IBM announced withdrawal from marketing of the model 3592-C07 mainframe tape controller. What are the implications of this move? If you want to order a new “build from scratch” mainframe system, there are no longer any options for attaching physical tape to a system Z. Of course, used tape controllers will be available on the used market, and IBM hasn’t withdrawn maintenance, so you can still get physical tape from your used equipment vendor. What IBM is really saying is that the future of tape is virtual. It may be all disk-based like the IBM TS7720, or have physical tape to offload virtual volumes like the IBM TS7740, but what the mainframe is going to see and understand are virtual 3490 tape drives and volumes.
So in that respect, mainframe high-density (3590 and 3592) physical tape is dead. You no longer need to figure out how to stack multiple disk volume backups onto high-density tape media to get the best utilization of expensive cartridges, or figure out what to do with the application that backed up 47 MB of transaction journals and needed that data sent off-site in case of a disaster. You can load your initiators up with backup jobs that dump DASD volumes to virtual 3490s until you saturate the VTS channels or control unit. But if the data does need to be stored off-site, you have a new challenge: How exactly do you get virtual data off-site in an affordable manner?
For companies that currently write backups to physical tape to be sent off-site for disaster recovery, or that archive data on off-site media for long-term retention, there are two options:
If you have physical tape back-ending your virtual tape solution, duplex copies of each virtual backup tape can be created and the duplexed copy exported for off-site storage
Implement replication of the backup data to your disaster recovery site. For an all-disk virtual solution, replication to a disaster recovery site is the only option
Sending duplexed copies of the back end data off site will not affect your RPO, but there may be an impact on your RTO. If your current backup strategy is based on stacking disk volume dumps on high-density native tape cartridges, you will have restore jobs to run at your disaster recovery site that are mirror images of the jobs you used to dump the data, so the flow of data between dump and restore should be in lock-step. But when using a virtual tape system, you have little control over how data is placed on the exported cartridges, and at your disaster recovery site, each mount of a virtual backup tape is going to force a recall from the exported physical tape. You may run into serious contention issues trying to access the exported virtual volumes.
If replication is your preferred option, network bandwidth is going to be required to get the data from here to there. How much bandwidth will depend on how much data you have, how often it gets backed up, and what type of window you’re willing to tolerate to complete the transmission. Consider that if your current physical backups complete every evening by 2:00 a.m. but the tapes don’t get picked up from dispatch until 8:00 a.m., you actually have a 6-hour window to get your data off-site. If you scale your network bandwidth so that the transmission completes in 6 hours, you are getting exactly the same off-site times that you have today.
Replication changes the cost dynamics for your backup and recovery strategy. Media and off-site storage costs are replaced with bandwidth costs. Since bandwidth is a monthly recurring cost, features that reduce replication bandwidth requirements, such as compression and deduplication, become important considerations in choosing a virtual tape solution.