Centralize storage arrays and management

Posted on December 01, 1999

Centralize storage arrays and management

Jim Straw

Demands on data are increasing rapidly, in large part due to audio, graphics, Internet, and e-mail applications. Protecting these assets is critical, and ready access to data is imperative. But storage is also a performance element of information technology, not just a repository for data, images, and audio files. One approach to protecting data while maximizing performance is to centralize your storage resources.

With applications and storage capacities expanding at a rapid rate, a storage system must be able to easily expand capacity and handle a growing number of I/O operations cost-effectively. While storage area networks (SANs) and clustering are positive directions, can existing disk arrays be easily integrated into these environments? Moreover, are software suites available to tailor storage solutions to fit users` requirements and budgets? Traditional server-attached storage is inefficient and hard to manage. But centralized storage is efficient, manageable, and cluster-ready.

In these environments, a disk-array subsystem that incorporates hardware and software to allow dual connections from a single server to the central storage unit is more advantageous than one that offers a single connection. Dual paths provide load balancing across the server connections, plus an element of fault-tolerance.

Open interfaces are a must with any centralized storage system. In FC-AL environments, for example, you can attach several servers to the same storage subsystem through a hub or a switch. To avoid data corruption, users can assign specific volumes to specific servers on the loop. Centralized storage systems that allow administrators to perform logical unit number (LUN) mapping and masking ensure that data is assigned to specified servers and that the servers can access volumes with the correct LUN number.

Some SANs and centralized disk-array systems include monitoring components that allow users to configure storage from anywhere in the world through either a dial-up modem connection or a LAN connection. This is a major benefit for corporations with multiple locations. IT managers can monitor operating characteristics, environmental specifications, performance data, and system alerts. Virtual drives can be created, deleted, expanded, or moved among servers from anywhere in the world. An SNMP link can provide monitored statistics and information to third-party network management packages such as HP OpenView, IBM Tivoli, or CA Unicenter.

Ease of use is also critical, and virtualization of storage volumes promotes ease of management. Applications, spread among a number of servers, often need dedicated volumes. From a storage architecture viewpoint, it can be advantageous to map a dedicated volume to a physical set of disk drives within the storage device. Virtualization helps relieve the nightmares of assigning LUN addresses to physical drives.

Virtual management

Management software should improve capacity planning and load balancing for the entire enterprise because all servers attached to the storage device share storage. Virtualization within a centralized subsystem allows administrators to easily create copies of virtual volumes. While copies are created, the source drive remains online with no disruption to users, thereby creating a "zero backup window." Users can copy a virtual drive to be backed up and assign it to a dedicated backup server or tape library.

This type of backup scenario can be performed at any time during the day, even when network traffic is at its heaviest. Network administrators don`t need to perform backups after hours or worry that backup operations didn`t run.

Other applications for this type of management package should also be included (such as Y2K or other non-production testing) without fear of data corruption, creating multiple versions of critical data and changing RAID levels on the fly. And when volumes are removed, the storage capacity is recycled by the system and is instantly ready to be assigned to another server.

Being able to either point a server to a volume or move the volume from one server to another, with no impact to the network, changes the way users operate their businesses. Processes that once took hours now take minutes because the volume itself doesn`t have to move.

Other key considerations are scalability and disk-drive usage. With storage virtualization, software must allow multiple servers to share the same storage volumes. This is key to server clustering and file sharing. Application and server failover for clustered configurations can be performed through any third-party clustering software, such as Veritas` FirstWatch, Mercury`s SANergy, Novell`s High Availability Server, or Microsoft`s Cluster Server. File sharing is often a necessary approach to manipulating large image files generated by prepress, medical, etc., applications.

Click here to enlarge image

For each virtual drive, management software stripes data across all physical drives, allowing simultaneous access to all drives.

Jim Straw is director of OEM sales for XIOtech Corp. (www.xiotech.com), in Eden Prairie, MN.