Ethernet in Metro and Long Haul Networks
(continued)

IEEE 802.17 Resilient Packet Ring (RPR)

The Institute of Electrical and Electronics Engineers (IEEE) is working on a new specification that describes how Ethernet can exist in a ring architecture, which is how most metro fiber is currently deployed. 

SONET is a powerhouse if you want dedicated point-to-point bandwidth with a sub-50 ms. protection scheme. However, it is not very efficient when it comes to supporting bursty data communications.  Suppose we use a SONET connection to directly connect two routers using a technique called POS (Packet over SONET).  Now let’s assume that the link is as STS-3c. To the router, the connection looks like a high speed PPP link. In reality, the router must send 2,340 bytes of data, 8,000 times per second, no more, no less. Here is the wasteful part. If the router cannot fill the STS-3c connection with data because it has none to send, it must send “filler” packets and intentionally waste bandwidth because SONET demands that you fill your payload 8,000 times per second.

This waste of bandwidth opened the door for Resilient Packet Ring (RPR) technology.

RPR supports dual counter-rotating ring topology, like SONET.  RPR will offer protection switching in less than 50ms, but it won’t waste half the bandwidth when protection is not needed. Since each RPR node must stand alone, there will not be “master node” like we have seen in other routing and switching protocols.  RPR is not like FDDI, where the device that puts a packet on the ring is the one that ultimately must strip it off. Instead, the frame will be stripped at the destination and it will no longer consume bandwidth on the ring, making downstream segments available for other traffic. This means that RPR will support multicast traffic because any multicast frames will simply drop and continue (to borrow another SONET term). 

The RPR nodes must periodically communicate control messages to one another indicating available unused capacity at a given point in time. RPR nodes process the incoming control messages and determine whether any positive or negative adjustments need to be made to the rates at which each active queue is drained onto the ring. To determine available capacity, each node continually monitors the fill rate of its queues.  When the queues start to clog up, the node signals the upstream nodes to slow down.  

Cisco is marketing their version of RPR, called Dynamic Packet Transport (DPT) and is based on the Spatial Reuse Protocol (SRP), a Cisco-developed MAC-layer protocol that utilizes a dual counter-rotating, ring-based network topology, where both rings carry data and control messages, concurrently, to maximize bandwidth. A control packet is sent with each data packet on counter-rotating rings.  There are no timeslots like SONET TDM tributaries, but rather statistical packet-based multiplexing.

<< Previous page     Next page >>

Page 5 of 7