gregpf@momenta (Greg Presedo-Floyd) (12/22/90)
I have questions on "Disk duplexing". Please email me directly with any/all comments/opinions you have on the following topic. I will summarize in a later post. 0) Our server environment: We have Adaptec 1540 SCSI boards. They drive Fujitsu 1 Gig external winchesters. Netware 386 3.0 (moving to 3.1) . Ethernet NE2000 and WD8003 coax and 10BaseT cards. 1) What is "disk duplexing" ?: GOAL: A robust/redundant server "Disk duplexing" , as I understand it, allows an entire SCSI chain (disk controller, all disk(s) connected to that controller) to be continuously and transparently have the same data written to each redundant side. If something on one side goes bad, operations continue and the OS sends a warning message out. I think I need to go to an Adaptec re-seller and get a Netware "disk duplex" NLM to do this. 2) QUESTIONS about "duplexing" : Is my specific hardware "duplex-able" ? Will this work ? Is it working at your site ? How hard was it for you to set up ? Has anyone had a disk/controller failure and had this save them ? Anyone know a source for sales and support of this in the SF Bay area ? A supplimentary thing that I will do to make our server operations robust is to have an identical second server, also disk duplexed, operating and on-line or at least powered down and standing by. Will these 2 things make my server as robust as possible ? More suggestons ? Thanks ! -- Greg Presedo-Floyd "Representing myself, only" Momenta Corporation 295 North Bernardo Ave Mountian View, CA 94043 uunet: {apple,leadsv,hpdtc}!momenta!gregpf
kenh@techbook.com (Ken Haynes) (12/26/90)
In article <1990Dec21.234852.21220@momenta> gregpf@momenta (Greg Presedo-Floyd) writes: >I have questions on "Disk duplexing". > >0) Our server environment: > > We have Adaptec 1540 SCSI boards. They drive Fujitsu 1 Gig external > winchesters. > > Netware 386 3.0 (moving to 3.1) . Ethernet NE2000 and WD8003 coax and > 10BaseT cards. Make sure that you have backed up all your data *before* you upgrade. The migration path to 3.1 using the generic SCSI driver NLM requires that you repartition the drive. This will cause all data to be lost! If you got the 3.0 generic SCSI fix, you've already done this.... assuming you've installed the fix. > >1) What is "disk duplexing" ?: > > "Disk duplexing" , as I understand it, allows an entire SCSI chain > (disk controller, all disk(s) connected to that controller) to be > continuously and transparently have the same data written to each > redundant side. Correct. > > If something on one side goes bad, operations continue and the OS > sends a warning message out. > > I think I need to go to an Adaptec re-seller and get a Netware > "disk duplex" NLM to do this. Nope. You only need another disk controller card and drive to duplex. You set up the duplexing you would use the install.nlm program. Refer to the installation manual under load INSTALL pg. 131 of NW3.1 Installation. > >2) QUESTIONS about "duplexing" : > > Is my specific hardware "duplex-able" ? It should be. > Will this work ? Is it working at your site ? It does work. It is working at my site, although not in exactly the same configuration. > How hard was it for you to set up ? Piece of cake. > Has anyone had a disk/controller failure and had this save them ? Yep. Although not since we've upgraded to NW386. It did save me on a 2.15 SFT system. > A supplimentary thing that I will do to make our server operations > robust is to have an identical second server, also disk duplexed, > operating and on-line or at least powered down and standing by. Server duplexing is not far away. Until you can mirror servers, what you are doing is the best bet. > > Will these 2 things make my server as robust as possible ? > More suggestons ? > Not only will it make your server more robust, it will increase performance as well. NW386 will assign read/write operations as efficiently as possible performing what is called a "split seek", or getting the data from both disks at the same time but different parts of the file. In other words, it will schedule disk 1 to get sectors 1, 3, 5, and 7 while it schedules disk 2 to get sectors 2, 4, 6, and 8. This decreases head seek time and thus increases performance. Good luck! Ken -- ****************************************************************************** Network Support Services: UUCP: {nosun, sequent, tessi} kenh@techbook