fido2100 3-Port Industrial Ethernet DLR Switch with IEEE1588
Data Sheet
April 10, 2013
Because of cut through forwarding, non-IP frames from all transmit queues and IP frames from
transmit queues 2 and 3 will face a delay of 2.8 µs, other non-IEEE 1588 frames will face a delay of
4.9 µs and IEEE 1588 time synchronization frames will face a delay of 7.3 µs through the fido2100,
when the transmitting port is free of contention. Cut through forwarding is disabled for a transmit port
when the receiving port is operating at 10Mbps speed and the transmitting port is operating at 100Mbps
speed. In this case, store and forward is used.
2.5 Quality of Service
In order to support high performance applications such as CIP motion, the fido2100 enforces quality of
service based on IP Differentiated Services Code Point (DSCP). The fido2100 implements four
prioritized transmit queues per port. Frames received with DSCP 59 are queued in the highest priority
transmit queue (queue #1). Frames received with DSCP 55 are queued in the second highest priority
transmit queue (queue #2). Frames received with DSCP 47 and 43 are queued in the third highest
priority transmit queue (queue #3). Frames received with other DSCP values are queued in the lowest
priority transmit queue (queue #4). In addition, DLR protocol frames are queued in the highest priority
queue 1. When a port is ready to transmit the next frame, the highest priority frame is chosen from the
current set of queued frames for transmission based on strict priority ordering. Within a given priority
queue frames are transmitted in FIFO order.
2.6 Unicast and Multicast Address Filtering
The fido2100 implements 256 entry dynamic unicast MAC address learning and filtering mechanism.
The unicast MAC address learning table is implemented as 4-way learning/look up table of 64 rows.
The 48-bit source MAC address from a frame received on any external port is hashed into 6-bits by
XOR operation of every sixth bit. The hash index identifies the row where one of four locations that is
free is used for MAC address learning along with associated external port. When none of them is free,
the 4th way location is always overwritten. Learning is performed on source MAC address only from
frames with a unicast or multicast destination MAC address. Automatic aging of learned entries will
happen in 4- 6 minutes, unless learning is refreshed within that period.
When a frame with unicast destination address is received through any port, 4-way simultaneous
lookup is performed from entries at hash index location of destination MAC address. If the address is
found, the frame is forwarded only through associated external port. If the address is not found, then it
is forwarded through all other external ports. When a frame with unicast destination address is received
through one of the external ports and the destination address matches host CPU MAC address, then that
frame is forwarded only to internal port.
The fido2100 implements multicast filtering on the internal port for frames received through external
ports and no multicast filtering is implemented for the external ports. The multicast filter consists of a
2048 bin hash table. The received multicast destination address is passed through a 32-bit Ethernet
19
support@innovasic.com
1-505-883-5263
Document #: IA211111101-04
UNCONTROLLED WHEN PRINTED OR COPIED
1-888-824-4184