Tape Library Sizing

17. Jul, 2009 0 Comments

tapelibrary As a Storage and Data Protection consultant, I often have to look at my customer’s tape library capacity and assess current state and future state sizing needs. Sometimes I find that the backup environment has been reconfigured to meet increasing capacity demands with existing equipment capacity. For example, an admin can no longer keep 2 backup cycles worth of tapes in the library for recovery, or make tape copies for offsite storage; rather they shift to sending primary backup tapes offsite.

The practice of keeping at least 2 backup cycles in the library and making tape copies that are sent offsite, is a best practice that I’ve discussed at length with my colleagues. My intent here is to include these best practices and build them into an approach for calculating tape library sizing. I’ll leave it up to you to determine if these best practices are right for your environment. If you’re interested in reading about what constitutes a best practice, I suggest you read this blog on best practices.

With that said, let’s dive in. At a high level here are the following areas of library sizing that need to be calculated.

Slots needed for Operational Recovery tapes
Slots needed for Tape Copy
Slots needed for Scratch Tapes
Slots needed to accommodate Growth Rate

Drives needed for Operational Backup
Drives needed for Tape Copy
Drives needed for daily restores
Drives needed for expandability and growth

So let’s take a deeper look at calculating the library slot need

The library should be sized to hold at least 2 BU cycles for Operational Recovery. This should be enough to service most recovery requests without having to recall a tape. If the schedule is 1 weekly full and daily incrementals, this would equate to enough slots to hold 2 weeks of tapes. If you have a monthly full schedule, then you need 2 months of tapes in the library.

Library should have slots available for Tape Copy. Tape Copy should be done to protect against a disaster (DR), lost media, or media read issues. The slots available for tape copy should be dependant on the tape ejection cycle and worst case need for scratch tapes if the tape handler is out sick. If the tapes eject nightly and the worst case is that there is no one there for two days, then that would equate to two days of tapes. Let’s assume the tapes needed for 1 full and 1 incremental backup as the worst case.

The tape library should have enough slots available for scratch tapes to protect against the tape handler not performing his function. If the worst case is two days, then let’s say tape slots equivalent to 1 full and 1 incremental.

So how do you calculate the need for library slots? The overall slot need = (operational recovery need + Tape copy need + scratch need + accommodation for growth rate for remaining years of depreciation cycle of library)

Now let’s look at the tape drives. The methods of calculation here describe the raw capacity need and do not consider dedicated drives that are idle and unusable by other backup, restore, or copy jobs. I also assume that you are able to satisfy the minimum required throughput for these drives.

Operational Backup tape drive requirements math = (full backup in MB / BU window in seconds) / (Tape drive native speed in MB/s * average compression ratio) = drives needed. This would be a positive integer that is at least 1 drive rounded up

Example: window = 8 hrs, Full = 10TB, Drive = LTO-4 (120MB/s native). (10485760MB / 28800s) / (120MB/s * 2) = 2 LTO-4

Tape Copy: There are obvious cost and backup window considerations when deciding on how to copy backup tapes and send them offsite. Depending on your approach, the calculation for determining the number of drives need would be different for each method of tape copy
- Inline Tape Copy – would need 2 tape drives for each concurrent backup job (10% increase in backup)
- Disk based backup (staging) with copy vault, cloning, or other copy to tape from disk solution: no additional drives needed
- Tape to tape copy – would need 2 tape drives for each concurrent tape copy job

Restores require at least 2 drives for the backup data zone for occasional restores. If you restore more frequently, then 2 per library. Ideally backups should not be running during the daytime when most restores occur, but you may need to consider restoring while backups and/or tape copy is running

Overall library tape drive need would = (the need for Backup + the additional drive need for tape copy + restore drives need + accommodation for growth rate for remaining years of depreciation cycle of library)

After running through these calculations, you should have a pretty good, in order of magnitude, idea of how to size a tape library to meet your growing capacity needs for backup to tape.

- James Brissenden, GlassHouse Senior Consultant, Storage and Data Protection