Choosing the right data storage systems is important and requires some planning considering that SAN storage systems do not come cheap. Consolidating storage into an iSCSI SAN can improve your disk space utilization and simplify an admin’s life when having to provision new servers or storage in the future.
Once having chosen an iSCSI storage area network, it’s necessary to choose the heart of your SAN, the disk array. iSCSI arrays may be purchased from vendors including iSCSI specialist companies such as Nexsan, EqualLogic, LeftHand Networks and EMC. Pricing for iSCSI arrays range from $5,000 to several hundred thousand dollars.
First off, one must plan out how the new array will be used. How many terabytes are needed for a disk-to-disk backup target or data storage for your new Exchange or SQL Server cluster? Is storage needed for a few servers, or are you implementing a SAN to support 100 blade servers for your Web and terminal server? Answer these questions and you can put together a list of features and requirements your iSCSI array must have to support your needs.
Disk space is key
For a disk backup system you’ll need a lot of space. A basic iSCSI array that presents its RAID storage sets as iSCSI logical units (LUNs), is one option, however, an iSCSI LUN can only be accessed by one host server at a time, and these systems can only handle a few host server connections that need large blocks of disk space.
Smarter systems allow admins to create one or more RAID sets and allocate different LUNs from them. One can create a 1 TB RAID-5 set out of seven 250 GB drives and assign 50 GB to one server and 800 GB to another server leaving the rest of the space free for ad hoc or development servers. Remember to check how many servers a prospective array will support. Some are limited to as few as eight servers, which would make them the wrong solution for a Web server farm of 50.
Take advantage of snapshots
Exchange, SQL Server and other transactional systems running of a SAN should take advantage of a SAN array’s snapshot capabilities. In the event of a server crash or a corrupt database, a server can be restored in minutes to a snapshot rather than the hours it could take to restore a large database even from a disk backup. Snapshots can even be taken every hour while the server is active so data loss, if any, is substantially smaller.
Plan for the future
Plan for expansion and upgrades. Traditional systems allow admins to add capacity by hanging additional shelves of drives off the controller module, though one must still have to buy a controller big enough to handle eventual growth or face a costly forklift upgrade when the SAN is outgrown. Other systems allow multiple drive arrays and controllers to form a single virtual array. As arrays are added, so are controllers, with cache memory and gigabit Ethernet interfaces increasing system performance. You may start small and grow to 100 TB or more without the forklift.
If trying to consolidate, one can combine NAS for file storage and iSCSI for block storage on a single integrated device. iSCSI traffic should be isolated from general network traffic, so make sure that an integrated system has enough gigabit Ethernet interfaces to provide the level of redundancy you need on both networks.
Consider replication. Some iSCSI arrays support snapshot replication where a disk array at the primary site takes periodic snapshots of your sensitive data LUNs. It then replicates the changed blocks to another array at the disaster recovery site. Other sans support synchronous replication, where all changes to the primary array are sent to the secondary array in real time. Others can also do either.
There are numerous iSCSI array options and solutions. Figure out what’s important and talk to some vendors to help plan the best solution that’s right for you.