ATM VIRTUAL SERVICE CONNECTION

 

BACK TO	ATM TABLE OF CONTENTS
	ATM VP/VC

 

The following aspects will be considered:

• VP, VC channels concepts
• Transport service quality contracts
• Multiservice traffic management
• Traffic and congestion control

ATM VP & VC CHANNELS

 The different notions of Physical Link, Transmission Path, Virtual Path and Virtual Channel are explained below. 

The physical layer can be on one hand as simple as a pair of wires and on the other hand as sophisticated as the SONET/SDH network made of optical links, add & drop multiplexes, cross connects. An end-to-end connection on such a network is called a transmission path (TP) and is composed of several digital sections interconnected through active devices such as cross connects. The quality of this physical transmission path is controlled via a special OAM (Operation, Administration & Maintenance) data flow so called " F3 ".

In ATM as in Packet networks (IP, X25 or Frame Relay), the connection between two users is identified by a logical address distinct from the physical address. When those users move, they can keep the same logical address, the network resolving the logical address translation to the new physical addresses. In addition, virtual connections based on logical addresses can share the same physical link or transmission path by virtual connections multiplexing.
In ATM two levels of virtual connections are provided : the Virtual Path (VP) and the Virtual Channel (VC). In each case a virtual connection is made of virtual links interconnected via devices such as multiplexes, cross connects (CC) or switches (SW).

As a result a VCC (Virtual Channel Connection) is made up of interconnected VCLs (Virtual Channel Link), a VPC (Virtual Path Connection) of VPLs (Virtual Path Link).
The Virtual Path and Channel have their dedicated OAM flows, respectively "F4" and "F5", at the link level (segment F4 or F5 flow) and the connection level (end-to-end F4 or F5). Another representation of TP, VP and VC is highlighted on the next figure.

A few comments on the previous figure:  1. a VCC is identified by linked VCLs; e.g. VCC (red) = VCL (VC=20, VP=1, TP=1) + VCL (VC=10, VP=1, TP=3)  2. a VPC is identified by linked VPLs; e.g. VP (green) = VPL (VC=3, TP=1) + VPL (VP=4, TP=3)  3. on a given physical link TP, a VPL is uniquely identified by the VPI value, a VCL by the pair (VCI,VPI); e.g. there is no ambiguity between the "red" VCL and the "brown" VCL, which share the same VCI value (20) but are on different VPs (1 and 2 respectively).  4. The values VCI and VPI have only local significance (on a given TP); e.g. the green VP has different VPI values on TP1 and TP3.  5. At the crossing point between two TPs, the VPs can be directly cross connected if they transport the same VCs (case of the "green" VP). If not the VC's within the input VP must be demultiplexed, cross connected (or switched) and remultiplexed on the output VP.

Why two virtual levels of connections VP & VC ?
The VP level is equivalent to an "end-to-end line" connecting two distant users. The VP connections will be provided by public or private operators. They will support the end-user VC connections that interconnect terminals and applications within a private ATM VC network.

VP & VC connections characteristics:
The following types of connections are possible:
- permanent (PVP, PVC) or switched (SVP, SVC), with SVP quite an uncommon feature,
- point to point or multipoint, symmetric or assymetric, multicast or unicast
A connection not only identifies the traffic source and destination but also the traffic Quality Of Service (QOS) guaranteed by the network. Therefore between the same source and destination there might be several VC connections (one for data, one for voice, one for video).
The QOS traffic guarantee is a distinctive feature of ATM compared to IP, X25 and even Frame Relay, which indeed makes possible the integrated networking of data, voice and video and multimedia networking. We will now expand on this very important aspect.

 TRANSPORT SERVICE QUALITY CONTRACTS

ATM as the universal transport medium for data, voice, video, must be able to support many end user applications with different transmission requirements (throughput, transit delay, error rate, continuous real time flow) and at the same time be "bandwidth management efficient" (statistical multiplexing) to reduce cost of transmission.
To avoid an unmanageable profusion of transport services dedicated to all those all different user applications, a limited set of services classes (for the time being for) has been defined by the ITU and the ATM Forum. We reuse here the ATM forum definitions.

Transport service classes	Application Types
CBR (Constant Bit Rate)	Circuit emulation, Real Time constant bit rate audio & video
Rt-VBR (Real Time -Variable Bit Rate)	Packetized variable bit rate audio & video (teleconferencing, multimedia applications)
(nrt)-VBR (non real time)-Variable Bit Rate	Connection Oriented data transfer - interactive data applications (transaction processing) & Frame Relay transport
UBR (Unspecified Bit Rate)	Connectionless Data Transfers as IP, SMDS
ABR (Available Bit Rate)	Reliable data transfer with variable rate
Notes:  - CBR: similar service to digital leased lines.  - VBR: suitable for bursty traffic that requires bounded transit delay (either average for data or maximum for audio or video) and low cell losses  - UBR: similar to the "best effort" service (IP) with no guarantee on throughput, transit delay and cell loss..  - ABR is "best effort plus" with a minimum guaranteed transmission rate (MCR) and a low cell loss ratio. User source has to adjust the transmission rate according to the available network bandwidth

For each service class, are defined attributes: traffic parameters (so called the traffic descriptor) to be specified by the end-user (within network allowed ranges) and the network guaranteed QOS parameters. Those parameters can optionally be specified by the end-user (within network authorized ranges).
The combination of the traffic class, the traffic descriptor and the QOS parameters is the traffic contract of a VP or VC connection, the basis of the mutual transport agreement between the end-user and the network (or service) operator on that connection.

Traffic contract		Traffic class
		CBR	rt-VBR	VBR	UBR	ABR
Traffic descriptor	PCR	specified	specified	specified	optional	specified
	CDVT	specified .25ms*	specified .25ms*	NS	NS	NS
	SCR	NA	specified	specified	NA	NA
	MBS	NA	specified	specified	NA	NA
	MCR	NA	NA	NA	NA	specified
QOS parameters	CLR (CLP=0)	specified 1.7 E(-10)*	specified 1. E(-8)*	specified 1. E(-7)*	NS	specified
	CLR (CLP=1)	specified	specified	specified	NS	specified
	CTD	Max specified 0.5 ms*	Max specified 5 ms*	Mean specified	NS	NS
	CDV	specified .15ms*	specified .15ms*	NS	NS	NS
* : typical values - NS: not specified  CLR : Cell Loss Ratio - CLP: Cell Loss Priority (1=low priority cell) - CTD: Cell Transit Delay - CDV: Cell Delay Variation  PCR: Peak Cell Rate - CDVT: Cell Delay Variation Tolerance (not a user parameter) - SCR: Sustainable Cell Rate (maximum average rate) - MBS: Maximum Burst Size (maximum number of cells at PCR rate) - MCR: Minimum Cell Rate

 

MULTISERVICE TRAFFIC MANAGEMENT

Multiple connections of different service classes are multiplexed as shown on the next figure within the link bandwidth. It is assigned first to the CBR connections, at the PCR values, then the VBRs ,at the SCR, the ABRs, at MCR. Additional Bandwidth will be reserved for the VBRs to sustain PCR rate during the MBS bursts. No reservation is normally done for UBRs. They will use the bandwidth left-over by the other connections taking benefit on the statistical multiplexing gain but without any guarantee.

As shown above multiple service-classes VC connections can be multiplexed on a VP connection. The service-class of the VP connection must at least be equal to the highest VC service-class (CBR VP for multiplexed CBR/VBR/UBR VCs). Multiple service-classes VPCs can be multiplexed on the same physical link or the transmission path (TP).

Several traffic control policies, summarized in the following table, are followed within an ATM network in order to fulfill the traffic contracts of all the subscribers.

Policies	Connection	Transmission
Traffic Control	Call Admission Control (CAC)	User Parameter Control (UPC)
		Traffic shaping
		Rate based Flow Control (ABR)
Congestion Control	Call Blocking  Call disconnect	Selective Cell discard
		Explicit Forward Congestion control (EFCI)
		Resource management (RM) cell

 

• The UPC is implemented at the network user connection to ensure that the user traffic meets the traffic contract: cells exceeding the PCR or the average SCR rates will be discarded, among the remaining, those exceeding the MBS burst size are tagged as low priority cells (CLP=1).
• The Traffic Shaping is used to minimize occurrence of overloads when several input links are multiplexed on an output link and minimize cell jitter. It smoothens (by ime spreading) the different flows generated on the link, still fulfilling the flow traffic contracts.

Note : Combining buffering, multiplexing and traffic shaping provide the best performances.

• The Rate based Flow Control mechanism defined in ABR regulates the traffic flow sent by the source according to the network load
• When congestion is about to occur, defense mechanisms are used to avoid it :

. the Selective Cell discard where CLP=1 cells are first distroyed and then CLP=0 cells of less priority traffics (UBR)
. the EFCI indication to signal to the end user the network congestion ; the network input traffic must then be reduced
. the RM cell in ABR to explicitly require lower a user input traffic rate
• The traffic and congestion control is also done at the connection level, where a new user will be accepted (via the CAC) only if the network can withstand the user required traffic contract. If the network starts to be congested selective call disconnection will be done.

BACK TO	ATM TABLE OF CONTENTS
	ATM VP/VC