[Search for users] [Overall Top Noters] [List of all Conferences] [Download this site]

Conference ssdevo::hsz40_product

Title:HSZ40 Product Conference
Moderator:SSDEVO::EDMONDS
Created:Mon Apr 11 1994
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:902
Total number of notes:3319

771.0. "HSZ50 - Maximum_cached_transfer setup" by VAXRIO::LEANDRO () Wed Feb 19 1997 16:21

    Hi,
    
    Can anyone explain in details how the switch  MAXIMUM_CACHED_TRANSFER
    works? I saw many topics about that but no one conclusive or clarify.
    
    We know that we can set from 0 to 1024 (blocks). Default is 32
    (blocks). How to set this :
    
    01 HSZ50
    01 cache 64MB
    WBC enabled
    3 striped mirrorsets :
    
    stripe1 = mirror1+mirror2 = mirror1 ( rz29b + rz29b) + mirror 2 (rz29b
    + rz29b)
    
    strip1=stripe2=stripe3
    
    Total 12 disks - 3 storagesets : D100 , D200 e D300 - 8GB each.
    
    What is a good settings for MAXIMUM_CACHED_TRANSFER?
    
    It's OPENVMS + INGRES. Customer says that READ of big tables is the
    question.
    
    Thanks.
    
    Leandro.
T.RTitleUserPersonal
Name
DateLines
771.1Match it to the Average Transfer sizeKERNEL::LOANEComfortably numb!!Wed Feb 19 1997 17:1811
>    Can anyone explain in details how the switch  MAXIMUM_CACHED_TRANSFER
>    works? I saw many topics about that but no one conclusive or clarify.

    This  parameter  is  used  to  stop  (for  example) a large I/O from 
    removing useful data from  the  cache;  any  I/O  larger  than  this 
    parameter  will  NOT  use cache. Before you play with the parameter, 
    it's useful to know what the I/O workload on  the  drives  is;  find 
    out  the Average I/O Transfer size from DECPS or stare at the DEVICE 
    DISPLAY in VTDPY for as long as  you  can  stand  it....this  should 
    give  you  an idea of whether the default 32 is too low (on a device 
    by device basis). The performance boost CAN be dramatic.
771.2I will try...Tks.VAXRIO::LEANDROThu Feb 20 1997 11:084
    Thanks.
    
    Leandro.