> OTOH I would not entertain running a parallel sysplex over DWDM. A former employer did. Distance about 22km. Due to IBMs requirement for parallel sysplex pricing for at least one machine being on a different box. That distant lpar was (obviously) sharing ISGLOCK, and response times from ISGLOCK to the distant lpar were in the 200-400ms range as opposed to (I believe) 20-40ms for response times on the same physical box. In any case, the difference was so huge that I was never able to show response times from all lpars in the sysplex for one structure in the same graphic and have the graphic be meaningful. And the GRS structure has the fastest response times in a parallel syysplex. If it is slow, other structures will be even slower.
> Meanwhile the consequences of disrupted sysplex communication are too > grave, including wait stating the entire sysplex. Too great a price to pay > for whatever advantage you're reaching for. True. There was at least one of those, including a split sysplex on re-IPL of the distant lpar. Which resulted in taking down the (so far) unaffected (local) lpars. Which eventually resulted in an unscheduled sysplex-wide IPL. Didn't we just talk in another thread about human errors being far more of a risk than hardware failures? Combine both hardware failure and sysprog mistake, and you have an even greater mess.(It really upset someone when I pointed out his mistake from the remnants of hardcopy log available instead of opening an ETR to IBM.) Barbara ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: INFO IBM-MAIN
