What you will learn from this tip: Three steps you should take in order to build the best LUNs for your storage environment. Click here for a downloadable copy.
1. Conduct a storage inventory.
What kinds of information will you be storing on this system and how is it related? This breaks down into an application inventory (DBMS, Microsoft Exchange, etc.) and a file type inventory. Many applications have different kinds of files. For example, it's not uncommon to have log files and index files, as well as data files in a single application. Break your storage needs down by application and file type to determine which storage architecture, LUNs and RAID arrays will most efficiently support your enterprise.
2. Decide which type of RAID array you want for each file type.
It used to be that you put all the files for a particular application into a single RAID array in a single LUN. As server loads grew, this became increasingly inefficient, because the different kinds of files have very different storage characteristics. With Microsoft Exchange, for example, the data files containing messages are best stored on a RAID-5 array for the combination of reasonable storage size and performance. The transaction log, which is much smaller, is best kept in a mirrored array, such as RAID-1 or RAID-10.
3. Pick the right LUN-to-RAID ratio.
Most SANs can bind multiple LUNs to a single RAID array. So the natural tendency is to bind as many LUNs of the same type to the same array in the interests of economy. However the performance implications of the LUN-to-RAID ratio need to be considered carefully.
For file sharing among multiple users, where the reads and writes are pretty much random, mapping several LUNs on a single RAID works well. However, in applications where each read or write is followed by an acknowledgement, such as Microsoft Exchange or databases with their commit process, having several LUNs using the same RAID array can bottleneck performance. Performance degradation is caused by the LUNs contending for the RAID's resources. In those cases, mapping just one LUN to an array will result in better performance.
About the author: Rick Cook has been writing about mass storage since the days when the term meant an 80 K floppy disk. The computers he learned on used ferrite cores and magnetic drums. For the last 20 years he has been a freelance writer specializing in storage and other computer issues.
This was first published in December 2004