Ron Hawkins writes: >2:00 am for which timezone, and middle eastern countries do not >find Sunday outages acceptable. 24x7 really means 24x7. Give me >the version without an outage..
Right, that version is called Parallel Sysplex. However, *reducing* the duration (and risk) of planned service outages is still a useful improvement in many situations, even if for some reason you cannot get Parallel Sysplex (which can *eliminate* service outages) implemented right away. Hence the TogglePlex suggestion. (No, that's not an official name. Maybe it ought to be? :-)) >And it's not a new idea. The Sysprog I worked with in Singapore >was doing this over 10 years ago on two frame Parallel Sysplex. >Swing over to one LPAR country by country on Friday night, IPL. >the empty LPAR to the new Maint level, and then swing back country >by country on Sunday night. I was thinking about a TogglePlex sans Parallel Sysplex. I'm sure it's not a new idea. But I see many customers that do not yet use a TogglePlex approach (or something like it), so I thought it might be worth mentioning (and giving an unofficial name; perhaps someone has a better name).. About the only cost-related issue I can think of with a TogglePlex is reserving some memory to facilitate the toggle. But that's not much cost at all -- and it might even be zero additional cost in the real world. For example, the memory could be temporarily borrowed from another LPAR that could tolerate being shut down for a relatively brief interval just before the toggle, such as a development LPAR. Also, in many cases the memory allocated to the uplevel LPAR could initially be less than what was allocated to the downlevel LPAR. If the toggle is scheduled at a relatively "quiet" time (in memory usage terms), you can start off with a smaller memory allocation and verify that the uplevel LPAR operates correctly. Once verified, you can shut down the downlevel LPAR (or, better yet, another LPAR) and (nowadays) dynamically add memory to the uplevel LPAR. (All that is easy enough, but if you happen to have z/VM it's even easier.) It is possible to maintain a group LPAR capacity softcap lid throughout the whole toggle operation (and beyond), so there should be no cost-related issues there. I'm not saying that a TogglePlex offers the same level of service as Parallel Sysplex. It clearly doesn't: there's still a (shorter) service interruption, and the operator has some more work to do. But a TogglePlex is a decent step forward along the higher availability continuum, in between a single production LPAR and a Parallel Sysplex. - - - - - Timothy Sipples IBM Consulting Enterprise Software Architect Based in Tokyo, Serving IBM Japan / Asia-Pacific E-Mail: [email protected] ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to [email protected] with the message: GET IBM-MAIN INFO Search the archives at http://bama.ua.edu/archives/ibm-main.html
