oberman%erd131.DECnet@LLL-ICDC.ARPA.UUCP (01/28/87)
>An earlier question about GPXs and performance led me to look at LRP, SRP, and >IRP values on several systems. Can someone explain to me why the LRPCOUNT, >IRPCOUNT, and SRPCOUNT values set in and reported by SYSGEN aren't the same as >the numbers reported by show memory? The show memory numbers always higher for >LRPs; higher for SRPs and IRPs on a GPX, but the same on a 750. Why? The size of the look-aside lists as well as pool space may be increased (but never decreased) by the system. The SYSGEN parameter LRPCOUNT, for exampl, is the initial number of LRPs to allocate from non-paged pool. The system may increase the number to LRPCOUNTV. The same is true for SRPs and IRPs. As the look-aside lists are icreased, there may be inadequate non-paged pool. The system may then increase this space. The expansion of these lists is NOT real efficient. If nothing else they are not likely to be contiguous. DEC recommends that you use MODPARAMS.DAT and the ADD_ functions to expand the initial values to where they will seldom, if ever, increase. You should also increase NPAGEDYN to where it seldom is less than 100000 bytes. A good way to see what your pool looks like is to SHOW MEMORY/POOL/FULL. DEC (or at least one of the developers at DECUS) recommends that you NEVER do an AUTOGEN starting at SAVPARAMS. He said to start at GETDATA or later and use the parameter INITIAL for P3 to prevent the system from looking for OLDSITE%.DAT files. This is because the use of the OLDSITE files can propagate undesirable parameters through AUTOGENs. See the VAXVMS handouts (Volume 2) from the SF DECUS for a good set of notes on how to maintain a good MODPARAMS.DAT file. and an overview of just what AUTOGEN does. R. Kevin Oberman LLNL arpa: oberman@lll-icdc.arpa (415) 422-6955 ------