1. Japan Telecom Information & Communication Labs
Achieving Multimedia QOS over Hybrid IP/PSTN Infrastructures: IP Traffic and Congestion Control
April 26, 2001
Susumu Yoneda
Japan Telecom Information & Communication Labs

2. Outline IP Transfer Capabilities Generic Traffic & Congestion Controls
Service models
Traffic descriptors
Conformance definitions
QoS commitments
Generic Traffic & Congestion Controls
Specific Mechanisms e.g., Diffserv, MPLS
Conclusion

3. IP Transfer Capabilities: ITU-T SG13 Draft Rec. Y.iptc
Dedicated Bandwidth (DBW) IP Transfer Capability
Statistical Bandwidth (SBW) IP Transfer Capability
Best-Effort (BE) IP Transfer Capability
IP Transfer Capability: a set of network capabilities provided by IP based network to transfer a set of IP packets under a given classification.

4. Service Models & Traffic Descriptors
DBW
SBW BE
Service
Model
Conforming Packets
Assure the negotiated QoS
Non-conforming Packets
Discarded
Delivered corresponding to the associated QoS
Delivered within the limits of available resources
All Packets
Forwarded by use of available resources
Traffic
Descriptors
Peak Rate,
Peak Bucket Size,
The maximum allowed packet size
DBW’s descriptors + Sustainable Rate, Sustainable Token Bucket Size

5. Conformance Definitions & QoS Commitments
DBW
SBW BE
Conformance
Definition
Packet Arrival
Conforming to the GBRA(Rp,Bp)
Packet Length
Not exceed the maximum allowed packet size
Conforming to the peak GBRA(Rp,Bp) and the sustainable GBRA(Rs,Bs)
QoS
Commitments
Specified Loss and Delay commitments
Include IP QoS Class 0 and 1
Specified Loss commitment
Include IP QoS Class 2
No absolute commitment

6. Generic Traffic & Congestion Controls
Traffic Control Functions
Network Resource Management
Admission Control
Parameter Control
Packet Marking
Traffic Shaping
Packet Scheduling
Congestion Control Functions
Packet Discard Control
Routing (Proposed)

7. Differentiated Services [DiffServ]
Two standard per hop behaviors (PHBs) defined that effectively represent two service levels
Expedited Forwarding (EF): A single codepoint (DiffServ value). EF minimizes delay and jitter and provides the highest level of aggregate quality of service. Any traffic that exceeds the traffic profile (which is defined by local policy) is discarded.
Assured Forwarding (AF): Four classes and three drop-precedences within each class (so a total of twelve codepoints). Excess AF traffic is not delivered with as high probability as the traffic “within profile,” which means it may be demoted but not necessarily dropped.

8. Diffserv Functions (1) Classifier Marker
Behavior Aggregate (BA): Uses only the Diffserv Code Point (DSCP) value
Multi-Field (MF): Uses other header info (like protocol, or port numbers, etc.)
Marker
Adds DSCP when none exists
Adds DSCP as mapped from RSVP reservation
Changes to Map from DSCP to IP TOS, or back
Changes DSCP as local policy dictates

9. Diffserv Functions (2) Meter Conditioner
Accumulates statistics, and provides the inputs to conditioning
Conditioner
Provides queue selection and treatment, policing (shaping traffic) by adding delay or dropping packets in order to conform to the traffic profile described in the SLA with destination or source (depending whether this is an egress or ingress point).
Authenticates the traffic for admission control.

10. MPLS Mechanisms
At the first hop router in the MPLS network, the router makes a forwarding decision based on the destination address (or any other information in the header, as determined by local policy) then determines the appropriate label value -- which identifies the Forwarding Equivalence Class (FEC) -- attaches the label to the packet and forwards it to the next hop.
At the next hop, the router uses the label value as an index into a table that specifies the next hop and a new label. The LSR attaches the new label, then forwards the packet to the next hop.

11. MPLS Routing protocols
Start with existing IGP’s
OSPF
IS-IS
BGP-4
Enhance to carry constraint data
OSPF-TE
IS-IS –TE
Distribute topology information only
The first part of automating the establishment of LSP’s is to have a routing protocol distribute topology information, and the current IGP’s OSPF, IS-IS and RIP do that.
However these protocols simply enable an individual router to decide the port that is on the shortest path to the destination IP address
Traffic engineering is seen as an essential element of running an effective large best effort IP network, and to address the needs of traffic engineering the routing protocols must be enhanced to provide much more data. For example the capacity of all the links between the ingress and egress node, the current utilization of each link, the delay across the link, whether the whole span has protection switching or not and so on. We may also want to set by management some link characteristics ie resource classes that allow the ingress LSR to include or exclude certain resources
Constraint based routing is the key to Traffic engineering
Constraint data
Link capacity,Link utilization
Resource class
Priority
Pre-emption etc
Constraint based routing is the key to Traffic Engineering

12. Label Distribution Protocols
LDP
CR-LDP
RSVP-TE
Hop by Hop routing
Ensures routers agree on bindings between FEC’s and the labels.
Label paths follow same route as conventional routed path
Explicit constraint based routing
Route determined by ingress LSR based on overall view of topology, and constraints
Traffic engineering
CoS and (QoS)
fast (50ms) rerouting

13. MPLS Shim Header Structure
...
Layer 2 Header
IP Packet
Label: 20-bit value, (0-16 reserved)
Exp.: 3-bits Experimental ( ToS)
S: 1-bit Bottom of stack
TTL: 8-bits Time To Live
Label
Exp. S
TTL
4 Octets
Label Switching
Look up inbound label + port (+Exp)
to determine
outbound label + port + treatment
Header operations
Swap (label)
Push (a new header)
Pop (a header from stack)
This is the basic MPLS packet format.
The 32-bit MPLS field is known as a "shim header". This comes from an engineering term - a "shim" is a thin strip of material used to makes parts fit correctly. When you fold up a beer mat to stop a bar table from wobbling you are using a shim!
The first 20 bits of the field actually represent the label. The next three bits are currently "experimental" and must be set to zero. The next bit indicates if this label is part of a stack of labels. If the S-bit is zero, then this is the only label.
The Time To Live (TTL) field is as per a normal IP packet, and is there for the same purpose (loop detection).
MPLS encapsulations are also defined for ATM and Frame relay.

14. Hierarchy via Label stack = Network scalability
Layer 2 Header
Label 3
Label 2
Label 1
IP Packet
Within each domain the IGP simply needs to allow the Boarder (ingress) routers to determine the appropriate egress boarder router
Reducing drastically size of routing table in transit routers
MPLS Domain 1
MPLS Domain 2
Let's look a bit more closely at those labels.
MPLS labels can be stacked one on top of another.
The way that this stacking is build up and stripped off can lead to nested MPLs domains, as I show above...
MPLS Domain 3

15. Dynamic-Bandwidth Setting
traffic
time
Link traffic monitor and dynamic-bandwidth setting.

16. Conclusion
Provide a summary of Y.iptc: IP Transfer Capabilities, Service models, Traffic descriptors, Conformance definitions, QoS commitments
How does it work with many other existing traffic engineering mechanisms?
Traffic engineering as well as congestion controls would work well when traffics are effectively monitored and conformance is checked.
Utilize Y.iptc for the conformance monitoring purposes.