The TS7720 Virtualisation Engine is the first mainframe VTL offering for System z without support for tape media. Instead, it can hold up to 70 TB native on 300 GB, 15,000 rpm Fibre Channel drives. Up to three TS7720s can be joined in a grid, and customers have the option of configuring the grid for three-way high-availability failover or to look like one pool of storage. With three boxes linked together, the system can support up to 210 TB. The system could hold up to 600 TB of capacity with compression turned on, according to Charlie Andrews, IBM marketing director of storage systems.
However, Andrews said the disk cache in this instance is not intended for long-term retention of data. Mainframe systems are more likely to store active data on tape cartridges than open systems, and hierarchical storage management (HSM) processes will probably still be used to write archives to tape for long-term retention. The TS7720 is aimed at data sets that might not require such a long shelf life but do require quicker access from nearline disk.
IBM is also adding disk cache to the TS7740 Virtualisation Engine. Customers will be able to add two more drawers to existing systems for a maximum of six drawers with a capacity of 9 TB. New customers can get a new set of disk shelves that use 300 GB disks for a total of 13 TB. The TS7740 will also now support IBM's TS1130 tape drive, including the 1 TB cartridge and the TS3500 high-density tape library.
Andrews didn't rule out future integration with the data deduplication IP acquired with Diligent Technologies into the mainframe product line. "We are going to leverage the deduplication IP broadly, and clearly that's a consideration," he said.
FICON updates for DS8000
The DS8000 high-end disk array, which is increasingly being positioned in support of IBM mainframe products with IBM's open systems XIV clustered block storage product on the scene, has moved even closer to the z10 with this announcement.
"DS8000 vs. XIV has been the topic of many conversations lately between storage analysts and IBM storage executives," wrote Illuminata analyst John Webster in an email to SearchStorage.com. "Mainframe connectivity is a good place to start with the analysis because there is clear differentiation – DS8000 is mainframe and open systems, which XIV is open systems only. IBM I think wants to keep that differentiation clear at least for the near future."
"However, I do believe that XIV is an architecture that will stand up well as time goes on and will do so better than the architecture of the DS8000 on both the open systems and the mainframe sides of the house."
Clipper Group analyst Mike Kahn said he didn't see a change in IBM's positioning for its products, but use cases for different arrays may be growing more specialized.
"At one time, all high-end arrays were mainframe centric and to some extent they still are," he said. "The flip side is that the idea of enterprises using only one kind of array for all applications -- I think we're beyond that. There are very few enterprises out there anymore trying to provide that kind of simplicity."
Updates to the FICON and Fibre Channel protocols on both the z10 side and the DS8000 side mean that the mainframe can use standard Fibre Channel ports to transfer FICON commands and data. Andrews described the updates as "cutting down on the chattiness" between Fibre Channel and FICON protocols. Lowering this overhead during data transfers has boosted performance in internal testing from 350 MBps over a 4 Gbps connection to 525 MBps and will also improve the efficiency of the transfer of small files in OLTP environments.
With FICON and Fibre Channel brought closer together, will there be further convergence with Fibre Channel over Ethernet (FCoE)? "It's too early to predict the end game yet," Andrews said. The way the market is moving in servers, networks and storage is a separation between converged physical layers and logically separated virtual layers, "and I don't see us having a whole new separate hardware infrastructure in the data center just for FICON, he said"
However, Andrews admitted a new converged network does raise potential performance and cost concerns. "Today, FICON delivers very high performance that needs to be maintained across any new physical interweaving of connectivity and needs to be done at an accessible price point," he said.
"This is one of those things that has been 'going to happen' for a long time," Kahn said. "It's still on the horizon in front of us."
Another update to the DS8000 connects a new data path among sites in a three-way high-availability configuration for disaster recovery. Previously, the three sites were joined with synchronous replication using Metro Mirror to a nearby location and with asynchronous Global Mirror replication over distance. If the primary site failed over to the secondary site, a feature IBM calls HyperSwap, syncing the new primary to the third site through Global Mirror wasn't done automatically.
After the system had failed over to the secondary location, Global Mirror would have to be restarted to the third location even if most of the data was already there, Andrews said. A new feature called z/OS Metro/Global Mirror incremental resync will automatically establish a new connection between the secondary and tertiary sites after HyperSwap.