CEG5101

Multiprotocol Label Switching

and Data Center Networks
Lecturer:
Mohan Gurusamy
Associate Professor
Electrical and Computer Engineering, NUS
Reference and Source of Contents and
Figures:
J.F. Kurose and K. W. Ross, “Computer
Networking: A Top-Down Approach”,
Pearson Publishers, 7th and other Editions
CEG5101-Notes-11
MPLS-DCN GMohan / NUS 1
MULTI PROTOCOL LABEL
SWITCHING (MPLS)

CEG5101-Notes-11
MPLS-DCN GMohan / NUS 2
MPLS
 Provides a notion of connection-oriented service in IP
networks
 Label switched paths (LSPs) are set up before data
transmission starts
 All packets from a traffic flow or connection will
follow the same path (traverse the same sequence
of routers)
 Facilitates resource (bandwidth) reservation, setting
up a backup connection in the event of a failure
 Facilitates traffic engineering
 Routing is based on just the destination address and also
the source address

CEG5101-Notes-11
MPLS-DCN GMohan / NUS 3
Multiprotocol label switching
(MPLS)

 initial goal: high-speed IP forwarding using fixed

length label (instead of IP address)
 fast lookup using a short fixed length identifier
 borrowing ideas from ATM Virtual Circuit (VC) approach
 but IP datagram still keeps IP address!

Exp: experimental bits, S: end of series of stacked labels

PPP or Ethernet
MPLS header IP header remainder of link-layer frame
header

label Exp S TTL

20 3 1 5
CEG5101-Notes-11
MPLS-DCN GMohan / NUS 4
 MPLS capable routers
 a.k.a. label-switched router
 forward packets to outgoing interface based
only on label value (don’t inspect IP address)

MPLS forwarding table distinct from IP forwarding tables
 flexibility: MPLS forwarding decisions can
differ from those of IP

use destination and source addresses to route
flows to same destination differently (traffic
engineering)

re-route flows quickly if link fails: pre-computed
backup paths (useful for VoIP)
CEG5101-Notes-11
MPLS-DCN GMohan / NUS 5
 MPLS versus IP paths
R6
D
R4 R3
R5
A
R2
 IP routing: path to destination
determined by destination address IP router
alone
 Only one path to a destination from a
given router
 MPLS routing: more than one path
from a router to the same destination
CEG5101-Notes-11
MPLS-DCN GMohan / NUS 6
 MPLS versus IP paths
entry router (R4) can use different MPLS
routes to A based, e.g., on source address
R6
D
R4 R3
R5
A
R2

 IP routing: path to destination IP-only

determined by destination router
address
 MPLS alone
routing: path to destination can be MPLS and
based on source and destination address IP router
• fast reroute: precompute backup

routes in case of link failure

CEG5101-Notes-11
MPLS-DCN GMohan / NUS 7
 MPLS signaling
 modify OSPF, IS-IS link-state flooding protocols to carry info used by MPLS routing,


e.g., link bandwidth, amount of “reserved” link bandwidth
 entry MPLS router uses RSVP-TE signaling protocol to set up MPLS
forwarding at downstream routers

RSVP-TE
R6
D
R4
R5 modified
link state A
flooding
CEG5101-Notes-11
MPLS-DCN GMohan / NUS 8
Example: LSP routing
 Refer figure in next slide
 R1-R4 are label switched routers (LSR)
 R5-R6 are standard IP routers
 Label switched paths (LSP) routed on the paths
 R6-R4-R3-R1-A
 R6-R4-R3-D
 R5-R4-R2-R1-A

CEG5101-Notes-11
MPLS-DCN GMohan / NUS 9
MPLS forwarding tables
in out out
label label dest
interface
10 A 0 in out out
12 D 0 label label dest
interface
8 A 1 10 6 A 1
12 9 D 0

R6
0 0
D
1 1
R4 R3
R5
0 0
A
R2 R1 in out out
label label dest
in out out
interface
label label dest 6 7 A 0
interface
8 6 A 0 CEG5101-Notes-11
MPLS-DCN GMohan / NUS 10
Example: LSP routing
(Contd.)
 Refer figure and example in previous slides
 Add a new path R5-R4-R3-R1-A. Choose the next
available label on the links. State the new entries made
in the LSR’s forwarding tables
 R4 table: <_ , 9, A, 0>
 R3 table: <9, 7, A, 1>
 R1 table: <7,8,A,0>

CEG5101-Notes-11
MPLS-DCN GMohan / NUS 11
MPLS vs SDN

 Will SDN dominate or replace MPLS

in future?

CEG5101-Notes-11
MPLS-DCN GMohan / NUS 12
DATA CENTER NETWORKS

CEG5101-Notes-11
MPLS-DCN GMohan / NUS 13
Data Center Networks
 Data Centers

Large number of hosts (servers, CPU+memory+disk
storage)

Could be hundreds of thousands of servers

Provide content services such as web pages, search
results, email, video streaming

Massively parallel computing infrastructure (eg:
distributed computation for search engines)

Provide cloud computing to other companies to serve
IT needs

Cloud providers: Amazon web services (AWS),
Microsoft Azure, Google Cloud, Alibaba Cloud

Data Center Networks interconnect the hosts
CEG5101-Notes-11
MPLS-DCN GMohan / NUS 14
Data center networks
 10’s to 100’s of thousands of hosts, often closely coupled, in close proximity:

e-business (e.g. Amazon)

content-servers (e.g., YouTube, Akamai, Apple, Microsoft)

search engines, data mining (e.g., Google)

 challenges:
 multiple applications, each
serving massive numbers of
clients
 managing/balancing load,
avoiding processing, networking,
data bottlenecks

Inside a 40-ft Microsoft container,

CEG5101-Notes-11 Chicago data center
MPLS-DCN GMohan / NUS 15
Data center networks
Hierarchical architecture
Each rack has 20-40 blades (hosts)
interconnected by top-of-the-rack ToR
Ethernet switches
10 Gbps, 40 Gbps, 100 Gbps links
Each ToR connects to two Tier-2 switches
Internet
(not shown) to prvide redundancy for high
availability
Border router
Load Load
balancer Access router
balancer

Tier-1 switches
B

A C Tier-2 switches

TOR
switches
Server racks

1 2 3 4 5 6 7 8
CEG5101-Notes-11
GMohan / NUS 16
MPLS-DCN
Load Balancer
 Data center hosts multiple applications
 Each application may be hosted at more than
one server host
 A publicly visible IP address for each application
 Load balancer receives external client requests
 Use destination port number and destination IP address
 Distributes requests to the hosts hosting that
application balancing the load
 Provides a NAT function. What? Why?
 returns results to external client (hiding data center
internals from client)

CEG5101-Notes-11
MPLS-DCN GMohan / NUS 17
Data center networks
 rich interconnection among switches, racks:
• increased throughput between racks (multiple routing paths possible)

• increased reliability (availability) via redundancy

Tier-1 switches

Tier-2 switches

TOR
switches
Server racks

1 2 3 4 5 6 7 8
CEG5101-Notes-11
GMohan / NUS 18
MPLS-DCN