Storage Area Networks (SAN)
SAN stands for Storage Area Network. It is a technology for delivering diskspace to servers. Most modern infrastructure architectures these days contain SAN's.
A SAN consists of a large amount of disks in a so-called storage array. A large amount can be between dozens of disks and hunderds of disks. With today's disk sizes a SAN can contain multiple terabytes of data (1 terabyte = 1024 gigabyte).
Connectivity
The disks are connected to one or more diskcontrollers. The controllers can be based on SCSI, iSCSI, Fibre channel, SATA or some other technology.
The controllers are connected to one or more SAN switches using Fibre channel. The servers are also connected to the SAN switches, through HBA's (Host Bus Adapters).
Fibre channel is a special SAN technology for efficiently sending disk-blocks of data over fibre cables.
Backup
SAN's often have tapelibraries connected to them for backup purposes. A backup can be done by the SAN itself. Data is copied from disks to tape without using servers. This way both the servers and the LAN experience no extra load during backups.
Virtual disks
A SAN virtualizes all disks and tapedrives that are normally installed separately on every server (Direct attached Storage - DAS).
The SAN presents disks to the operating systems on the servers. This is usually not one-on-one. The SAN's disk controller splits up all disks in small pieces (physical extends). From these small pieces, new virtual disks (LUN's - Logical Unit Numbers) are composed and presented to the operating systems. The operating systems don't know about the physical disks, they just work with the LUN's as if it were disks.
The SAN can place the physical disks in some type of RAID array for reliability and/or speed. For instance 8 disks of 72GB in a RAID5 configuration can be presented as 1 virtual disk of 500GB. The operating system only sees the one very reliable disk and has no knowledge about the disk really consisting of more physical disks.
Cloning and snapshotting
Typical SAN functionality is cloning and snapshotting. With this functionality it is possible to "freeze" data. Data is not changed for some time, so a backup can be made, or a copy can be made for BI (Business Intelligence) or datawarehouse purposes.
Cloning means that the SAN creates a copy of a (virtual) disk. This looks like making a RAID1 mirror disk. This clonedisk can be split-off of the SAN, so a backup can be made of it.
Snapshotting looks like cloning, but snapshotting is somewhat more intelligent. A snapshot takes a point in time from which no writing to the disks is permitted anymore. All writing from that moment on is done on a separate place in the SAN. The disk has still read-access. When an operating system reads data that was just written (and put on the separate place in the SAN), the data is retrieved from the special place on the SAN automatically. This is a fully transparent process, the operating systems have no knowledge of the snapshot technology at all.
Because no data is written to disks during the snapshot, a backup can be made from the disks. As soon as the snapshot is removed, all data is written to the original disks again.
A large advantage of snapshots compared to cloning is that clones take relatively much time to create, and cost much diskspace, because all data must be copied before a clone can be used. A snapshot is available the moment the snapshot is started. If a snapshot is not used for too long, or is not much data is written, the snapshot takes not much diskspace as well.
Software
All of the above can be arranged using standard SAN management software. This is where SAN vendors differ. The quality of the SAN management software deters much of the ease of use of the SAN.
Large SAN vendors are HP (EVA series), EMC, IBM and Hitachi.
This entry was posted on Friday 05 January 2007