> 1. For another project we are investigating if it would be > beneficial to > have a 2.4 kernel under Z/VM 4.2 and VM TCP/IP with > hipersockets and virtual > lan etc. Do current users of these mechanisms find they are > reliable and > perform well ? Or would IUCV be recommended for example ? > Comments would be > appreciated.
Yes, although with the caveat that the hipersocket support is still a bit difficult to understand and that it is very "black box" -- you don't know what's happening under the covers, and unless you're a serious assembler jock, you won't *ever* understand it. Once you get the driver set up once, it's fantastic, and from an operations and network management perspective, guest LANs are the only way to go. Use CTC or IUCV only when necessary. > 2. Do shops who run Linux in production generally use VM mainly as a > hypervisor and use a Linux toolset to control and manage the > Linux guests, > eg. alert processing if a guests fails, security, and dasd > backups. Or is a > VM toolset generally used, or even a mix ? We use both. Some automation (such as startup/shutdown) is better/easier done from the VM side. Things that require knowledge of what's happening inside the Linux systems are better done from the Linux side. You rely on Neale Ferguson's VM OpenEdition ports of a lot of the standard Unix services (syslog is particularly useful) and the Linux hcp device driver for issuing commands from Linux to the underlying hypervisor to communicate between the two layers. From the VM side, you use SCIF, PROP, REXX, RXSOCKET and the VM SSL support to interact with things VM and Linux. TRACK is also incredibly valuable, and as soon as someone has time to teach it how to interpret Linux control blocks, it'll be even more useful. -- db
