adept technology GRouter3, Руководство пользователя

Страница: 52 / 69

time. Decreasing it increases network traffic and load on the router. A practical lower limit is 100
ms.
Timeout Cushion : is sets the time period in milliseconds of latency cushion for the time out
for failure detection of the monitoring packets. In other words if aer a time equal to HeartBeat
Time + Cushion, both monitoring packets are not detected by a router then a monitoring failure
has been deemed to have occurred. e routers then go to an active diagnostic mode. e
cushion should always be less than the HeartBeat Time but greater than the expected latency due
to propagation delays. e default is 200 ms.
AutoSync Time : is sets the time period in milliseconds between automatic synchronization
attempts from the twin to the inactive twin. e default is 5000 ms.
Diagnostic Retrie s: is sets the number of retries that the active diagnostic interrogation
request/response message will use. A diagnostic is sent out from each interface (709 and IP)
whenever a monitoring failure occurs. If the interrogation packet fails aer
Diagnostic Retries
number of retries then a fault of the associated network interface will have been deemed to have
occurred. is will generate an alarm. e default is two retries. If spurious faults occur it may be
because
Diagnostic Retries is too low and the diagnostic responses are getting lost due to
collisions. e odds of lost packets due to collisions decrease significantly for retry counts above
Four.
Initial Arbitration Count : e arbitration count is a 64 bit number. e redundant twins use an
arbitration count encapsulated in the monitoring packets to determine which member of the pair
should be active. e twin with the highest count wins the arbitration and goes active while the
one with the lower count will go inactive. If both have the same count then they both pick
random counts until one wins the arbitration.
On boot up both routers will default to active. e ensuing arbitration will result in one of the
routers going inactive. is menu option can be used to guarantee that a particular router will
win the boot up arbitration on the next reboot. e desired active one should have the higher
Initial Arbitration Count . Use this menu option to set the Initial Arbitration Count appropriately.
e arbitration count is incremented twice per HeartBeat Time . e relative difference between
initial arbitration counts should be set big enough to account for any variable latency in boot up
time. e default is zero. If both nodes are set to zero, which ever node boots up first will go active
and start incrementing its arbitration count. e other node will also go active but because it
booted up later its arbitration count will be lower and will lose the arbitration and go inactive. e
arbitration count will eventually roll over to zero. us on the next arbitration aer roll over the
active and inactive nodes will switch. Given that the arbitration count is a 64 bit number, for a
HeartBeat Time of 1 second and an Initial Arbitration Count of zero, the rollover time is more
than 292 billion years.
To reiterate, the initial arbitration count is only going to have an effect if there is an arbitration on
boot-up. An arbitration only occurs when both nodes are in active forward state. In order to force
the inactive node to be active one must set the arbitration counts on both nodes and then reboot
both nodes.
Powerup in Forward Mode : On boot up both routers will default to active. As a result, they could
both forward packets thereby resulting in a spike of duplicate traffic until arbitration completes.
-52-