FM6000 Series - 10G and 40G Adaptable Ethernet Switch
The FocalPoint FM6000 series is a family of fully-integrated wire-speed 10G Ethernet switch chips that include enhanced features specifically for the modern data center such as low latency, high scalability, L3 routing, Priority Flow Control (PFC), Enhanced Transmission Selection (ETS), Quantized Congestion Notification (QCN), and FCoE, along with support for interconnect topologies such as TRILL and SPB, as well as emerging virtualized networking standards such as 802.1bg (Edge Virtual Bridging) and 802.1bh (Port Extenders). In addition to enhanced functionality, the FM6000 series contains up to 72 10G Ethernet ports and the ability to group ports as 40G MLD interfaces. All of this is accomplished while maintaining the best-in-class latency and throughput demonstrated by the FM2000, FM3000 and FM4000 families of the FocalPoint product line.
The devices in the FM6000 series that support per-port data rates up to 10G include:
FM6000
Part Number
|
Max
Bandwidth |
SGMII
Ports |
XAUI
Ports |
XFI/KR
Ports |
FM6232 |
320G |
72 |
24 |
32 |
FM6248 |
480G |
72 |
24 |
48 |
FM6264 |
640G |
72 |
24 |
64 |
The devices in the FM6000 series that support per-port data rates up to 40G include:
FM6000
Part Number
|
Max
Bandwidth |
SGMII
Ports |
XAUI
Ports |
XFI/KR
Ports |
MLD/KR4
Ports |
FM6316 |
160G |
72 |
16 |
16 |
4 |
FM6324 |
240G |
72 |
24 |
24 |
6 |
FM6332 |
320G |
72 |
24 |
32 |
8 |
FM6348 |
480G |
72 |
24 |
48 |
12 |
FM6364 |
640G |
72 |
24 |
64 |
16 |
FM6372 |
720G |
72 |
24 |
72 |
18 |
Note: The ports identified as "SGMII" can be independently configured to operate at 10M, 100M, 1G, and 2.5G rates, using the SGMII standard.
New Alta Architecture
The FocalPoint FM6000 series devices are based on the groundbreaking Alta architecture, which is comprised of two key technology blocks:
- RapidArray™ 300nS shared-memory switch element
- FlexPipe™ Adaptable frame processing pipeline
RapidArray
The RapidArray™ 300nS shared-memory switch element can support 720Gbps of simultaneous read and write bandwidth through GHz dual-ported memory and highly-efficient crossbars. With RapidArray, frames are stored only once, providing the lowest possible latency, efficient sharing of a precious shared memory resource, and industry-leading multicast performance.
By implementing a true output-queued shared-memory architecture, RapidArray also delivers outstanding congestion management. With no internal congestion, RapidArray eliminates the need for extra ingress buffers and queues. In addition, multicast packets are stored only once, eliminating the need for additional output queue storage capacity.
The RapidArray memory can be logically divided into multiple shared memory partitions, which, combined with Priority Flow Control, can provide lossless operation for protected traffic classes such as storage. Each egress port can also be logically configured into multiple class-of-service queues that can be scheduled with strict priority or ETS-compliant DWRR, and support over 1 Billion packets transmitted from memory every second. Egress queues can also be individually shaped to limit downstream bandwidth contention. RapidArray also supports a variety of memory watermarks and congestion feedback mechanisms to avoid system-level head-of-line blocking and congestion hot-spots.
FlexPipe
Completely new in the Alta architecture is the FlexPipe™ low-latency frame processing pipeline, which can parse, modify, and apply multiple rules to traffic at over 1 Billion packets per second, in a completely deterministic manner, and can be upgraded in the field to support future data center networking protocols as they emerge. RapidArray can maintain this Billion-pps performance with all of its advanced features enabled (no corner cases).
As a frame arrives, the parser can be configured to place selected header fields into channels that are read, consumed, modified, or regenerated at various stages in the pipeline. The configurable logic has performance similar to fixed-function ASIC blocks, and provides deterministic performance under all conditions. The atomic tables can be configured through the FocalPoint API to support various matching, forwarding and scheduling functions. All of this is achieved at 1 Billion packets per second and less than 300nS cut-through latency.
With a combination of RapidArray 300nS shared-memory switch element and FlexPipe adaptable frame processing pipeline, the Alta architecture offers the highest, most deterministic, and most flexible Ethernet switching performance in the industry.
|
