Westermo U200 Operator manal - Manual (Page 19)

V4.5
www.westermo.com
U/R/T200 series
- 19 -
7 Fast Re-configuration of Network Topology (FRNT)
7.1 Introduction
The Westermo OnTime 200 series is available with redundant ring technology. This
eliminates network failure caused by fibre or copper failures on the trunk ports (ring ports).
The speed of ring recovery is an essential part of designing your network. The Westermo
OnTime ring solution can recover from a failure in only 30mS if such a failure does occur.
When used in conjunction with redundant power supplies a very reliable system can be
designed.
Standard Ethernet networks would collapse and fail if normal office based Ethernet Switches
were formed into a complete ring. This failure is commonly referred to as a ‘broadcast storm’
as Ethernet Packets have multiple routes on a network to communicate to devices. Usually,
an incorrect type of packet broadcasts (or floods) over a network and causes hosts to
respond all at once, typically with wrong responses. This starts the process over and over
again; hence your network crashes.
7.2 FRNT version 0
7.2.1 FRNT version 0 principles
The FRNT version 0 protocol is similar to the IEEE Spanning Tree Protocol (STP) except for
the following:
Each switch in a ring topology has knowledge of the network topology, see figure below. I.e.
not only its neighbouring switches as is the case for STP. A FRNT topology change event
packet will be sent directly to the focal point switch in case of a topology change (e.g. a link
loss or a link establishment), while a STP implementation will only send STP control packets
one network hop. The focal point switch will, based on the received topology change event
packet, generate a topology change command. This packet is sent to each member switch in
the ring. The time it takes from the occurrence of a topology change until the corresponding
topology change event packet is received on the focal point is typical a fraction of millisecond
(ms) or a few milliseconds (ms) at the most, even though there is 50 switches on the path
between the topology change detecting switch and the focal point and the network load on the
links are high (e.g. 50 % of full wire speed). Thus, the switch latency in the no load scenario
is 15 microseconds (µs), while a conservative estimate in case of 50 % load is 70
microseconds (µs). The most time consuming part in case of a topology change is MAC table
update procedure. The MAC tables on each switch must be updated in case of a topology
change. This operation takes approx. 20 milliseconds (ms) and is independent on the
number of switches in the ring.
Note:
Similar proprietary network redundant protocols are often based on polling instead of event
controlled handling of a topology change. This will introduce a slower establishment of a new
topology. Another aspect is link re-training. A proprietary network redundant protocols that
are based on link re-training will suffer from a delay of 300 milliseconds (mS) or more.