The MLXe chassis is built for capacity. It supports a s 15.36Tb/s back plane, 960Gb/s per slot, with 32 100Gb ports or 256 10Gb/s ports at wire speed. It can aggregate up to 64 10Gb Ethernet ports into a single trunk. The MLXe can also forward 4.8 billion IPv6 packets per second. The MLXe includes Multi-Chassis Trunking (MCT) for active/active failover and can work with older MLX and XMR networking blades. The MLXe is available in 4, 8, 16, and 32 slot chassis starting at $22,245 for the 4-slot chassis.
MCT lets IT connect core and edge switches to multiple upstream switches, so that if an upstream link or switch fails, the other path can be used automatically. Without MCT, having multiple paths can cause loops in the network, leading to problems like broadcast storms. The spanning tree protocol automatically removes loops in the network and when a one link fails, it can enable the backup link automatically. However, that means you can only use half your capacity. MCT puts your entire infrastructure into play.
Until multi-path protocols like the IETF's Transparent Interconnect of Lots of Links (TRILL)--which Cisco is shipping in a pre-standard version for the Nexus platform as FabricPath or the IEEE 802.1aq Shortest Path Bridging--are ratified standards and in shipping products, proprietary solutions like Multi-Chassis Trunking enable multiple active paths simultaneously increasing over all capacity and enabling fail-over. MCT connects two switches over a single physical link or multiple links aggregated together to share state and forwarding information. The linked switches act as a single virtual switch. Cisco does something similar with its virtual port channels (VPC) and virtual switching system (VSS). Brocade says failover is less than one second and targets 200 ms as a goal for every failover.
Brocade is also announcing the 100GbE Blade for the MLXe, MLX, and XMR chassis. The blade supports two 100Gb Ethernet interfaces with 802.3ba protocols for 40 and 100Gb. Cisco's CRS-3 and Juniper's T-1600 routers both have single-interface 100Gb Ethernet cards. The 100 GbE cards are initially targeted at service providers, which are already blowing past 40Gb Ethernet links and looking at 100Gb, as well as the largest data centers reaching 10Gb capacity. However, as data center networking starts to consume more bandwidth, 10Gb Ethernet and aggregated 10G links will become saturated and 40 and 100Gb will start to be deployed. The interfaces are C-form factor pluggable (CFP), which are larger than the small form factor pluggable (SFP) interfaces IT is used to with Gb fiber interfaces. They use 10 lanes of 10Gb traffic in both directions, which combined can achieve 100Gb/s. Multi-mode fibre can run 200 meters while single mode can extend to 10 kilometers.