BACKUP STRATEGIES
TRADITIONAL BACKUP SYSTEMS
In the past, the almost universal method of providing backup for data storage was by way of a Tape Backup System.
Over the years these systems have continued to grow in capacity, matching the advances in Disk Storage.
As disk storage grew from Megabytes to Gigabytes, tape sub-systems also grew in size.
In the more recent past, the growth of disk sub-systems has been explosive, whilst Tape Sub-systems have peaked at around 800Gb.
Actually, this is a slight exageration as the size is 400Gb/800Gb where the raw capacity of the tape is 400Gb, uncompressed and the assumption is that MOST data can be compressed 2:1 allowing 800Gb on a single tape. Generally this is not true and a more typical storage would be 600Gb, and in many cases would be the 400Gb size.
For systems with less than 800Gb of Disk Sub-system required for backup, the existing tape systems are adequate, if not speedy.
For systems requiring more than 800Gb of tape Storage, we can move to the systems which use Auto-Loaders and accommodate multiple-tape backups for a single session.
Tape sub-systems using Auto-Loaders tend to be significantly more expensive than the single loader equivalents.
There is however, a second significant problem and that is the speed of these tape sub-systems, both for backup and for recovery.
Once we get to very large systems, the time for backup can exceed the backup cycle period.
it is no use to have a system that requires 24 hours to perform a daily backup.
Again, in the past it was mostly possible to restrict the amount of data to be backed up to a small amount and time was not a consideration and size was not a condiseration.
As disk sizes have moved beyond 80Gb, and then 300gb and then to the 1Terabyte range, we see an increasingly rapid move towards storage sizes which eclipse the comparable tape backup systems.
LARGER BACKUP SYSTEMS
One of the recent innovations in backup storage was the introduction of the Disk-Disk-Tape backup system.
The initial backup would be done at very high speed using Disk-Disk backup using ever more sophisticated techniques to minimize the amount of actual data copying.
The initial Disk-Disk backup could be as little as a few minutes and in many instances much less than 1 hour.
Of course, storage size is a consideration for the total amount of time.
Once the Disk-Disk backup is completed, then the Disk to Tape function could be initiated and to ensure that time is minimized, there could be MULTIPLE tape drives running concurrently to backup the disk data.
Example:
We have 5Tb of data for backup and tape capacity of 500Gb.
If we employ a sub-system which has 10 tape drives attached, we can backup the 5Tb of data in the same time as is required to write a single tape of 500Gb.
This is not possible with a traditional Disk-Tape system as we do not know the backup size until the full backup is taken.
Once the disk copy is performed, we can calculate just how the data will be distributed across the tape drives AND know that the image on disk is FROZEN.
In any event, these larger backup volumes tend to be beyond the scope of what can be accomplished in the Small/Medium business environment.
Worse, the amount of data to be subject to backup is increasing rapidly.
Whereas a few years ago it would be highly unusual to have a system which stored 1Tb of data [1000Gb], these systems are now becoming very common.
It is possible to buy systems with 4Tb of storage for use as a home server and at home user prices!
Tape storage backup becomes a problem.
Clearly new solutions are required to handle these larger systems.
RECENT CHANGES IN BACKUP SOLUTIONS
Up until the Windows 2003 server, Microsoft has provided the NTBACKUP solution with the option to backup to a number of devices including Tape Sub-Systems.
With Windows 2008, Microsoft has quietly dropped support for Tape Backup.
This is not an accident, but clealry a pointer to the inability of tape Systems to keep up with the requirements of the ever rapidly increasing size of storage requirements.
In Windows 2008, the emphasis is on Disk-Disk systems.
Disk systems will fail!
Why would we trust our data to Disk when we know it will fail at some point in time.
The tape systems of the past have been reliable.
Unfortunately, this is not strictly a true statement.
Tapes have a very limited shelf life.
Good practice says you should replace your rotated tapes every year, or at maximum 2 years.
Relaibiality of data stored on tape over 2 years is not guaranteed.
In fact there are VERY FEW systems that have an ARCHIVE quality of data storage.
Microfilm is one of the few media which has an ARCHIVE quality with the ability to store information for over 100 years.
Microfilm is not very suited towards data storage in the current era.
But equally, MOST users have no need for archival storage of data.
Those that do have this requirement must invest VERY HEAVILY in suitable dedicated systems which specialize in ARCHIVE requirements.
MOST BACKUP SYSTEMS are used to be able to recover data quickly and easily, and in MOST CASES are interested only in recovery of the most recent backup of data.
Sometimes the most recent backup of data may need to be what happened a few seconds ago.
For DISASTER RECOVERY, a mechanism is required for RAPID recovery of information to the most recent point in time, which may be hours, or days, or perhaps months.
The requirement for OLDER backup is usually quite rare and perhaps sometimes it is necessary to go back 1 year.
With proper planning of on-line data storage, this can become a minimal problem or even a non-problem.
However, by restricting the span of time that backups must accomodate, we can also change the concept of how the back data is stored and used.
RAPID DISASTER RECOVERY BACKUP
Once we look at the NON-ARCHIVAL form of backup, we can look at alternative methods of backup and recovery which are outside of the traditional TAPE BACKUP scheme.
The systems also fit with the direction Microsoft [and others in the future] are taking in managing and creating backup storage.
A. FIRST LEVEL OF DISASTER RECOVERY:
At this level we strive to ensure that there will be NO DATA LOSS in the event of a single failure of a storage device.
This is typically accomplished by using some form of RAID implementation.
Although there are now a wide vairety of RAID chopices, they basically fall into 2 possible solutions.
RAID 1: This is a simple MIRROR copy of the data. requires twice the storage capacity of the data to be stored. Failure of 1 single disk drive results in ZERO loss of data.
When the failed disk drive is replaced, it is automatically mirrored again and fail safe operation resumes.
RAID 5: This is a more complex structure which requires a minimum of 3 storage devices with NO DATA LOSS in the event of a single failure of a stroage device.
The storage capacity is more efficient than the RAID 1 option. We have a storage capacity of the sum of the disk drives minus one drive.
When a failed disk drive is replaced, it is automatically regenerated and fail safe operation resumes.
When HOT PLIG drives are used, disks can be replaced without the need to power down the system.
When NON HOT PLUG drives are used, the system must be shut down first, the drive replaced and then the system restarted.
Functionality is the same in both cases.
B. SLIGHTLY HIGHER LEVEL OF DISASTER RECOVERY
More recently, larger operations are worried that the loss of 2 drives concurrently, or near concurrently is a possibility, and such a failure would result in irrecoverable data loss unless there is a backup copy of the entire system somewhere. This issue can be resolved using a RAID 6 system.
RAID 6: Similar to RAID 5 except there are 2 redundant drives. The failure of 2 drives will not result in loss of data. Total storage capacity is thus the sum of all of the drives minus 2 drives.
This solution is being used more and more often in a variety of situations which demand a much higher level of security against data loss such as in Virtual machine enviroments where there may be multiple users of Virtual Disk sub-systems and those subsystems may be higher than 5Tb in size. Size is not so critical as the issue of recovery.
A and B SUMMARY:
The systems of disk stroage discussed in situation A and B only deal with LOCAL server storage and do not address the problems of OFF-SITE storage or SERVER FAILURE resulting in data corruption.
C. DISTRIBUTED FILE SYSTEMS
To protect against data loss in a given SERVER, it is often advisable to have data REPLICATED across 2 or more servers, even if those servers are int he same physical location.
Data Replication acorss servers in a WAN environment is also a common solution for getting data to the office closest to the user.
Windows Server offers a Distributed File Service [DFS] which allows data to be instantly replicated between servers.
In the event of a server failure, users will still have access to the latest data from the remaining operational server[s].
This is relatively simple to implement in systems running Windows Server 2003 R2 or later.
The cost of doing this in a multi-server environment is zero in equipment and only a slight impact in time and processing efficiency.
D. DISTRIBUTED FILE SYSTEMS and SAN STORAGE
An alternative to having DFS replicate data across servers is to use a STORAGE AREA NETWORK system which is connected to the system LAN [Gigabit or higher speed].
The function is similar to option C, but also covers the case where there are multiple server failures [but no SAN failure].
C and D SUMMARY:
The use of DFS provides an additional level of protection and recovery of system data beyond that obtained in A and B.
Caution should be applied. In the event of application corruption [from whatever source] which results in data corruption will be instantly replciated across the various systems and there may be no recovery from this form of data destruction. Protection from this situation requires generational data stroage, versioning or a series of other protection mechanisms.
E. MOVABLE STORAGE:
After implementation of the systems described in A through D, we still need some form of generational backup which can be physically detached from the active system, and hence free from the possibility of corruption from any nefarious on-line activity.
As mentioned earlier, TAPE is no longer a preferred media, and removable storage will need to in th form of a DISK SUB-SYSTEM which is built for rugged operation and easily moved without possible physical damage. The problems here are that the size of the storage must be sufficient for convenient copy/transport.
If we are keeping a backup of 1Tb of on-line data, then the movable storage system must be AT LEAST the same size.
If we are keeping GENERATIONAL backups, we may need larger capacities, totally depending on the defined backup strategy.
The simplest of systems would be a 100% backup on a daily basis. This would require 7 backup disk devices, each the full capacity of the on-line storage.
This provides ONLY 7 days of storage.
If we require only a full backup weekly and an incremental backup daily, then we would need a full size storage device for the weekly backups and a smaller dvice for each of the 6 daily backups. However, the cost of the smaller drives may not be substantial and in reality 7 drives the same size would be probably recommended. One or two spare drives shoudl always be held for emergency in the event of a drive failure.
MOVABLE STORAGE STRUCTURE.
Many of the removable storage devices recognize the problems of failure and are configured as RAID 1 devices.
As mentioned above, this allows for the failure of a single drive within the removable storage media so that a single failure results on no loss of data.
When using removable storage this is the optimum approach.
There are many variations on what type of system can be used for the movable storage.
Price varies significantly.
Performance varies significantly.
Choices for the movable storage include USB2.0 devices, FireWire devices, eSATA devices and LAN devices.
Choices include single drive devices and RAID 1 drive configuration devices.
Price/Performance is a driving factor in the choices.
F. SPEED OF RECOVERY CONSIDERATIONS:
If continuity of business is a major concern, then the choice and configuration of backup and recovery is not a simple choice.
First step is to define HOW FAST the recovery of operation must be.
F1. INSTANT RECOVERY:
In this case, it is mandatory that there be a mirrored distributed system located in some physcially separate secure location.
Systems must be implemented which allow for Internet access to information.
This requires a systems approach which AT ALL TIMES is dependent on a CLOUD TYPE of service provision.
Local:
This is readily achieved with minimal changes to existing infrastructure.
First step is that ALL users are operating in a Remote Desktop Environment to the data servers.
Second step is that ALL user data is located on the servers - preferably in a Virtual Machine environment.
Remote:
Third step is to ensure there is at least ONE mirrored server located somewhere else which is ON-LINE to the local servers and is part of the distributed file system such that all data is being concurrently updated on a transaction by transaction basis.
Users:
Must have access to computer devices somewhere which also have internet access and can access to either the local or remote servers.
In the event of failure of either the Local or remote site, data access is still possible.
An advantage of this implementation is that users can be located ANYWHERE IN THE WORLD and be able to work as though they were in the local office desk.
The problem is one of suitable security in such an environment.
F2. BACKUP RECOVERY:
Given the above system structure, then the Data Backup should be implemented to provide INSTANT data copy either local or remote.
The speed of the recovery wil be dependent on the speed of the devices and their method of connection to the infrastructure.
F3 GENERATIONAL RECOVERY:
The ability to go back multiple generations of information needs to be more carefully and extensively planned.
There are a multiple number of choices for such staged backup/recovery and these MOST LIKELY involve some form of Tape Storage.
Such systems would need a good degree of definition and planning to ensure that they meet he business needs.