13. ST9990V SPC-1 Performance Results October 1, 2007 News Release: The ST9990V platform achieves the highest SPC-1 benchmark result in enterprise storage history , eclipsing results of every single enterprise storage system ever tested with 200,245K SPC IOPS. This nearly doubles the performance levels achieved by competing enterprise storage systems with a single controlle r. Leads industry with 3.5M peak cache IOPS. A 75% - 300% advantage over other enterprise systems in the market. # 1 in Performance
14. ST9990V SPC-1 Performance Results What this new SPC-1 benchmark result means for customers is: Greater Productivity : Sustain significantly more business transactions than any other enterprise storage system in the market, boosting sales and profitability Increased Efficiency : Greatly improve application response time to single digit millisecond levels and s upport more users, more applications, and handle more capacity on a centrally managed platform that leverages years of mature and reliable microcode Lower TCO : Run a multitude of applications, such as Microsoft Exchange and Oracle Database 11 g , concurrently on a single storage controller Reduced Risk : If your applications don’t perform well then your business suffers—performance definitively matters to business success .
19. ST9985 V Switched Loop Details ST9985 V introduces a switched back-end, where the drives are still on arbitrated loops but the switch logic bypasses all targets on that loop except for the one disk being addressed
23. Hardware Comparison of ST9985V vs ST9985 – Hi-Star Net Data Data path bandwidth of Crossbar Switch architecture Net: 8.5GB/s Data path bandwidth of Crossbar Switch Architecture Net: 8.5GB/s ST9985 ST9985 V
24. Hardware Comparison of Crossbar Switch Architecture Net Control Control path bandwidth of Crossbar Switch Architecture Net: 3.6GB/s The number of paths from CHA/MIX to each SM is 12. Shared memory is installed on BASE PCB Control path bandwidth of Crossbar Switch Architecture Net : 4.8GB/s The number of paths from channel adapter/disk adapter to each SMA is increased from 12 to 16. Shared memory is installed on SMA PCB. ST9985 ST9985 V
25. Hardware Comparison – Cache Memory In order to improve access performance to CM, CMA, ST9985V raised the transfer rate of the memory path, advanced memory control method, and achieved twice the hardware performance as compared with Network Storage Controller Model ST9985 . ST9985 ST9985 V
29. ST9990V SPC-1 Performance Results October 1, 2007 News Release: The ST9990V platform achieves the highest SPC-1 benchmark result in enterprise storage history , eclipsing results of every single enterprise storage system ever tested with 200,245K SPC IOPS. This nearly doubles the performance levels achieved by competing enterprise storage systems with a single controlle r. Leads industry with 3.5M peak cache IOPS. A 75% - 300% advantage over other enterprise systems in the market. # 1 in Performance
30. ST9990V SPC-1 Performance Results What this new SPC-1 benchmark result means for customers is: Greater Productivity : Sustain significantly more business transactions than any other enterprise storage system in the market, boosting sales and profitability Increased Efficiency : Greatly improve application response time to single digit millisecond levels and s upport more users, more applications, and handle more capacity on a centrally managed platform that leverages years of mature and reliable microcode Lower TCO : Run a multitude of applications, such as Microsoft Exchange and Oracle Database 11 g , concurrently on a single storage controller Reduced Risk : If your applications don’t perform well then your business suffers—performance definitively matters to business success .
44. The Competition How They Stack Up! (see notes page for detailed annotations) NO NO NO YES #15 YES Scalability NO YES YES YES ? ? ? ? NO #4 YES EMC InVista Dynamic Volume Mobility NO YES YES NO NO YES #13 NO NO NO #5 YES #2 VERITAS VxVM NO YES YES #16 MF support YES YES YES Open Systems support YES YES YES Heterogeneous support YES NO YES Leverage virtualization platform ( required for efficient data volume migration ) NO YES YES #14 Extensive selection criteria ( enabled by device mgmt. integration ) YES #12 YES #11 YES #10 Application/host level support ( required for deep selection criteria ) YES #9 NO YES #8 Integrated with configuration manager ( supports deep selection criteria ) YES #7 NO YES #6 Tiered storage support ( mgmt interface for data volume migration ) NO #4 NO #3 YES Storage controller based migration ( based on virtualization architecture ) YES YES #1 YES Non–disruptive IBM SVC Migration Softek TDMF HiCommand Tiered Storage Manager Migration Function
On the previous Hitachi Universal Storage Platform™ model, all of the disks attached to a back-end director PCB were on eight shared loops. A transfer to the last disk on a loop (could be 32 or 48 disks per loop) would have to pass through bypass logic on all of the disks in front of it. The Universal Storage Platform VM introduces a switched back-end, where the drives are still on arbitrated loops (all Fibre Channel disks use Fibre Channel Arbitrated Loops (FC-AL, not FC-SW fabrics), but the switch logic bypasses all targets on that loop except for the one disk being addressed. This reduces the propagation time by some microseconds (not noticeable), but the primary effect of this change is improved accessibility to individual disks. Each HDU has four such FSW switches, two on the front side (32 disks) and two more on the back (32 disks). Each pair of switches is either an “upper” or a “lower” switch as seen in the figure above.
On to customer successes...
Note: On the graphic above, CHA stands for channel adapter
Cache memory (CM) is installed in BASE-PCB It is only used for rack model, and is not compatible with FC model. • 4G CM: install DIMM in which 512Mbit DDR-SDRAM is mounted. • 8G CM: install DIMM in which 1Gbit DDR-SDRAM is mounted. CM capacity -------- min.4GB – max.32GB addition unit: 4GB/4DIMM CM capacity -------- min.8GB – max.64GB addition unit: 8GB/4DIMM . Cache memory (CM) is installed in CMA-PCB The following 2 types of CM are supported. • DKC-F610I-C4G: consists of DIMM in which 512Mbit DDR2-SDRAM is installed. CM capacity -------- min.4GB – max.32GB addition unit: 4GB/4DIMM • DKC-D610I-C8G: consists of DIMM in which 1Gbit DDR2-SDRAM is installed. CM capacity -------- min.8GB – max.64GB addition unit: 8GB/4DIMM • Memory intermingling specification------ intermingling within a CMA-PCB is impossible. CM (CM-DIMM) is installed on the CMA PCB. 1 set of 2CMA PCB/set can be installed in the subsystem. Up to 16 CM-DIMM can be mounted on 1 CMA PCB, and up to 32GB(16GB when using C4G) can be installed on 1 PCB. CMA is not preinstalled in DKC as a standard component, so the order for it is indispensable. NSC55 USP VM
On to customer successes...
ST9985V delivers a Updated Storage Platform that is a follow-on to the ST9985. Because it takes on some of the new technology within the ST9990V it has been consolidated into the SUN ST9900 family. Introduces support for 750GB SATA II HDDs as a new lower cost storage tier. Ideal for Tier 2 storage such as File (large) based data or short term tape backup cache. 11.25TB Raw storage in a single disk drive enclosure. SUN ST9900 Tiered Storage Manager takes advantage of the capacity!. ST9985V built-in RAID 6 ensures reliability 4Gb/s FC ‘End-to-End’. Switch FC architecture on the back-end. 15-20% faster than ST9985 (4GB/s architecture, CPU, etc.) SUN ST9900 Dynamic Provisioning software (“Thin Provisioning”). SUN ST9900 Dynamic Provisioning was introduced on the ST9990V in May. The ST9985V also enjoys Internal Thin Provisioning of internal storage as well. ST9985V and SUN ST9900 Dynamic Provisioning also introduce thin provisioning for external storage as well. SUN ST9900 is the 1 st vendor to deliver this flexibility to maximize storage resources while simplifying storage management. ST9985V also provides ‘Day 1’ Support for VMware ESX on internal storage as well as virtualized external storage for true ‘end-to-end’ IT virtualization. Again, SUN ST9900 is the first company to deliver this capability to help maximize resources. VMware ESX Server support combined with SUN ST9900 Dynamic Provisioning enables IT environments the ultimate in flexibility to align resources to business needs. Simplified software packaging – discussed in more detail shortly. ST9985V Configuration Flexibility - Choice of either: Storage controller and high performance storage for Tier 1 applications Storage controller only for virtualization services Replacement / Follow-on to ST9985.
On to customer successes...
On to customer successes...
On to customer successes...
All internal processors are 2x the performance of the USP At the core of the USP V is the third generation, non-blocking, switched architecture. The Universal Star Network is a fully fault tolerant, high performance, non-blocking, switched architecture. Each USP V model uses the same Universal Star Network architecture, but will have different performance characteristics based on the installed components. The data Cache system has the same path speeds and counts. The Shared Memory system has been significantly upgraded over the USP version, with 256 (up from 192) paths operating at 150MB/s (up from 83MB/s). On the USP, the CHA PCBs had 8 Shared Memory paths, and the DKAs had 16. Now, all PCBs have 16 Shared Memory paths. The Cache memory only contains user data blocks, whereas the Shared Memory system holds all of the metadata about the internal Array Groups, LDEVs, external LDEVs, and runtime tables for various software products. There can be up to 256GB of Cache and 32GB of Shared Memory. BED Switched Loop Details On the previous USP model, all of the disks attached to a BED PCB were on eight shared Loops. A transfer to the last disk on a loop (could be 32 or 48 disks per loop) would have to pass through bypass logic on all of the disks in front of it. The USP V introduces a switched back end, where the drives are still on Arbitrated Loops (all FC disks use FC-AL, not FC-SW fabrics), but the switch logic bypasses all targets on that loop except for the one disk being addressed. This reduces the propagation time by some microseconds (not noticeable), but the primary effect of this change is improved accessibility to individual disks. Each HDU has four such FSW switches, two on the front side (32 disks) and two more on the back (32 disks). Each pair of switches is either an “upper” or a “lower” switch as seen in Figure 21.
On to customer successes...
These are very recent analyst quotes from Bob Passmore of Gartner Group, he can ignore Hitachi’s storage virtualization solutions no longer. Passmore even states that Hitachi has “no competition”.
Self explanatory
TDMF is non-disruptive for mainframe only VxVM requires additional paths and HBA’s for host-level mirror if performance or redundancy is not to be impacted (most of the time) TDMF is a host-based software migration solution IBM SVC and EMC InVista are appliance/switch-based migration solutions. VxVM is an LVM host-based solution HiCommand Tiered Storage Manager offers the deepest selection criteria and tiered storage user interface IBM Multiple Device Manager and the SVC GUI provide user interface, however data migration is enabled by CLI only at this time. HiCommand Tiered Storage Manager integrated with HiCommand Device Manager database IBM Multiple Device Manager supports virtualization (SVC), device management (array config) and replication (FC, PPRC, etc.). HiCommand Tiered Storage Manager supports deep selection criteria, one of which is LDEVs by host/application TDMF provides a ISPF dialog interface and extensive selection criteria (i.e. VOLSER, UCB device addr, etc.) IBM Multiple Device Manager and SVC GUI support visualization of the host, however data migration enabled by CLI only VxVM provides a single host-based interface only HiCommand Tiered Storage Manager offers the most extensive selection criteria. This is enabled by the base USP/UVM/HDvM architecture TDMF scales for Mainframe only as TDMF Open Systems Edition only scales at the host level, due to TDMF being host-based HiCommand Tiered Storage Manager does not support Mainframe environments today. Volume Migration does however