DiffServ CoS Simulation -- IP VPN Case Study
Whatever the end-to-end mechanisms used for ensuring Quality of Service (QoS)
for SIP individual application flows, this QoS is dependent on the QoS
offered by service providers in the crossed networks. With DiffServ, individual flows
are classified and aggregated when entering the backbone. This white paper analyses the simulation
of classes of services over an IP/MPLS backbone combining DiffServ and Traffic Engineering.
White Paper: Simulating DiffServ Classes of Service over an IP/MPLS Backbone
|
The object of this paper is to present a case study based on a simulation,
in order to
provide a practical perception of how a VPN Service Provider runs and maintains
delay-sensitive applications, such as video or voice, over its IP/MPLS network.
IPVCoSS (IP VPN CoS Simulator) has been developed uniquely for feeding the content
of this White Paper, hence having the full control of traces and performance reports.
Simulating DiffServ Classes of Service
over an IP/MPLS Backbone
-- January 2005 (572Ko, 45 pages)
|
Figure 1: Logical Topology
|
Figure 1 shows an SP (service provider) network that offers VPN services to
3 customers: Red, Blue and Green. The internal backbone links connecting the
edge routers to the core routers are limited to STM-1 in order to cope with the
traffic throughput used in the context of this case study. The other backbone
links between core routers are either STM-1 or STM-4, thus creating the conditions
for over-sizing or under-sizing. This core network topology ensures a minimum of
resiliency since each PE (Provider Edge) router is connected to two P
(Provider core) routers and each P router is in turn connected to two other P
routers. However, the links in dotted lines will not be considered for this case
study, not to complicate the simulation survey.
The distances (not shown) between the backbone routers represent a realistic regional
network, either national or international. We consider primarily MPLS VPN services
over this network but other services such as Internet Transit could be offered
as well. In some scenarios we will introduce extra flows crossing the SP network
and we can consider that these flows could belong to traffic related to other VPN,
or Internet Transit, services.
|
Figure 2 shows the ingress and egress points of the generated IP flows.
We can assimilate them to the source and destination of each micro-flow by
considering that a host (or for instance a video encoder/decoder) is immediately
connected to the CE (Customer Edge) router.
EF (expedited forwarding) flows (V, W, X, Y) are based on UDP transport and are isochronous.
They require very low jitter and their initial throughput must be maintained.
All AF (assured forwarding) flows (M, N, O, P, Q) are based on TCP transport
and therefore TCP slow start will be normally applied as well as
congestion avoidance mechanisms if any. These flows have a fixed packet size.
BE (best effort) flows (A, B, C, D, E) are based on UDP transport and generated in several volume occurrences separated by regular gaps, in order to create traffic bursts
|
Figure 2: Physical Topology and IP Flows
|
Links to WP's References
| [1] |
QoS Support in MPLS Networks (Victoria Fineberg, MPLS/Frame Relay Alliance)
|
| [2] |
Supporting differentiated service classes in Large IP Networks (Chuck Semeria, Juniper)
|
| [3] |
Supporting differentiated service classes: Queue Scheduling Disciplines (Chuck Semeria, Juniper)
|
| [4] |
Supporting differentiated service classes: Active Queue Memory Management (Chuck Semeria, Juniper)
|
| [5] |
Supporting differentiated service classes: TCP Congestion Control Mechanisms (Chuck Semeria, Juniper)
|
| [6] |
Supporting differentiated service classes: MPLS (Chuck Semeria, Juniper)
|
| [7] |
DiffServ - The Scalable End-to-End QoS Model (Cisco)
|
| [8] |
Advanced Topics in MPLS-TE Deployment (Cisco)
|
| [9] |
Virtual Leased Line Services Using Cisco MPLS DiffServ-Aware Traffic Engineering (Cisco)
|
| [10] |
Voice Trunking and Toll-Bypass Trunking Using Cisco MPLS DiffServ-Aware Traffic Engineering (Cisco)
|
| [11] |
NS-2 Network Simulator - http://www.isi.edu/nsnam/ns/
|
| [12] |
RFC 2018 - TCP selective Acknowledgment Options (Mathis, et al.)
|
| [13] |
RFC 2309 - Recommendations on Queue Management and Congestion Avoidance in the Internet (Braden, et al.)
|
| [14] |
RFC 2474 - Definition of the DS Field (Nichols, et al.)
|
| [15] |
RFC 2475 - An Architecture for Differentiated Services (Blake, et al.)
|
| [16] |
RFC 2581 - TCP Congestion Control (Allman, et al.)
|
| [17] |
RFC 2582 - NewReno modification to TCP's Fast Recovery algorithm (Floyd & Henderson)
|
| [18] |
RFC 2597 - Assured Forwarding PHB Group (Heinanen)
|
| [19] |
RFC 2697 - A Single Rate Three Color Marker (Heinanen & Guerin)
|
| [20] |
RFC 2698 - A Two Rate Three Color Marker (Heinanen & Guerin)
|
| [21] |
RFC 2702 - Requirements for Traffic Engineering over MPLS (Awduche, et al.)
|
| [22] |
RFC 3031 - MPLS Architecture (Rosen, et al.)
|
| [23] |
RFC 3086 - Definition of Differentiated Services Per Domain Behaviors (Nichols & Carpenter)
|
| [24] |
RFC 3246 - An Expedited Forwarding PHB (Davie, et al.)
|
| [25] |
RFC 3260 - New Terminology and Clarifications for Diffserv (Grossman)
|
| [26] |
RFC 3270 - MPLS Support of Differentiated Services (Le Faucheur, et al.)
|
| [27] |
RFC 3272 - Overview and Principles of Internet Traffic Engineering (Awduche, et al.)
|
| [28] |
RFC 3290 - An Informal Management Model for Diffserv Routers (Bernet, et al.)
|
| [29] |
RFC 3564 - Requirements for Support of Diffserv Aware MPLS Traffic Engineering (Le Faucheur & Lai)
|
|