DMARTIN@UAFSYSB.UARK.EDU (Daniel P. Martin) (02/05/90)
I suppose this isn't exactly on focus with the stated topic of discussion for this list any more, but what the heck... Yes - SFS is used to contain service for TSAF, GCS, and AVS under R6 VM. Based on my own experience, this has been both good and bad. Good, in that I don't have to worry (as much) about problems of the "disk full" variety when servicing these products. Bad, in that if SFS breaks and I don't have a backup copy of a working version tucked in my hip pocket, then I'm suddenly and thoroughly impaled on a helical inclined plane. Since the service procedures isolate system materials in a separate SFS service virtual machine, it's fairly easy to maintain a fall-back copy of SFS support code at a known working level, and just execute the svm from a non-DCSS copy of the code. (What's that? You applied maintenance without a backup?? But "mother" told you to never never do that...) Re work files -- Yes, you can easily wind up with work files allocated on SFS rather than on mdisk, but this seems to be by virtue of transparency rather than by agressively imposed bad design. The standard CMS search order is apparently imposed against all accessed file modes, without regard to the media type of the particular file mode. (An unintentional part of our "stress testing" on SFS consisted of watching one of our guinea pigs run an ADA compile with work files allocated in SFS. Slowed him down, exercised the server, and convinced me that I wouldn't want very many users doing this simultaneously.) Our experience with SFS has been a mixed success. I've had SFS up and going for a little over a year now, first under HPO 5.0 + CUNY SFS support mods, and under HPO 6.0 since the first of the year. The in-house users that have converted over to it like it. We're just beginning to release external users into it, and haven't had time to gather any significant results yet. However, after installing approximately 400 student users into the server last week, I expect feedback fairly soon. On the down side, I've had an open problem with the IBM support center on an SFS bug since 15 August '89. Support from the CMS change team has been singularly unimpressive. The problem (APAR VM38356) was initially opened as severity 2, and was upgraded to severity 1 on 15 January. I'm still waiting on a solution, and it's dragged on *so* long that it's not even much fun to whack my SE over the head with the combined pile of notes, descriptions, and nasty remarks about OCO any more. On the topic of SES, anybody else on VM R6 ought to go take a look at WSC FLASH 9005... Hmm. I've rambled on long enough. Maybe there's a more appropriate forum for continuation of this discussion? Anybody care to suggest one? -dan. Daniel P. Martin - Systems Programmer III University of Arkansas Main Campus - Fayetteville BITNet: DMARTIN@UAFSYSB InterNet: DMARTIN@UAFSYSB.UARK.EDU or @PostOffice.UARK.EDU VMSHARE: -UA