Storage Charging 2011 - 2012
For the period Oct 2011 thru Sep 2012 CiCS have reviewed the regime & have kept the four charge bands, with differing service levels & applicability.
The area of backup, recovery & archive under review during 2011-2012. This is reviewing admin data, student data, research data - long term/short term requirements/ regulatory requirements etc. It is highly likely that we will try to *NOT* backup to Tape wherever possible - this is not scalable/cost effective. We will aim to provide a choice whether a third copy is held at a third site.
As before, the four main categories below are for general purpose filestore (not high transaction, i/o rate) - but there are sample costings for higher performance - together with a rider.
Storage Types
Type A
- Data Mirrored between two centres - high availability.
- Data recovery time 2 days if major disaster.
- Snapshots retained for 28 days (depending on rate of change).
- Backup to tape for disaster recovery (DR) - once per week held for 4 weeks.
This provides protection against physical disaster, machine room loss, filer failure, disk corruption, software failure, admin blunder, user error.
Data is held in at least two locations (monthly tapes are at a third site) and also on different media.
Snapshots provide protection for admin/user error (if notified in time). In the event of major disaster (loss of primary site) the recovery point will be to the last mirror sync point (1 hour, typically).
Typical use - primary data store for essential live, active data
Additionally (January 2011) two variants of this storage have been added.
- Type A above is SATA disk with a SATA mirror.
- Type A** FCAL/SAS disk with FCAL/SAS mirror.
- Type A* FCAL/SAS disk with SATA mirror
Any thought of the use of FCAL/SAS needs to be seriously discussed with the CiCS storage team.
Currently CiCS storage charging is based on storage volume (bytes) - but consideration must also be given to the i/o demand of the "application". An "application" that is too demanding (ie imperils the storage service for other users by requiring excessive i/o capacity), may be requested to move to a higher tier of storage (ie SATA to FCAL/SAS) or have its own storage. The CiCS storage team will help the "application" owner determine whether the use of SAN storage can be optimised eg by tuning the application, offloading temporary/swap space to local disk, suggesting they liaise with the application vendor about the run time overhead.
Type B
- Data held at one centre (not mirrored between centres).
- Data recovery time after Type A if major disaster (3 days - 4 weeks).
- Snapshots retained for 28 days (depending on rate of change).
- Backup to tape for DR - once per week held for 4 weeks.
- Backup to tape for DR - once per month held for one year.
Snapshots provide protection for admin/user error (if notified in time). Data is vulnerable to a loss of site and major malfunction in a highly resilient hardware/software platform, in which case the recovery point may be up to 1 week away.
Typical use - data less critical than Type A, user has copies of data elsewhere or can regenerate it.
Type C
- Data held at one centre (not mirrored between centres).
- Data recovery time after Type B if major disaster (4-12 weeks).
- Snapshots retained for 28 days (depending on rate of change).
- Backup to tape for DR - once per month held for one year.
Snapshots provide protection for admin/user error (if notified in time). Data is vulnerable to a loss of site and major malfunction in a highly resilient hardware/software platform, in which case the recovery point may be up to 1 month away.
Typical use - long term archive where there is a requirement for data to be available on spinning disk for ensured accessibility. For maximum protection, users need to synchronise movement of data on to this store and deletion of original data, with the tape backup regime.
Type D
- Data held at one centre (not mirrored between centres).
- If major disaster - provide new empty space after recovery of Type A systems and reserving space for Type B.
- Snapshot retained for 28 days (depending on rate of change).
- No backup.
Snapshots provide protection for admin/user error (if notified in time). Data is vulnerable to a loss of site and major malfunction in a highly resilient hardware/software platform.
Typical use - easy access reliable managed filestore, where data can be recreated in the event of loss.
For Type A & Type B storage the snapshot reserve area is 20%. For Type C & Type D the reserve is 10%. This area is in addition to the area requested for data storage. Should the rate of change of a data volume exceed these reserves CiCS will discuss this with the customer concerned. Snapshot space is typically consumed by changes/deletions not additions