RVic
asked on
iSeries SNA over TCP IP AnyNet support
Hi! I have 2 iSeries(os 6.1) in different data centers . Speed of file transfer is about 200 mbit per sec. through 1000Gb adapter when I send 1 file by 1 batch job (SBMJOB CMD(SAVRSTOBJ OBJ(SAVFILE) LIB(SAVLIB) RMTLOCNAME(SYSTEM1)) ). The transfer speed grows to 800 mbit per sec. when I send 5 files in the same time by 5 batch jobs through 1000Gb adapter. Why speed so little in the firs case? May be SNA has any limitations speed for transfer 1 file?
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
ASKER CERTIFIED SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
A single big SAVF isn't likely to be split up evenly over 20 disk units, and since it is a non-database object it will probably be paged in using a single process unless you create some custom code, or cache it in memory ahead of time.
This isn't a data set in the sense used in talking about MapReduce - it is an archive file. MapReduce isn't going to be much help with raw transfer of a big archive. Really just need a good parallel transfer mechanism, hopefully reading underlying data from RAID array(s) with plenty of disk arms.
This isn't a data set in the sense used in talking about MapReduce - it is an archive file. MapReduce isn't going to be much help with raw transfer of a big archive. Really just need a good parallel transfer mechanism, hopefully reading underlying data from RAID array(s) with plenty of disk arms.
You can try pure tcp ftp if it is about a file alone, maybe it is a bit faster than SNA (at least one indirection less in any case)
SOLUTION
membership
This solution is only available to members.
To access this solution, you must be a member of Experts Exchange.
You can name splitting data set into pieces in modern posh name = MapReduce ;)