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

Conference ssdevo::hsj40_product

Title:HSJ30/40 Product Conference
Moderator:SSDEVO::EDMONDS
Created:Mon Jul 12 1993
Last Modified:Fri Jun 06 1997
Last Successful Update:Fri Jun 06 1997
Number of topics:1264
Total number of notes:4958

1200.0. "redundant HSJ40 improves performance??" by PTOSS1::MINSHALLJ () Fri Jan 31 1997 09:27

    I am having an Alpha/VMS/HSJ performance problem (see VMSNOTES #120
    for details).  
    
    In brief, backup/image from a particular NONredundant HSJ causes
    a 40-80 times increase in response time for a file copy operation
    to another drive on the same HSJ.  There is very little activity
    on the other drives on the HSJ while the backup is going on.
    The I/O rates and thruput is much lower than during the day.
    
    It has been suggested that redundant HSJs can share cache, and that
    adding another HSJ to make a redundant pair would alleviate the
    problem.  This hypothesis is attractive since the performance
    problem doesn't seem to exist for backups from other redundant
    HSJs in the cluster.
    
    Is it likely (or possible) that a redundant HSJ could dramatically
    improve performance while BACKUP is going on?
    
    
    (Note: this problem does not seem to occur on other (redundant)
    HSJs, even when copying files to the disk that is being backed up!)
    
    All HSJs at V27j
    
    Thanks,
    Jerry
T.RTitleUserPersonal
Name		DateLines
1200.1Not what you thinkSSDEVO::RMCLEANFri Jan 31 1997 10:0815
>>    It has been suggested that redundant HSJs can share cache, and that
>>    adding another HSJ to make a redundant pair would alleviate the
>>    problem.  This hypothesis is attractive since the performance
>>    problem doesn't seem to exist for backups from other redundant
>>    HSJs in the cluster.
    
>>    Is it likely (or possible) that a redundant HSJ could dramatically
>>    improve performance while BACKUP is going on?
    
    
I don't know what your problem is but redundant HSJ's do NOT share cache
in a dynamic manner.  The only thing that happens is that if a controller
fails the other controller is able to access the data for failover purposes.
Redundant HSJ's will help only if the backup is on one controller and the
disks in use are on the other controller.
1200.2thanksPTOSS1::MINSHALLJFri Jan 31 1997 15:563
    Thank you.
1200.3I'm fogged.GEM::SHERGOLDA FOOL'S living PARADISESat Feb 01 1997 01:3616
    
    In brief, backup/image from a particular NONredundant HSJ causes
    a 40-80 times increase in response time for a file copy operation
    to another drive on the same HSJ.      
    
    
    <<<	This statement does not clearly state what you are trying to do or
    <<<	what this 40-80 times increase is. A 40-80 increase from what? and
    <<<	what response? I am afraid your briefness has lead to a very
    <<<	unlcear definition of your problem. It would appear to me you are
    <<<	doing a disc to disc copy using backup but I do not see what you are
    <<< comparing this with to define the response anomaly. Are you saying it
    <<< takes 40-80 times longer to do this when both discs are on a single HSJ
    <<< than when they are different ones?
    
    Keith
1200.4see VMSNOTES 120PTOSS1::MINSHALLJTue Feb 04 1997 08:0217
    re .3
    
    The details are in VMSNOTES #1200.  
    
    A summary is that the command "COPY file D400:file" normally takes
    6-8 seconds.  However, this same command can take 100-400 seconds to
    complete when a single backup occurs from another disk which is on
    the same HSJ as D400:.  This situation seems to be unique to
    a nonredundant HSJ.  It doesn't seem to happen to other HSJ's which
    are redundant.
    
    Someone had suggested that the HSJ's can share cache.  I just
    wanted to know if that could be a possible explanation for
    the performance problem.  This was answered in an earlier reply.
    
    
    Jerry