Echelon LNS - Manual (Page 283)

LNS Programmer's Guide 269
The following section describes some of the other differences in behavior between
standard network interfaces and high performance network interfaces.
Addressing
All devices that communicate on a LonTalk channel must have a Neuron ID and a
domain/subnet/node address. When using a high performance network interface, these
addresses are stored on the PC containing the network interface. Standard network
interfaces store these addresses on the network interface.
When a standard network interface receives a LonTalk message, it forwards the message
to the PC only if the destination address in the message matches one of the network
interface’s addresses. High performance network interfaces must forward all messages
received by network interface to the PC, as the LonTalk protocol stack running on the PC
is responsible for decoding network addresses. This means that high performance
network interfaces will send many messages to the PC that are ultimately discarded by
the PC. This may have performance implications, depending on the link between the
network interface and the PC.
On the other hand, a standard network interface performs more processing than a high
performance interface. This could also affect the overall throughput, depending on the
network interface’s processing speed.
Some network interfaces support making uplink calls when they receive messages
addressed to the interface ( i .LON 10 Ethernet Adapter, i .LON 100 Internet Server, and
the SLTA-10 models). These are all standard network interfaces, since high performance
interfaces do not store their address on the interface. Consult the SLTA-10
documentation for more information on the SLTA-10. For more information on the i .LON
10 Ethernet Adapter and the i .LON 100 Internet Server in this document, see Using
xDriver Interfaces on page 271.
LonTalk Transactions
Sending a message using the request, acknowledged or unacknowledged/repeat
messaging services initiates a LonTalk transaction. An acknowledged or request message
transaction is completed when the acknowledgment or response to the message is
received, or when the transaction times out (based on the message transaction timers
and number of retries). An unacknowledged/repeat message transaction completes when
all of the repeats have been sent.
High performance network interfaces support multiple outstanding transactions for
subnet/node and group addressed messages. This means that LNS can send a network
variable fetch request using subnet node addressing to different devices simultaneously,
without waiting for responses to previously sent requests. This can substantially increase
the total polling throughput. LNS utilizes a secondary source subnet/node address when
sending Neuron ID/broadcast messages so that it can have a single outstanding Neuron
ID or broadcast message transaction, without impacting subnet/node or group addressed
messages. When using a high performance interface, LNS can support thousands of
simultaneous transactions. However, LNS dynamically adjusts the maximum number of
simultaneous transaction when there are signs of network congestion, such as late
responses and message failures. Note that LNS does not support multiple outstanding
message transactions that use the same destination address, and it does not support
multiple outstanding transactions that use Neuron ID or broadcast addressing.