1. Scheduling and Quality of
Services (QoS)
Advanced Telecommunication Network
(ET5187)
by
Aris Cahyadi Risdianto
23210016
2. Review Scheduling
• Scheduling and Qos : Qontrolling input to output
• Packet Classification : Same Class (FIFO/LIFO)
or Different Class (Lost/Delay Sensitive, Class)
• Queuing System : Scheduling control
• Loss Sensitive Scheduling : HoL, PBS, POB,
RED
• Buffer size : small assured delay, but loss cell
• Delay Sensitive Scheduling : Upper Bound
Method
3. Review QoS
• QoS as Technological Lever : over installed
resources or controlling traffic in the network
• QoS as Commercial Lever : sub-optimal controlling
resources = loss revenue
• QoS : performance, availability, reliability and
security (L3 QoS inspired by ATM)
• Evolution architecture : Integration IP and ATM
(Dual-Mode, I-PNNI, Ipsilon, IETF-MPLS)
• RSVP : IP Signaling Protocol, Path and
Reservation Messages
• IntServ : guaranteed and controlled-load service
4. Upper Bound Method
• Used for solving CAC (Call Admission Control)
problem
• Some assumption :
o Each arrival process satisfies with certain
business constrain
o Service time for cell/packet is deterministic and
proportional
o Scheduling rule is used to generate QoS for
class k with minimal μk ("fair" rule to prevent
blocking another class getting served)
5. Upper Bound Method (Cont.)
• Queue count is maximum difference between
inflow and outflow (λk and μk)
• If queue > 0, class served by minimal rate (μk)
• Number of queue bounded by burstinest σk
provided if λk ≤ μk
• Buffer size bounded by sum of burstinest all flows,
so loss can be guaranteed
• Maximum delay bounded by burstinest divide by
inflows, so delay can be guaranteed
6. Upper Bound Method (Cont.)
• Remarks on upper bound method :
• Zero packet loss only guaranteed for admitted
packet (satisfied with burstinest constrain), if
not packet will be lost
• Delay guaranteed are deterministic because all
stochastic assumed to be bounded or
deterministic
• Upper Bound Method more optimal than N*D/D/1
queuing for scenario where N not identical and
independent CBR resources
7. Generalized Processor Sharing (GPS)
Differ from fair policy including minimum service rate and
excess capacity allocation
Provide inherent fairness (measurable amount resource
reserved for each class based on weight
Work-conserving discipline, ideal for small amount of data
from I different jobs
8. Generalized Cμ-rules (Gcμ rules)
Powerful, dynamic scheduling rule which view QoS from
different angle such as posses delay function as monetary
cost
Founded from 3 fact in the Queuing theory:
Total workload invariant for work-conserving scheduling
rules
Class workloads “live on the faster time scale” than total
workload process
Well behaved heavy traffic limit systems, class workload
process “Converges”
Distribute total workload over different class to minimize
delay cost rate at each point
9. Generalized Cμ-rules (Gcμ rules)
With “lagrange” optimization problem, the solution defines
as mapping g intepreted as switching curve of Gcμ-rules
parameterized by scalar W
11. Differentiated Services (DiffServ)
• Threat each class differently on per-hop behaviour
(PHB)
• Class differentiation rather than flow differentiation
(more scalable)
• Provide QoS more natural than IntServ which inline
with Internet
• Bandwidth Broker use to managed inter-domain
resources for providing end-to-end QoS
12. Differentiated Class
• IP DSCP format:
• Two different PHB Class, except BE (Best Effort) :
Expedited Forwarding (EF) = virtual leased line or
point-to-point connection
Assured Forwarding (AF) = better best efforf
13. DS Class: Expedited Forwarding (EF)
• Absolute dedicated BW independent from other
• Guaranteed BW for providing low packet loss, low
latency and low jitter
• Implement with Priority Queue and Strict Policing
• EF behavior : departure rate EF traffic must equal
or exceeded configurable rate
• Guaranteed BW means excess traffic must be
discarded (strict policing)
14. DS Class: Assured Forwarding (AF)
• “No Free Lunch”, better service for one class,
expense of other service
• 4 Class with 3 class each based on drop
preferences
• Level forwarding assurance based on resource
allocation, load offered and drop preference
• Implement with weighted Round-robin (WRR),
Weight Fair-Queue (WFQ), and drop technology
(RED/WRED)
15. Shortcut Routing to MPLS
• Traditionally Internet routing create problem,
because size of route, per-packet lookup
burden network, bottleneck
• Solution : Eliminate L3 processing by L2
packet forwarding (Shortcut Routing)
• IP over ATM : mixed CL(connectionless)/CO
(connection oriented) for best effort traffic
• 3 Approach : flow driven, topology driven, and
Explicit shortcut
16. Layered Routing
• Top level routing by IP (OSPF), route between
nodes by ATM layer Routing (I-PNNI)
• ATM change the path based on available resources,
OSPF rediscover low weight link regularly => Hop-
by-hop path different next-hop nodes
• More vulnerable to loop, L2/L3 routing loop is
hidden at both L2/L3
• Transient loop for CO/CL environment
• I-PNNI the ultimate solution, but the standard never
finished
17. Flow Driven Shortcut
• Short messages use CL (connectionless)
because connection setup costly
• Long duration high-traffic use CO (connection-
oriented) for header efficiency
• Pareto Law : 20% flows are long and constitute
of 80% bytes
• Decision between router and switch is
complicated
• Ipsilon Switching : decision based on Ipv4
header (TCP = switch, UDP = route)
18. Topology Driven Shortcut
• Special ATM-VC setup to “shortcut” number of
router
• Integrated switch & router individually decide
to shortcut
• Sources and destination path stored in
“shortcut” forwarding table
• CO/CL forward together with QoS
differentiation
• The approach is Cisco Tag-Switching
19. Multiprotocol Label Switching
(MPLS)
• TDP (Cisco) & LDP (IETF) : signaling protocol
for routing to “shortcut” based on MPLS tag
• Support explicit routing to provide QoS
constrain routing
• Based on LDP, construct label forwarding
table (LIB), similar to ATM VPI/VCI
• Adopt label stack approach, up to 3 labels
including “push”, “pop”, and “swap”
20. Multiprotocol Label Switching
(MPLS) continue..
• Separated control and forwarding with Traffic
Engineering (TE) can mapped into label
• Flexible to form FEC to build VPN for any
other medium didn't support labelling
• Traffic Engineering : redirect, balance and
restoration the path
• “Forwarding with the clue”, the clue give next-
hop downstream router, the current router
end-up with IP lookup
21. Generalized MPLS (G-MPLS)
• Extension of MPLS for other packet switched
as IP packet
• TDM/Optical Lamda can be formed
• Redesign MPLS protocol and optical switching
without optical-electronic conversion
• Extend control plane for legacy equipment:
Simplification O&M
Efficiency and Faster
Higher Flexibility
22. Generalized MPLS (G-MPLS)
Summary
• LMP assigned to manage critical network by
mapping time slot, lambda, or port into label
• Extension to OSPF for advertising availability
of optical resources
• Enhance IP signaling RSVP to setup LSP
accross
• Scalability features such as hierarchical LSPs
23. Generalized MPLS (G-MPLS)
Example
IP Network (left) and
SDH Network (right)
Each SDH has link
capacity of 2 Mbps
Three different
configuration
originate by GMPLS
switching in the SDH
nodes