What lessons can the storage industry take from the mainframe? [We can learn] that consolidated storage is easier...
By submitting your email address, you agree to receive emails regarding relevant topic offers from TechTarget and its partners. You can withdraw your consent at any time. Contact TechTarget at 275 Grove Street, Newton, MA.
to manage, with fewer human resources, requires less electrical energy and other environmental costs, and provides better chances for coordinated disaster recovery. In the case of September 11, nearly all mainframe data was recovered within 6 days. Most PC data was lost forever. What are some of the challenges of mixing mainframe storage and open systems storage? First, the operating systems that run on mainframes are open, including OS/390, z/OS, Linux on S/390, z/VM and VSE. Few people realize that OS/390 is a UNIX-branded operating system, and complies with many of the open industry standards that other platforms comply with. IBM prefers the term distributed systems, relating to servers that have not yet been consolidated onto a mainframe. One challenge is that the Storage Area Network (SAN) technology that IBM invented in the early 1990's is based on ESCON, and industry standard serial protocol. The SAN second generation is Fibre Channel Protocol (FCP), which is based on SCSI. Although mainframes now support FICON, which is ESCON protocol over Fibre Channel cabling, the challenge is that switches, routers and directors can only support either FCP or FICON, but not yet both. Switch vendors have announced plans to offer "intermix" directors that support both FCP and FICON over common fabric. Also, IBM has made a "statement of direction" to support FCP directly from its eServer zSeries line of mainframe servers. These announcements should address this challenge. What lessons can the storage industry take from the mainframe? The world of mainframe better understands the importance of disaster recovery plans, 24x7 applications and efficient usage of tape resources. Have the concepts of hierarchical storage management been correctly adapted by open systems administrators? HSM has taken on two forms in the open systems environment. The first being server HSM where the movement of backed up data is from one storage medium to another, based on policies. The second is client HSM where seldom used or large files residing on users machines are automatically migrated to near line storage (in this case TSM server controlled storage), and then automatically recalled whenever an application needs them. This creates a virtually infinite storage space for end users. Are tape solutions as efficient in the mainframe environment than open systems market? The mainframe environment provided much of the foundation for distributed systems tape solutions. In fact, the mainframe solutions can be seen as being most efficient. What are the top 3 concerns in the mainframe storage market? One, getting LAN-free backups so they work between mainframes and distributed systems. Two, sharing resources efficiently across the mainframe and open-systems. Three, a common storage resource manager that can view both. Are there any misconceptions about mainframe storage? Misconception 1: Tape is dead. Tape is very much a vibrant technology, and ideal for disaster recovery. Misconception 2: Disaster recovery is too expensive. No, different levels provide different levels of support. Disk mirroring is more attractive than bankruptcy.
For more information:
If you have your own questions for Trica and Michael, be sure to join us in March 19, in SearchStorage's 'Storage management tips and tricks' forum. They will be there all day on Tuesday answering your mainframe storage questions.
Check out our sister site Search390 for more on mainframe technology.
Scroll through a bunch of mainframe storage links generated by a SearchStorage TargetSearch.
About Tricia Jiang:Tricia Jiang is a technical attache for IBM Tivoli Storage Manager. She acts as a technical liaison between IBM Tivoli Storage Manger customers, development and marketing.
During Jiang's tenure at IBM she has held a variety of positions. She was hired to work in ADSM technical support in Tucson, Arizona. She then worked as a developer for the IBM 3466.
About Michael Pousson:Michael Pousson is a technical attache for IBM Tivoli Storage in San Jose, California. He acts as a technical liaison between IBM Tivoli Storage Products customers, development and marketing. Pousson joined IBM in 1966 and has worked with data management and storage management software development in San Jose and Palo Alto, California, and in Tucson, Arizona. During Pousson's career at IBM he has held a variety of positions. Is it true that retrieving files from a mainframe has typically been slow? If so, have there been any enhancements to speed up this process?
No. Mainframes are designed for high-bandwidth. Retrieving mainframe data from distributed systems involves TCP/IP, and as such Gb ethernet can help in this respect. Retrieving data from disk or tape can run at FICON speeds and distances, also based on Gb technology. How much of the mainframe storage market has moved, or do you see moving to Networked storage?
Nearly all mainframe storage is already networked with ESCON and FICON facilities. Non-networked storage (called OEMI or parallel channels) based on copper cabling, are fading from mainframe use, and the new z800 mainframe does not support non-networked storage. What are the top 3 concerns in the mainframe storage market?
One, lack of industry standards for advanced functions, like disk mirroring and copy functions. SNIA and others are trying to address this. Two, the lack of enterprise storage resource management tools that are integrated across the enterprise (including mainframes and distributed). There are over 21 companies, including IBM, developing solutions in this ESRM space. Three, data sharing. Although we can share resources today, different platforms cannot share data with proper serialization to prevent overwrites or inadvertent access. The Enterprise Storage Server (ESS) disk subsystem is a good example of shared resources, in that the same ESS can be shared among multiple systems, but mainframe data is not shared directly with distributed systems, and rather allow sharing indirectly, such as with NFS, CIFS, or application-specific connectors.