DAC-FF The Ultimate Fibre-to-Fibre Channel External RAID Controller Solution for High Performance Servers, Clusters, and Storage Area Networks (SAN) DAC-FF - Technology 100MHz RISC Processor 256MB ECC Cache 2 Fibre Channels to Host PP250299.ppt 4 Fibre Channels to Drives 3 Technology Fibre Technology Guaranteed Performance Dedicated XOR/ Cache controller Up to 256 MB ECC protected User Data Cache Burst Cache Transfer rate: over 190 MB/s Proven system architecture PP250299.ppt Fibre Channel Host connectivity Data Transfer rates of 100MB/s per channel over a 30 Meter serial link Distances of several kilometers via Fibre Hubs and switches Fibre Channel Drive connectivity 4 Channels for up to 500 drives; storage: 25.0 Terabytes RISC processor -- 100 MHz i960RN Common Management software 4 Architecture FC IOP FC IOP PCI Bus 0 32 - 256 MB SDRAM North Bay i960 RISC Processor 16 MB DRAM PCI Bus 1 FC IOP PP250299.ppt FC IOP FC IOP FC IOP 5 Hardware DAC-FF: RISC Processors (i960RN 100MHz) Firmware: 2MB Flash EEPROM Up to 256 MB ECC Data Cache Dual 64 bit PCI Bus design 5 1/4” form factor PP250299.ppt 2 Copper FC-AL host channels Burst cache transfer rates: Over 190 MB/s 4 Fibre Channel drive channels Up to 125 devices per Fibre channel - 500 devices per controller Up to 25.0TB (Seagate 50GB drives) capacity per controller 6 Board RISC Processors (i960RN) 256 MB ECC cache XOR engine 2 FC-AL channels ( to Host) DACFF 4 FC-AL channels ( to Drives) PP250299.ppt 7 DAC-FF Features RAID 0, 1, 0+1, 3, 5, 10, 30, 50 and JBOD Write Through and Write back or read-ahead support Active/Active ----- Failover/Failback for High Availability Variable Stripe size per controller Variable Cache Line size per controller Mirrored Cache for Data Integrity Clustering Support PP250299.ppt Windows NT, UNIX, SCO UNIX, NetWare Designed to fit in an Industry standard 5 1/4” half-height form factor 8 Advanced RAID Features DAC-FF Fault Tolerance: PP250299.ppt Automatic detection of Failed drives Automatic and transparent re-build of replacement drive On-line RAID expansion Hot Spare drive support Hot Drive Swapping support SAF-TE and SES enclosure management RAID configuration stored on disk and NVRAM Add a LUN; Expand a LUN 9 Simplex Performance and Capacity Performance SX Burst IO’s Per Second Burst (cache) Bandwidth (MB/s) Sustained (disk) Bandwidth (MB/s) 1,800 45 30 SF FL 4,100 97 52 4,100 97 52 FF 5,800 190 190 Note: Sequential read performance, small blocks, no coalescing driver. Storage Disks Terabytes (9 GB) Terabytes (18 GB) * Single host channel PP250299.ppt SX SF FL FF 75* 7 1.4 90 .8 1.6 60 .5 1.0 500** 4.5 9.0 ** Single ported disks 10 Performance DAC-FF Sustained transfers to Disk 190MB/s 150MB/s Burst Cache Transfer rates PP250299.ppt RAID 5 Reads RAID 5 Writes Reads Writes 190MB/s 190MB/s 11 Global Array Manager™ (GAM) Network Security Levels Online capacity expansion Remote Instrumentation PP250299.ppt Guest Access Admin Access DMI SNMP SMART Support Wizard Configuration Flash upgradable F/W OS-Independent Server API Performance Monitoring Controller Statistics Drive Statistics Historical Statistics 12 Benefits Unparalleled Performance RISC Processors--i960RN 100MHz Fibre Channel to host interface at 100MB/s per channel Fibre Channel to drive interface at 100MB/s per channel Mylex Firmware + Software Flexible Storage & Fault Tolerance Feature rich RAID NT and Unix Clustering Storage Area Networks (SAN) Mirrored Cache RAID expansion (M.O.R.E.™1) Configuration & Remote Management Global Array Manager™ (GAM) Client Server GUI Create, Manage, Expand and Monitor PP250299.ppt 13 Target Applications Designed for $30K to $200K servers Designed for Clusters PP250299.ppt Departmental to Enterprise NT Servers Low-end to Mid-Range UNIX Systems 3,000 to 30,000 IO’s per second 10’s of GB’s to 10’s of TB’s NT (Microsoft Cluster Server), NetWare ORION and UNIX clusters Support for Storage Area Networks (SAN) Active / Active Duplex Configurations Concurrently operation --> performance Transparent failover / failback --> non-stop operation Mirrored cache architecture --> guaranteed data integrity 14 Active /Active -- Fibre Interconnect Concurrent I/O Services from single or multiple servers to dual controllers increase performance Controller failover requires dual host ports Fibre silicon does not support multiple port addresses Active Reserved Controller 0 PP250299.ppt Active Reserved Controller 1 15 Failover -- Fibre Interconnect With multiple host ports, when a controller fails failover is initiated Reserved port of the surviving controller assumes port ID of the active port on the failed controller Active Reserved Controller 0 PP250299.ppt Active Reserved Controller 1 16 Failback -- Fibre Interconnect When the failed controller is replaced, port ID from the reserved port on the surviving controller is transparently returned to the active port on the new controller Active / Active operation is resumed Active Reserved Controller 0 PP250299.ppt Active Reserved Controller 1 17 Scaleable Fault Tolerant Configurations FC HBA FC HBA FC HBA FC HBA FC Hub Active Reserved Controller 0 PP250299.ppt Active Reserved Controller 1 Active Reserved Active Reserved Controller 2 Controller 3 18 Wolfpack-II Parallel Application Computing FC HBA FC HBA FC HBA FC HBA FC HBA FC HBA FC HBA FC HBA Reserved Active Reserved FC Switch Active Reserved Controller 0 PP250299.ppt Active Reserved Controller 1 Active Controller 2 Controller 3 19 Parallel App Clusters - Wolfpack-II N-Node, Switched Fibre, N-FF’s Arrays, FCAL to Disk FC HBA FC HBA FC HBA Port 0 Active Port 1 Active FF Controller 0 PP250299.ppt FC HBA FC Switch Port 0 Active Port 1 Active FF Controller 1 Port 0 Active Port 1 Active FF Controller 0 Port 0 Active Port 1 Active FF Controller 1 20