Cisco's FabricPath: Multi-Path For Those Who Can't Wait For Standards
Cisco made a few announcements at CiscoLive that extends thier footprint in the data center and offers Cisco customers migration paths to 10Gb Ethernet. The biggest announcement is Cisco's pre-standard multi-path Ethernet protocol called FabricPath that makes better use of switched infrastructure. Cisco is pretty far ahead of other networking vendors and most data centers with FabricPath, but for early adopters, Cisco and their customers will get some first hand experience with multi-path Ethern
July 1, 2010
![Network Computing logo Network Computing logo](https://eu-images.contentstack.com/v3/assets/bltde8121fc52c5c8f3/blt3f3d0318f746b1c2/65a530e4187606040a1d8b8c/placeholder.png?width=1280&auto=webp&quality=95&format=jpg&disable=upscale)
Cisco made a few announcements at CiscoLive that extends thier footprint in the data center and offers Cisco customers migration paths to 10Gb Ethernet. The biggest announcement is Cisco's pre-standard multi-path Ethernet protocol called FabricPath that makes better use of switched infrastructure. Cisco is pretty far ahead of other networking vendors and most data centers with FabricPath, but for early adopters, Cisco and their customers will get some first hand experience with multi-path Ethernet. FabricPath is another step along Cisco path to virtualzing the data center network.
FabricPath is a pre-standard implementation of the IETF Transparent Interconnection of Lots of Links (TRILL) that provides a standard way to support multi-path Ethernet network. Today's Ethernet networks are built on a single-tree model where there is only one path between any two nodes. Multiple paths between nodes create conditions for broadcast storms which can degrade a network. The Spanning Tree protocol is used to dynamically detect loops, or multiple paths, in Ethernet and disable one path restoring the single path architecture. The impact of a single tree meant that redundant and fault-tolerant networks were expensive to build since the full capacity of the network could not be utilized. The inefficiency grows in fat tree network designs where the bandwidth increases as you move toward the core. For example, a redundant 10Gb uplink to a core network switch would be unused due to spanning tree.
Multi-pathing uses a routing protocol to manages routes between nodes and suppresses broadcast storms. The competing IEEE 802.1aq Shortest Path Bridging provides similar functionality and is being backed by other network vendors. In both cases, the routing protocol IS-IS, establishes a network wide view of the layer 2 network, and lets each switch determine the best path through the network for frame. IS-IS also support a basic load balancing algorithm so that multiple paths can be uses simultaneously which should improve application performance and decrease the problems that congestion cause.
Cisco FabricPath is what the company calls a "pre-cursor to TRILL." Supported in the NX-OS on the Nexus platform, users can migrate to TRILL in a hitless fashion when the IETF standard is ratified and implemented in NX-OS, the company promised. FabricPath currently allows upto 16 simultaneous paths between nodes in the network. Those links could be 10Gb links providing 160 Gbps connectivity into the network fabric. The fabric architecture is flexible enough that it can support mesh network designs to suit any need. FabricPath, like TRILL and 802.1aq are meant to be very hands of requiring little network configuration or management. If you are running migrating to the Nexus platform, FabricPath will be a benefit. Most data centers won't need to move to a multi-path network until they start to deploy FCoE or start to dive into virtualization and data center automation. FabricPath is a licensed option starting at $25,000 and will be available Q3 of 2010.
Cisco is also announcing a FabricPath Switching System bundle, a starter kit that includes five Nexus chassis, the Fabricpath license, system validation and the Data Center Network Management software that can be purchases as a single SKU. Designed for high-density data centers, the design simplifies large scale roll out and is scalable to 160 Tbps. Pricing is not available at this time and bundle will be available in Q3 2010.Cisco also announced a new line card the Nexus 7000 F-series module which support 32 1Gb or 10Gb auto-sensing ports, and allows the Nexus 7000 to scale to 512 ports total. The F-series module support 320Gbps switching with 5 micro-seconds latency. The F-series module to Nexus 7000 oversubscribes 1.4:1, supporting 230 Gbps to the backplane. The module supports TRILL and Data Center Bridging protocols and uses approximately 10 watts per port. The 1Gbps/10Gbps auto-sensing port provides an easy migration path to higher speeds as data centers ramp up 10Gb Ethernet. The F-series is prices at $35,000 list price and will be available in Q3 of 2010.
You May Also Like
Maximizing cloud potential: Building and operating an effective Cloud Center of Excellence (CCoE)
September 10, 2024Radical Automation of ITSM
September 19, 2024Maximizing Manufacturing Efficiency with Real-Time Production Monitoring
September 25, 2024