7 things that are very wrong about tape based mainframe backup


7 things blue@2x What's inside

The mainframe backup architecture was designed in the 1970s. Back then, physical tape was the only practical solution for creating additional copies of data due to costs and limited capacity of primary storage. As years went by, physical tapes evolved to virtual tapes, enabling automation, improved tape utilization and shorter backup times.

However, mainframe backup software has not changed much since the old days, and still treats tapes as physical media and not virtual. For this reason, the mainframe backup architecture still suffers from many limitations, inefficiencies and incurs high costs, which are not really required in modern backup architectures.

7 Very Wrong Things about mainframe tape-based backups:

  1. Physical tapes require manual handling, hard to utilize capacity in an efficient manner, slow, and unreliable.

  2. Virtual tape systems still emulate mainframe tapes therefore they must connect to the mainframe over FICON - a proprietary connectivity protocol. There are only few storage vendors that still make FICON-attached virtual tape system - mainly IBM, EMC, Oracle, and Luminex. This oligopoly drives up costs of virtual tape systems in comparison to SCSI or IP attached virtual tape systems - up to 10x higher.

  3. Tape is a serial device that can only be accessed by one application at a time - writing a data set to tape must be done sequentially, it is not possible to employ parallelism to improve write speeds. When reading a data set from tape, the tape and drive are held exclusively by the reader. This means one tape cannot be read by two applications in parallel, and the total number of applications reading/writing from tapes is limited by the number of tape drives. Additional applications must wait for the tape or drive to become free.

  4. Backup and archive to tapes require legacy data management interfaces that run on CPs and incur high CPU utilization, which often impacts MLC pricing when in R4HA monthly peak

  5. Due to high costs of compute, compression and encryption are delegated to the VTL and not done by the backup software. This means data that could have been saved by compression is sent over the wire, which in turn drives up the need for more robust, more expensive infrastructure to sustain throughput. The data is also sent insecurely to the drive, could be tapped or tampered with, in a way that likely violates modern-day security regulations.

  6. Recycling is required to consolidate data spread across many tapes - high cpu and data movement through host to consolidate tapes. Without recycling tape data is not freed and cannot be reused by the VTL and the backup software keeps managing large numbers of tapes even though the data expired.

  7. Unless the virtual tape library copies the virtual tapes to physical tapes, virtual tape data actually remains written on disk drives inside the virtual tape library. This means the data is still immediately accessible to applications running on the mainframe. - susceptible to ransomware…