Datasheet

© 2006 Cisco Systems, Inc. All rights reserved.
Important notices, privacy statements, and trademarks of Cisco Systems, Inc. can be found on cisco.com.
Page 2 of 9
BENEFITS
The Cisco SFS 7008P offers the following customer benefits:
• 96 ports of non-blocking 10-Gbps (4X) InfiniBand server connectivity with full bisectional bandwidth (1.92 Tbps)
• Powered ports to enable flexible copper and optical interface configurations
• Cisco Rapid Service Architecture to allow service and replacement of all active electronics without removing cables
• Director-class high availability and stateful failover, eliminating service disruption
• InfiniBand 1.2 Compliant
• Embedded Cisco InfiniBand Subnet Manager enables convenient “plug and play” operation for small clusters that can scale to support large
HPC clusters
• Optional external high-performance Cisco Subnet Manager for very large HPC environments
• Comprehensive performance and fabric diagnostics tools in a fully managed switch
• Easy configuration, monitoring, and maintenance in-band and out-of-band
FEATURES
The following features are included with the Cisco SFS 7008P:
• High-performance, ultra low-latency switched 10-Gbps server interconnect
• 96 non-blocking, 10-Gbps full duplex (4X) InfiniBand ports
• Full bisectional bandwidth (1.92 Tbps)
• Rapid Service Architecture where all field-replaceable units (FRUs) are redundant and hot-swappable
• Passive mid-plane chassis that does not require any cables to be removed to service active components
• Dual redundant Cisco InfiniBand Subnet Managers with stateful switchover (SSO) to enable graceful recovery from system errors
• In-depth systems diagnostics and automatic recovery from many system errors
FABRIC DENSITY AND SCALABILITY
The Cisco SFS 7008P is an 8-slot chassis that supports twelve 10-Gbps ports per slot (Figure 2). The Cisco SFS 7008P can support up to 96 10-
Gbps (4X) non-blocking InfiniBand ports. When combined with other Cisco SFS 7000 Series InfiniBand switches, the Cisco SFS 7008P provides
the foundation for building very large HPC clusters consisting of thousands of nodes to support the most demanding HPC applications.
RAPID SERVICE ARCHITECTURE
The Cisco SFS 7008P delivers the performance, scalability, and director-class uptime required for the most demanding HPC environments.
Minimizing downtime requires rapid hardware serviceability and software upgrades. With fully redundant power, cooling, and system management,
every switch component is hot-swappable and supports automatic failover. The Cisco SFS Rapid Service Architecture provides a passive mid-plane
chassis design that isolates all active electronics on the front of the chassis, while all cable connections are located at the back. The Rapid Service
Architecture allows switch modules to be replaced without detaching a single cable to reduce the mean time to repair (MTTR) from hours to
minutes.
The Cisco SFS 7008P can also maintain multiple system-image versions and automate rolling upgrades and rollback between the redundant system
management CPUs. The Cisco SFS 7008P can tolerate and recover from an internal fabric management error without any service interruption.
Reliability is further enhanced by its ability to automatically upgrade any FRUs that are inserted into the switch to help ensure that the entire system
configuration is always synchronized. The embedded Cisco InfiniBand Subnet Manager constantly synchronizes operational information with other
Cisco InfiniBand Subnet Managers within the network so that if any one switch fails, another subnet manager can assume active subnet
manager responsibility.