I would like to add one other advantage that I took advantage of. When you define your hipersocket network, you can define a very large MTU size for the hipersocket network. This can eliminate a lot of segmentation/re-assembly overhead because the data transmission is broken up into fewer large pieces. This really can help elapsed time for transmission of large things like FTP of large files, NJE transmission of large reports, or perhaps DRDA between database systems, and I never noticed any downside for smaller things.
As mentioned previously, both networks can co-exist, and you can define the routing so that hipersockets are preferred. I don't have my definitions handy but we used VIPAs and so we could reach the zOS image via either of two IP addresses. We used 192.168.x.x addresses for the internal (hipersocket) network; I recommend making the last octet or the last two octets the same as those of the OSA IP address, but it's not required. You can test the difference in speed - find a large dataset (or create one), and FTP it using the OSA network and the other system's non-hipersocket IP address, then FTP the same file using the hipersocket IP address. ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
