If you can arrange to make all records the same length, you can pass one record to the senary (#5) input stream to replace the master record without storage leakage.
j. On 10 March 2010 17:32, Alan Winson <[email protected]> wrote: > I have just used yesterday's release of pipes on my job that ran out of > memory as mentioned in previous posts. I modified my pipe to use the new > lookup input stream 5 (replace master) instead of streams 3 (add master) > and > 4 (delete master). (Sorry if my stream numbers are off by 1, but I trust > you know what I mean.) > > I am pleased to report that the job did not run out of storage this time, > the output exactly matches the results I got when I circumvented the > problem > by using regular REXX code including COPY (UNPACK and linein, and the > resource consumption was, of course, vastly reduced. > > With linein circumvention: > CONNECT= 01:20:33 VIRTCPU= 060:22.32 TOTCPU= 060:33.28 > > With yesterday's pipes: > CONNECT= 00:25:45 VIRTCPU= 018:32.47 TOTCPU= 018:34.88 > > I used the maximum VMBATCH storage (1912M) in both of those runs. Now the > only remaining question is how much I can lower the storage size, but that > isn't important. > > Thank you, John! >
