I have seen some Model9 implementations during security assessments. Usually the S3 backup is taken to their on perm server that stores the data on a DS8xxx (or compatible) and mirrors that elsewhere using standard DS*xxx mirroring. So restore can be at the regular disk speed on site or on backup site.
ITschak ITschak Mugzach *|** IronSphere Platform* *|* *Information Security Continuous Monitoring for z/OS, x/Linux & IBM I **| z/VM coming soon * On Sun, Oct 29, 2023 at 7:29 AM Brian Westerman < brian_wester...@syzygyinc.com> wrote: > Yes, > > I have 4 smaller clients that use cloud tape connector who store the > second copy of migrated datasets to the cloud, as well as some DR related > items. Also, I have their DS8K DASD conduct backups of the full DASD > volumes directly to the cloud. The neat part of doing that directly is > there is no CPU overhead involved in the process. > > In a DR test, we reload the cloud volumes from the cloud, and we direct > HSM to use the second migrate copy instead of the primary one. It did seem > a 'little" slow, but was offset by the fact that we had the volumes > immediately available to be loaded. We had no tapes to transfer. > > Those sites could not afford and frankly did not need PPRC and a second > DS8K sitting around so they saved a lot of hardware costs. > > Obviously I'm over simplifying a bit, but if you put the thought into the > process, you can have a viable DR for frankly a very low cost. The biggest > cost issue for using the cloud isn't writing the data, it's reading it > back, depending on what your plan is, it can cost several times as much to > read as it does to write. > > One of the sites uses Adabas, and after we flashcopy the database to a > backup set of volumes, we then use CTC to write that saved database to the > cloud and as the PLOGs are created, they are also copied to the cloud. > > Obviously none of these sites are a bank, and redoing work between the > last PLOG (they cut them hourly) and the current time is completely > possible and not a real hardship. > > Writing to the cloud isn't as fast as writing to a VTS, but it's not super > slow either. Just make sure you have a zIIP processor so that the CPU > overhead on your actual CPs is kept in the trivial zone. Normally you > wouldn't have any production processes that write directly to the cloud. I > try to have HSM do most of the work where possible, even if that means > creating an extra backup copy of "important" production datasets. It's a > lot of work to set things up properly, but then most DR solutions are > pretty involved. If you don't spend a lot of time thinking it all through, > your DR test, and more importantly the actual DR will likely fail, so take > your time and think it all through. > > Brian > > ---------------------------------------------------------------------- > For IBM-MAIN subscribe / signoff / archive access instructions, > send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN > ---------------------------------------------------------------------- For IBM-MAIN subscribe / signoff / archive access instructions, send email to lists...@listserv.ua.edu with the message: INFO IBM-MAIN