FTP from a Dataspace or GETVIS - Intro

bitte in Deutsch

"Better is the enemy of good"

M. de Voltaire

It is obvious, that FTP is only as fast as your network is. But are you really using z/VSE to its limits and put as much stuff in memory as possible?
Here is a piece of code, that helps to speed up the preparation for the FTP by intercepting the write-operations of the file for FTP and rerouting them to memory. As a side effect the CPU-time of the FTP is also reduced.

I have combined PUN4IPT-techniques (pre-open VENDOR-Exit) with a driver for FTP and the result was FTP from memory.
And it works for both TCPs. The driver/input exit looks a little different, but that is it.

Initial testers found the natural limit of available GETVIS in the partition. Hence the limit was pushed to 4G, by placing the data in a data space.

The test version uses only GETVIS and is limited to 16MB. The production version allows everything that you allow for data spaces.

Addition after z/VSE 4.3 announcement

One item in the list:
  • Exploitation of innovative IBM System z10 technology
    and in this list
    • .....
    • Large page (1 megabyte page) support for data spaces
did catch my eye.
This product is (aside of PUN4IPT) one of the few products, that can help to explore the benefit of this hardware feature while on z/VSE.

More about it:
JCL as it is today (no REXX)
a REXX that FTPs (with CSI) from a data space
a REXX that FTPs (with BSI) from memory
JCL with just one line changed to use it
the code for the free version (limit = 16MB)

for E-mail to the author.
demo version.