Figure 14. Ethernet Growth Path - Dell PowerEdge M1000e Technical Manual

Figure 14. Ethernet growth path

GbE fabrics use single lanes per link over 1000BASE-KX. Initially 10GbE will be supported through
10GBASE-KX4, which uses four lanes per link, as does InifiniBand. As technology matures and
enables high level integration, 10GBASE-KR will replace 10GBASE-KX4 as the 10GbE backplane
transport of choice, enabling 10 Gbps over single lanes within the system midplane. In the future,
multiple lanes of 10GBASE-R per mezzanine will allow expanded Ethernet bandwidth up to 80 Gbps
per fabric mezzanine.
Assuming a full population of GE and 10GbE lanes in Fabrics A, B and C, the M1000e can deliver
theoretical total bandwidth of 8.4 Terabits per second.
The M1000e is designed for full support of all near-, medium-, and long-term I/O infrastructure
needs. While the M1000e system's bandwidth capabilities lead the industry, the M1000e is also
intelligently designed for maximum cost, flexibility, and performance benefit.
While Fabric A is dedicated to the server module LOMs, requiring Ethernet switch or pass-through
modules for I/O slots A1 and A2, Fabrics B and C can be populated with Ethernet, Fibre Channel, or
InfiniBand solutions.
I/O modules are used as pairs, with two modules servicing each server module fabric providing a
fully redundant solution. I/O modules may be pass-throughs or switches. Pass-through modules
provide direct 1:1 connectivity from each LOM/mezzanine card port on each server module to the
external network. Switches provide an efficient way to consolidate links from the LOM or mezzanine
cards on the server modules to uplinks into the network.
PowerEdge M1000e Technical Guide
29