative tests have shown ZERO improvement in execution times.

    Why should your new computer be at least twice as powerful as a 2 years old PC?

    • How does the clock speed compare?
    • Was floating-point performance a limiting factor with the older PC?
    • Was memory performance a limiting factor with the older PC?
    • Was disk performance a limiting factor with the older PC?
    • Was I/O performance a limiting factor with the older PC?
    • Was Oracle performance a limiting factor with the older PC?
    Which of those measurements led you to believe that your new computer would provide more performance? How did you justify spending 6000 Euro?

     

    Erwin:

    The tests done so far involve extensive floating point calculations and DB I/O (calculating the area of parcels, creating shapes for some 4000 polygons and doing the I/O for ORACLE). DB connection is done via ORACLE 11g2 (with latest patches). OS is XP 32 prof. SP3. Microstaton XM in the latest download version is used.

    4000 polygons doesn't seem like a lot of work, but database I/O would seem a possible bottleneck. I/O is mostly integer, so floating point performance may be irrelevant.

    Is Oracle running on the same computer as MicroStation, or via a network connection? If it's on the same computer you should set the Windows processor affinity so they are running on different cores.

    MicroStation, like most desktop applications, is mostly single-threaded (except for 3D rendering, which is not what you are doing). That single thread runs on a single processor. Of your 8 cores, 7 are not being used by MicroStation, and never will be.

    Erwin:

    The programs in question have been developed over almost 10 years and the company who is doing this says it would be difficult to isolate the relevant code so that it could be tested meaningfully in another environment.

    Are the programs in question running in Windows or on MicroStation? Can't the developer run a performance profile analysis on their code? There are tools for Windows that show processor, memory, I/O performance per process.

    Regards, Jon Summers
    LA Solutions

     
    Regards, Jon Summers
    LA Solutions

Reply
  • One Process: Multiple Processors

    Erwin:

    In my company GIS applications are very important. To improve performance some 6000 Euro have been spent on a new PC (2x4 core XEONS with 12 MB L2 Cache, Raid Controller, fast memory, etc.). Theoretically the new PC should be at least twice as powerful as a 2 years old PC which it replaces. Comparative tests have shown ZERO improvement in execution times.

    Why should your new computer be at least twice as powerful as a 2 years old PC?

    • How does the clock speed compare?
    • Was floating-point performance a limiting factor with the older PC?
    • Was memory performance a limiting factor with the older PC?
    • Was disk performance a limiting factor with the older PC?
    • Was I/O performance a limiting factor with the older PC?
    • Was Oracle performance a limiting factor with the older PC?
    Which of those measurements led you to believe that your new computer would provide more performance? How did you justify spending 6000 Euro?

     

    Erwin:

    The tests done so far involve extensive floating point calculations and DB I/O (calculating the area of parcels, creating shapes for some 4000 polygons and doing the I/O for ORACLE). DB connection is done via ORACLE 11g2 (with latest patches). OS is XP 32 prof. SP3. Microstaton XM in the latest download version is used.

    4000 polygons doesn't seem like a lot of work, but database I/O would seem a possible bottleneck. I/O is mostly integer, so floating point performance may be irrelevant.

    Is Oracle running on the same computer as MicroStation, or via a network connection? If it's on the same computer you should set the Windows processor affinity so they are running on different cores.

    MicroStation, like most desktop applications, is mostly single-threaded (except for 3D rendering, which is not what you are doing). That single thread runs on a single processor. Of your 8 cores, 7 are not being used by MicroStation, and never will be.

    Erwin:

    The programs in question have been developed over almost 10 years and the company who is doing this says it would be difficult to isolate the relevant code so that it could be tested meaningfully in another environment.

    Are the programs in question running in Windows or on MicroStation? Can't the developer run a performance profile analysis on their code? There are tools for Windows that show processor, memory, I/O performance per process.

    Regards, Jon Summers
    LA Solutions

     
    Regards, Jon Summers
    LA Solutions

Children
No Data