Competition suddenly is heating up at the top of the mainframe storage world. EMC introduced the high end DLm8000, the latest in its family of VTL products. This one is aimed for large enterprise mainframe environments and promises to ensure consistency of data at both production and recovery sites and provide the shortest possible RPO and RTO for critical recovery operations.
It is built around EMC VMAX enterprise storage and its SRDF replication and relies on synchronous replication to ensure immediate data consistency between the primary and target storage by writing the data simultaneously at each. Synchronous replication addresses the potential problem latency mismatch that occurs with the usual asynchronous replication, where a lag between writes to the primary and to the backup target storage can result in inconsistent data.
Usually this mismatch exists for a brief period. EMC suggests the issue, especially for large banks and financial firms—its key set of mainframe target customers—is much more serious. Large financial organizations with high transaction volume, EMC notes, have historically faced recovery challenges because their mainframe tape and DASD data at production and secondary sites were never fully in synch. As such, recovery procedures often slowed until the differences between the two data sets were resolved, which slowed the resulting failover. This indeed may be a real issue but for only a small number of companies, specifically those that need an RTO and RPO of just about zero.
EMC used the introduction of the DLm8000 to beat up tape backup in general. Physical tape transportation by third party records management companies, EMC notes, hinders recovery efforts by reducing what it refers to as the granularity of RPOs while dramatically increasing the RTO. In addition, periodic lack of tape drive availability for batch processing and for archive and backup applications can impair SLAs, further increasing the risks and business impact associated with unplanned service interruptions. That has been long recognized, but remember EMC is a company that sells disk, not tape storage, and ran a Tape Sucks campaign after its purchase of Data Domain. What would you expect them to say?
The DLm8000 delivers throughput of up to 2.7 GB/s, which it claims is 2.5x the performance of its nearest competitor. BottomlineIT can’t validate that claim, but EMC does have a novel approach to generating the throughput. The DLm8000 is packed with eight Bus-Tech engines (acquired in its acquisition of Bus-Tech in Nov. 2010) and it assigns two FICON connections to each engine for a total of 16 FICON ports cranking up the throughput. No surprise it can aggregate that level of throughput.
EMC has not announced pricing for the DLm8000. The device, however, is the top of its VTL lineup and VMAX enterprise storage tops its storage line. With high throughput and synchronous replication, this product isn’t going to be cheap. However, if you need near zero RPO and RTO then you have only a few choices.
Foremost among those choices should be the IBM TS7700 family, particularly the 7740 and the 7720. Both of these systems provide VTL connectivity. The TS7700 avoids the latency mismatch issue by using a buffer to get the most optimal write performance and then periodically synch primary and target data. “Synchronous as EMC does it for VTL is overkill,” says an IBM tape manager. The EMC approach essentially ignores the way mainframe tape has been optimized.
Among the other choices are the Oracle Virtual Storage Manager and Virtual Library Extension. Oracle uses StorageTek tape systems. The Oracle approach promises to improve tape drive operating efficiencies and lower TCO by optimizing tape drive and library resources through a disk-based virtual tape architecture. HDS also has a mainframe tape backup and VTL product that uses Luminex technology.
EMC is a disk storage company and its DLm8000 demonstrates that. When it comes to backup, however, mainframe shops are not completely averse to tape. Disk-oriented VTL has some advantages but don’t expect mainframe shops to completely abandon tape.
In other storage news, IBM recently announced acquiring Texas Memory Systems (TMS), a long established (1978) Texas company that provides solid state memory to deliver significantly faster storage throughput and data access while consuming less power. TMS offers its memory as solid state disk (SSD) through its RamSan family of shared rackmount systems and Peripheral Component Interconnect Express (PCIe) cards. SSD may be expensive on a cost per gigabyte basis but it blows away spinning hard disk on a cost per IOPS.
Expect to see IBM to use TMS’s SSD across its storage products as one of its key future storage initiatives. Almost every other storage vendor is incorporating some form of SSD or Flash into its storage architecture.
Finally, IBM today introduced the latest rev of its top end mainframe. The new machine, the zEnterprise EC12 (zEC12) delivers more performance and capacity that its predecessor, the z196. Both machines support multi-platform hybrid computing.
With a 5.5 Ghz core processor, up from 5.2 Ghz in the z196, and an increase in the number of cores per chip (from 4 to 6) the zEC12 indeed is faster. Supporting 101 configurable engines compared to 80 on the z196, IBM calculates the zEC12 delivers 50% more total capacity in the same footprint.