NovaNET - Storage Area Network (SAN) Support

The NovaNET Advantage

Integrated solution

NovaNET Alliance is unique in the industry as it is the only product that was designed to handle client-attached devices from the start.  This has meant that the transition to SAN was a natural progression; the designed-in integration means that NovaNET SAN support is seamless and hence is easy to setup and use.  The standard NovaNET Alliance installation can support SANs without any further software installs.

Easy Setup

Users' biggest fear with SAN solutions is the setup process;  NovaNET addresses device setup in an automated fashion.  Using NovaNET 's plug'n'play style device configuration, a device is initially configured to NovaNET but to only one system. NovaNET then has the ability to auto-configure the same device to other systems on the SAN.  This functionality is crucial on SANs where a library with several drives may be shared by many systems.

Transparent Operation

A tape drive or tape library is always controlled from the NovaNET server.  The setup is such that all devices are logically connected to the NovaNET server although the server does not necessarily need physical connectivity. The NovaNET server arbitrates the devices used so that only one backup/restore operation can take place to a tape drive at any one time.  If one NovaNET client on the SAN is doing a backup to a tape drive, then other NovaNET clients would need to use a different device or go into a queue until the transfer is completed.

From an operational standpoint, the fact the device is SAN-attached is transparent.   NovaNET will always try to use a device that has local attach or SAN connectivity thereby avoiding the network at all times.  This is done automatically and the user is not required to set up his backup jobs in any special fashion.


SAN Serverless backup

What is it ?

SAN Serverless backup provides a method for transferring data directly from the disk to the tape device without going through the host machine.  This works by having a SAN hub/router doing SCSI Copy commands.  This takes a physical block from the hard disk and copies to the tape.  

Benefits

The biggest benefit is reducing the amount of data the host has to process.   Normally data would be taken up to the host from the disk and back down from the host to the drive.  In SAN serverless backup the host only needs to send each SCSI COPY initiator command; although this reduces that amount of data to the host, the host is repeatingly sending the initiating COPY commands.

Disadvantages

This sort of backup is only appropriate to data where physical data copy is appropriate.  This is currently only applicable to filesystem backups.  With databases, data is backed up in a logical fashion which means that data is retrieved via the database not by going directly to the hard disk.  For instance, Oracle backups are done using RMAN, Informix using ONBAR,  Sybase and MS SQL also have an API for backups.  None of these APIs supports direct physical access.  Serverless backup is not suitable for database backups. 

Unfortunately even filesystem backups via this method have limitations.  Firstly, because data is backed up on a disk block physical level, restores of individual files and directories are difficult - in some cases impossible, needing the whole partition to be restored.  This also means that incremental backups are seldom possible - forcing the user to backup the complete filesystem every time.

Secondly, the technology requires a tight integration into various filesystems which is normally done via "backdoor" methods as most filesystems do not have APIs to support this. Some vendors do not allow this and interfering with the filesystem in this way invalidates the vendor's support for the OS. To further complicate matters OS vendors sometimes make minor changes to their filesystems that do not affect traditional backup methods, but invalidate the direct block access system.

Misconceptions

One misconception is that, because the host is not involved, this is the fastest method of doing backups.  This is untrue.  Having a host do a backup means that the host can buffer a certain amount of information; this means that if either the source (the disk) or the destination (the tape) should slow down for brief period, the buffering will smooth out the data transfer.  This kind of buffering is not available in a Serverless application where 8-32 Mb of buffering in a router/hub is inappropriate.   The main reason why benchmarks have shown an improvement is due to the difference in doing raw block level backups against logical filesystem backups.

NovaNET 's position on SAN Serverless backups

NovaNET believes that SAN serverless backups are going to evolve into a useful technology and the NovaNET product has provisions for its use.  Currently however,  filesystems have not been designed to work with this technology and the "back door" approach raises more problems than it solves.  Strategically NovaNET is ready for SAN serverless backup but is waiting for filesystem technology to mature enough for these sorts of solutions to be a realistic industry proposition.



About SAN Backup

SAN in Backup storage environment - What does it offer?

Fundamentally, SAN allows multiple machines connected to a SAN to share a storage device.  

In a backup environment this gives the following advantages:

  • Performance - Conventionally, when multiple machines use a single tape library, they transfer the data across the network to a machine that has connectivity to the tape devices; because each machine on the SAN can directly communicate with the tape device, transfers speeds are equivalent to having a device locally attached to the machine's SCSI bus.
  • Can reduce cost - Where performance is critical, it is necessary to have direct attachment to tape devices.  This could mean having several tape libraries connected to each machine; because SAN allows multiple machines to share the same hardware, the storage can be consolidated thereby saving cost.  There is a careful balance that needs to be achieved, as implementing SAN can be more costly than the hardware you are trying to share.
  • Flexible architecture - In a traditional environment some machines have the centralized storage devices.  If further expansion is required, it is often quite complicated to try to redistribute the resources. For the reason that SAN attached devices are shared, it is much easier to add hardware and adjust which machines use which devices.
  • Reduces network load - Although some sites have enough network bandwidth to allow backups to occur across their network, the burden on the network is often deemed unacceptable.  SAN removes the load of the network.

SAN, especially in the backup storage environment, is useful, it works and is ready now.