SAN Arrays

This will be a short sort of generic description os a SAN attached array. Arrays are generally classified into midrange and Enterprise classes. The difference is usually that enterprise class arrays offer more throughput and functionality. Also the vendors usually practically guarantee 100% availability and data protection. This guarantee usually comes with various caveats in terms of specific configurations and operations.

Relatively few vendors offer SAN Arrays. HP, STK, IBM, EMC, Hitachi and LSI Logic are the primary vendors. HP and Sun resell enterprise arrays from Hitachi and both companies sell their own midrange Array. Dell is reselling some of the EMC midrange arrays from mostly the lower end of the product line.

Physically an array consists of controllers and disks. The controllers offer front end connections for hosts/switches and backend connections to the disks. SAN arrays typically offer read/write cache which greatly enhances performance. The type and number of connections and the abilities of the controllers will define the array's characteristics. The array may also offer optional functionality for snapshots and data replication etc. Midrange SANs typically offer active/passive controllers where both controllers can make a LUN available to a host but only one can be active and the other is for failover only. Enterprise arrays usually offer active/active which can essentially double the throughput to the lun. Choosing midrange or enterprise array hardware is really a question of whether you can tolerate occasional downtimes. With midrange arrays this occasional downtime is really unavoidable and often unplanned. Regardless of vendor promises of uptime try to avoid using a midrange array for mission critical 24x7x52 applications. One thing I have seen happen on more than one midrange array is a situation where both controllers go offline. I have never seen this with an enterprise array like the HDS 9980.

Arrays must provide a management interface to allow management by SAN/Storage admins. Sometimes this functionality is built into the array controllers and in other cases a management appliance provides the management interface.In some cases the management is web based and others such as the Hitachi Thunder it is provided via a java client program. In the early days of scsi disk arrays management was accomplished via a terminal and a serial console connection. Console connections can still be found on modern arrays but are used for service or preliminary configuration.

Array Configuration

The following items are usually configured through the management interface.

  • Host Definitions - Usually there is a method of identifying hosts. Usually this is done by associating a hostname with that hosts HBA WWN. In some cases this is global and others it is configured per frontend port.

  • Storage/Disk Groups - Pools of storage are created from a group of disks. Storage is then allocated from the pools to form luns. Many LUNS may be allocated from 1 pool. In some cases the pool is virtualized as in the case or HP's EVA and in others the pools are created from user defined specs for various raid levels. For example with the Hitachi Thunder a raid group may be created as a raid-5 raid group. All LUNS allocated from that group will essentially be raid-5 disks. This is not the case with the EVA. VRaid LUNS can be allocated from the pool in either raid-0, raid-1 or raid-5. The HP EVA does a better job of virtualizing storage than any other vendor. Recently, Hitachi began offering an additional choice in raid group. The new group is called 6D+2P which is a raid-5 group with 6 data disks and 2 parity disks. This offers an extra measure of fault tolerance compared with traditional raid-5.

  • LUN Creation - Provides the ability to create LUNs from the storage pools.

  • LUN Presentation - Provides the ability to make a LUN visible to specific hosts.

  • Status Monitoring - In some cases snmp or mail reporting of status problems can be configured via the management interface.

  • Spare Disk Assignment - Specific Disks may be identified as spare. In the event of a disk failure the spare will automatically be used until the spare is replaced.


    You are visitor number 2886