These
are the first set of definitions of terms. Please comment or give additional
term definitions to promote a common understanding in our future
discussions. Some of
them are possibly voted on in the Orlando meeting , May 14-18, 2001.
The
final list will be included in the standard.
Best regards, Harmen
DELAY DEFINITIONS FOR MAC
STUDIES
Propagation
delay: Time required for
a packet to travel over the medium (for fiber this is 5 ms/km).
Ring latency: Time required for a packet to propagate once around the
ring
Queueing delay: Time between the arrival of an end
of packet at the MAC transmit buffer and the instant that this
packet becomes the head-of-the-line packet in the transmit buffer. This delay
is only caused by the node's own traffic.
Medium access delay:
Time required for a head-of-the-line
packet in the MAC transmit buffer to gain access to the medium. This
delay is only caused by the medium competition and the fairness
mechanism, not by the node's own traffic. This delay does not include the
packet transmission time.
Packet transmission time:
Time required to clock a packet onto
to the medium. This time calculates as t-packet [s] = packet-length [bit] /
bit rate on the medium [bit/sec].
Transit node delay: Time required to transfer an immediate node of the ring
between source and destination. It consists of a constant packet handling time
and a variable insertion or transit buffer delay.
Insertion buffer delay:
Time required for a packet to pass
through the insertion buffer operating in cut-through mode.
Transit buffer delay:
Time required for a packet to pass
through the transit buffer operating in store-and-forward
mode
Receive buffer delay:
Time between the arrival of a
begin-of-packet at the MAC receive buffer and the instant that this
packet is completely delivered to the next protocol layer.
Ring end-to-end delay:
Time required for a packet to travel
from a source to a destination node on the same ring. It consists of
the packet transmission delay, all transit node delays, and the
propagation delay from source to destination.
MAC end-to-end delay:
Time between the arrival of an end of
packet at the MAC transmit buffer of the source node and the time that this
packet is completely delivered to the next protocol layer of the destination
node on the same ring.
DELAY DEFINITIONS FOR INTERACTIVE
REAL-COMMUNICATIONS
Compression delay: Time required
to reduce the amount of the original information from an interactive
real-time (synchronous) source such as live video.
Packetization delay:
Time required to fill a
packet with information from an interactive
real-time (synchronous) source. For a 64 kbit/s voice source this is one
byte per 125 ms.
Protocol stack delay:
Time required to handle packets in the
protocol layers above the MAC.
Decompression delay: Time required to obtain the original information format from
the received packet(s) before relaying it to the acoustical and/or video
equipment.
Delay jitter: Delay variation of the packet transfer caused by the
queueuing and access delays in the source node, all transit node delays, and
the receive buffer delay in the destination node
Playout buffer delay:
Enforced delay at the receive side for
interactive real-time communication to achieve a constant end-to-end
delay. The appropriate delay value is calculated from the delay
jitter, whereby the calculation depends on the application.
User end-to-end
delay:
Total time delay between two users or
applications. It is the sum of all time components above the
MAC, those time components outside the considered ring, and the MAC
end-to-end delay between source and destination on the considered
ring.
DEFINITIONS ON MAC BUFFERS AND THEIR
OPERATION MODES
Transmit buffer: MAC buffer that
contains the packets waiting to be transmitted over the medium
Receive buffer: MAC buffer that
receives the packets addressed to the node
Insertion buffer: MAC
buffer operating in cut-through mode and being part of the
transmission path of the ring.
Transit buffer: MAC buffer operating in store-and-forward mode and being
part of the transmission path of the ring.
Cut-through mode: Operation mode to handle the MAC buffer in the transmission
path of the ring with the purpose to hold up an upstream packet for the
time that the node is transmitting a packet from its transmit buffer. Thus,
the filling of the insertion buffer is not necessary a complete packet.
Assuming that the insertion buffer has priority over the transmit buffer, then
the possibly partly buffered packet is immediately pulsed out again on
the medium. The additional insertion-buffer delay given by the amount of
data that had to be held up is then experienced by all passing packets until
the insertion buffer can be emptied during the absence of data on that part of
the ring.
Store-and-forward mode:
Operation mode to handle the MAC
buffer in the transmission path of the ring with the purpose
to buffer each transit packet completely before relaying it to the next node.
MAC Buffer scheduling:
Scheduling strategy within the MAC to
decide whether to transmit a packet from the node's transmit buffer or a
packet from the insertion/transit buffer. In case of ring QoS classes,
there are a number of priority buffers, both in the transmit and receive
parts as well as at the insertion/transmit buffer part.
Packet preemption: Operation to preempt a packet of a lower priority being
clocked out from the transmit or insertion/transit buffer in order to expedite
the higher priority packet. Preempting the lower priority is not destructive,
so that the preempted and preempting packets are both received at the next
ring node.
Ring QoS classes: Number of service classes that are supported on the ring by
the MAC. Each ring class has its receive buffer, its
insertion/transmit buffer, and its transmit buffer within the
MAC.
FAIRNESS PROTOCOL
DEFINITIONS
Simultaneous access: Nodes
geographically distributed around the ring are able to access the ring
simultaneously. The fundamental mechanisms are destination removal (stripping)
and the use of a buffer in each node on the ring transmission path
operated as cut-through or store-and-forward.
Destination removal: Method that
destination nodes remove the received packet from the ring.
Destination stripping: destination removal.
Spatial reuse: Simultaneous use
of different geographical parts of the ring. This is possible because of
destination removal (stripping).
Fairness protocol: Medium access control protocol to
ensure to all competing nodes have fair access to the medium. Each ring
is controlled independently.
Global fairness: Fairness based
on a mechanism that allows nodes to share the same amount of the
transmission capacity of the ring, independently whether their traffic
interfere or not
Local fairness: Fairness based
on a mechanism that coordinates the ring access of only those nodes that
interact during their packet transfers. Therefore, all nodes that do
not interfere are not throttled in their performance as is in the case of
global fairness mechanisms.
Bottleneck-link fairness:
Fairness based on a mechanism that coordinates the ring access
of only those nodes that use the same links for their packet
transfers.
Flow fairness: Fairness based on
a mechanism that coordinates ring access to individual traffic flows instead
of nodes.
Farness cycle: Constant or
dynamic control period of the fairness mechanism.
Rate control: Access control
method in which sources or flows periodically obtain transmission credits
(e.g. in number of bytes).
Backpressure control: Control
method to stop or throttle the data flow from the upstream node. On
a dual ring the control packet is sent on the counter-rotating
ring.
Round-trip delay: time
required for a control packet to reach it destination and the instant
that the control becomes effective.
GENERAL DEFINITIONS
Medium Access Control (MAC):
Function for each ring of a node for the purpose of coordinating medium access
between distributed nodes that compete for transmission on that
ring. For a dual ring each node has two MACs.
Unicast: Packets
are delivered to a single destination node.
Multicast: Packets
are delivered to a number of destination nodes.
Broadcast: Packets
are delivered to all destination nodes.
Quality-of-Service (QoS):
Service quality that has to be guaranteed in terms
of throughput, end-to-end delay, delay jitter, packet loss, and service
availability.
Connection-oriented: Form of
packet communication with a previous connection set-up to obtain a
virtual (logical) connection between source and destination.
Connectionless: Form of packet
communication without connection set-up.
Packet: Unit of transmission on
the medium (I assume this would fit better for RPR as frame)
Dual ring: Ring network
consisting of two counter-rotating rings comprising a number of nodes
interconnected by point-to-point transmission links. Nodes normally
select the clockwise or the counter-clockwise ring according to the shortest
path, i.e. the minimum number of transmission hops to their
destination.
Multiple rings: Ring network
consisting of more than two rings.
Source node: Node to which
the origin of the communication is attached
Destination node: Node to
which the destination of the communication is attached
Upstream node: Node
located before the considered node on the ring in data flow
direction.
Downstream
node: Node
located after the considered node on the ring in data flow
direction.
Port: Ingress/Egress attachment
of a node.
------------------------------------------------------------------
Prof.Dr.
Harmen R. van As Institute of
Communication Networks
Head of
Institute
Vienna University of Technology
Tel
+43-1-58801-38800
Favoritenstrasse 9/388
Fax
+43-1-58801-38898 A-1040
Vienna, Austria
http://www.ikn.tuwien.ac.at
email: Harmen.R.van-As@xxxxxxxxxxxx
------------------------------------------------------------------