Conference 7.286::digital

             <<< VAXAXP::NOTES$:[NOTES$LIBRARY]ALPHANOTES.NOTE;1 >>>
           -< Alpha Support Conference - Digital Internal Use Only >-
================================================================================
Note 23.0                       What About VUPs?                      No replies
PERFOM::HENNING                                      13 lines  20-JAN-1997 04:59
--------------------------------------------------------------------------------
    What About VUPs?
    
    One of the most frequently asked questions to our performance
    group is "How can I compare the newest system to my 10-year-old
    system"?
    
    A partial answer may be found at:
    
    	http://tlg-www.zko.dec.com/~henning/compare/compare.html
    
    This note has been cross-posted to several conferences.  The immediate
    stimulus for the webpage was DEChips note 630 and ALPHANOTES (1997)
    note 18.

    I was told that this was impossible and valueless.

    impossible, valueless, and frequently requested.  So the graphs
    referenced in .1 try to give some rough idea, not a precise comparison.

    In a pinch, use the "VUPS" numbers in DECPS PA so you can get on to the
    real work.  Be sure to include YMMV.
    
    FJP
    

    How about something more "real" like TPC benchmarks, which I believe
    have been done for both?
    
    (Okay, maybe not "real", but closer to something like a real situation)

    Most of the VAX TPC results used a different TPC benchmark than the
    current Alpha systems so there's still the apples and oranges
    comparison. You can't compare TPC-B to TPC-C to TPC-D.
    
    perfom::csg_reports: does have some older alpha-vax comparisons for the VAX
    7000 -> AlphaServer 7000 vis:
    
    ALPHA_AXP_VAX_COMPARISON.PS;6
    ALPHA_VAX_COMPARISON_940223.PS;3
    ALPHA_VAX_COMPARISON_940414.PS;1       
    ALPHA_VAX_COMPARISON_941221.PS;2
    
    Bottom line is anyone moving from a VAX 6000 class or lesser machine
    is going to see a big CPU performance improvement in going to any of
    the current generation of AlphaServer systems.  On top of that, many
    customers may also find they drop down a class from a licensing
    perspective ie VAX 6000 -> AlphaServer 4100, VAX 4500 to AlphaServer
    1000A.
    
    	Owen.
    

    listen to Henning. he knows this stuff.

    Thanks for the pointers. I will go through them and see what there is.
    
    Frank

    I suggest that you look at DIGITAL's capacity planning technology,
    this will provide you, and your customer the tools to plan how 
    performance will be gained for their environment.  We have
    two technologies, a Capacity Planner for Windows, provides
    quick sizing of systems and networks as well as the Enterprise
    Edition, Capacity Planner for UNIX/OpenVMS, which collects performance
    data from real workload, characterizes by user types, then build
    models where "what-if" analysis can be made.  Check out the internal
    web site at http://qnut.zko.dec.com
    
    steve
     

    The graph and article have been updated per various suggestions, and
    posted in a customer-accessible location:
    
    http://www.digital.com/alphaserver/performance/vups_297.html

    
    We have a solution!
    
    At the request of SBU Marketing in Europe, much work has been done by
    the Corporate Servers Performance Group (Lee Allison) to address this 
    very problem of comparing old to new.  The results are posted in the 
    Installed Base Intiative web site.  Please look in:
    
    http://sbu.mro.dec.com/ibi/	then select tools and resources.
    The comparison tables are called "SYSTEM PERFORMANCE DATA".  There are
    some updates to come on the 4100 and 8x00, so check back again in a
    while.
    
    Work is underway by David Gitner in the Corporate IBI team to produce a 
    pocket guide to be distributed to the whole field and partners.
    
    -Ken 
                                                   

    I just jumped over to the web site listed in .-1 and I am impressed by
    the completeness of the list.  I do have a few concerns...
    
    1. There is a very 'lite' definition of the tests that were run on the
    systems.  One could easily 'ASSUME' that these were official TPC tests. 
    There should be a strong statement that these numbers do not in any
    way, shape or form correspond or relate to any TPC numbers.  In fact,
    there is plenty of white space to the right of the table to reinforce
    this fact.
    
    2. There really should be a description of the hardware used so that
    accurate comparisions can be made.
    
    Regards,
    
    Jeff

    This information is also available for our Digital Business Partners on
    the web at
    	http://www.partner.digital.com/sbu/ibi
    
    It will also be published in the May 97 edition of rht Digital System
    Enhancement Guide.
    
    As far as feedback/comments on the information or any other aspect of
    the IBI site we welcome it!!!  Please use the "Feedback" button on the
    bottom of the web pages to send your comments.  This is the fastest way
    to get your feedback to the proper people.  Or, send comments to
    WHYNOW::FEEDBACK
    
    Bob Newman
    IBI Webmaster

    nit:
    
    > Corporate Servers Performance Group (Lee Allison) 
    
    Lee sits down the hall from me.  We are both members of the CSD
    Performance Group.  Lee was one of the reviewers for the VUPs article
    (which IBI is welcome to also add to their site if they wish).  Lee is
    out of town right now, but I'm sure she would agree with the
    sentiment that estimated tps must be clearly stated to be estimates and
    not in any way imply that they can be compared to offical TPC results.
    
    You won't get details of hardware config for these estimates.  They are
    based on Lee's (highly experienced) wetware estimates.