What is an optimum disk subsystem for Linux messaging?
Message stores, even for medium-sized organizations, tend to grow very large, very quickly. Message stores with hundreds of gigabytes and even terabytes of data are not uncommon. Storage Area Networks (SAN) and Network Attached Storage (NAS) systems combines multiple RAID arrays and disks into a single manageable disk system. SAN and NAS systems are connected to the messaging server via a high speed channel such as FibreChannel or an IP-based network. SAN and NAS systems give messaging systems additional flexibility for clustering multiple host servers accessing the single enterprise message store, providing additional levels of fault tolerance and scalability. The hardware components that comprise these Linux-based enterprise systems can be low-cost commodity devices connected through standard interfaces.
Linux has multiple disk options that can be configured and optimized to handle loads generated by messaging applications. Typically, enterprises manage large message stores using redundant arrays of independent disks or RAID technology. RAID enables multiple physical disks to appear to the operating system and messaging applications as a single logical hard disk. RAID arrays can be configured to solve a number of data management concerns such as performance and fault tolerance.
Among the most popular RAID configurations for Linux messaging applications are RAID-0, RAID-5 and RAID-1+0. Simple disk striping, known as RAID-0, can provide added performance for email applications that involve largely transient e-mail traffic as is common with Message Transfer Agents. Redundant disk configurations, such as RAID-5, can provide cost effective disk redundancy for email applications that rely on large message stores. Mission critical environments, that require the highest levels of performance and redundancy can be configured in RAID-1+0 configurations, combining disk striping and disk mirroring.
This was first published in March 2004