Tivoli Systems rolled out a storage technology earlier this month that lets network professionals share data among different host systems by building pools of storage that can be allocated by application, business need or degree of fault tolerance.
Code-named Storage Tank, the software ties any Unix, Windows NT or 2000, or Linux host to any storage system via traditional storage-area network switches to create a single virtual pool of data that can be shared by any host, user or application.
The technology is one of the first that treats data in an agnostic manner irrespective of the server file system and creates virtual pools of storage that can be assigned to different applications, class of user or importance of data. For instance, Network Appliance blends the Network File System used in Unix systems with the Common Internet File System used in NT and Win 2000, but does not apply attributes to the data that lets IT managers set policies. Other vendors, such as Compaq and EMC, add data to a common storage pool that is assigned to a single server.
"These vendors give you the ability to take disparate storage, pool it, and let you slice it and dice it and make it available to server A, B or C," says Steve Duplessie, analyst with Enterprise Storage Group. "But they stop short of [Tivoli's Storage Tank's] approach, which gives you additional attributes to create policy on."
For instance, IT managers might want to create three storage groups -- a group for very important data, less-important data and data that is not at all mission-critical. Then they might want to put their very important Oracle database data on an IBM Shark Storage array, their less-important data on a Compaq StorageWorks array and the remaining data on a just a bunch of disks (JBOD). The mission-critical Oracle e-commerce application running on NT would be stored on the IBM storage array and saved continuously with snapshot backups. Solaris or Linux servers could access the Shark and retrieve data for a data-warehousing site running Win 2000 and save the data to the Compaq StorageWorks array, which is backed up incrementally each day. At some time, when the data becomes less frequently used, it will be stored on the JBOD.
Also included in Storage Tank is the ability to set policies that control the migration of old data from one storage device to another or that could indicate where a certain type of data should be stored. Policies can let IT managers enforce service-level agreements.
Storage Tank is the culmination of eight years of development within IBM Research and Tivoli, and is based on technology that has only been available in high-end mainframe environments.
Tivoli will provide a developers' kit for operating system vendors that want their platforms to work with Storage Tank, and a program for application and database vendors that lets them certify their products with Storage Tank.
Dataquest says storage management technology will increase from about $740 million in 2000 to nearly $1.7 billion in 2004.
Tivoli will ship Storage Tank the second half of next year. Pricing is not available.