MPLS

Question 1

MPLS extended name

Answer 1

multi-protocol label switching

Question 2

Why MPLS was born? Which was the situation before it?

Answer 2

before MPLS the network was made of lots of protocols and components that were independent. That was causing a difficult management of the network

Question 3

What is the general schema of MPLS compared to the onion?

Answer 3

optical switching (to let IP run at a high speed) + IP on MPLS.

There is only one control plane so IP and MPLS share it

Question 4

Which are the other versions of MPLS?

Answer 4

MPlambaS: for optical networks
GMPLS: general, for any kind of network

Question 5

What is the extended name of ATM?

Answer 5

Asynchronous transfer mode

Question 6

On what a IP network and a MPLS network are based? What are the differences?

Answer 6

IP: on addresses. To find the right route there is the longest prefix matching -> bad because the table has to be scanned each time. IPs are unique

MPLS uses label: labels have to be unique only on a link so each. label can identify multiple destinations. they are used as index so there is the direct access -> good

Question 7

What type of network is a MPLS network?

Answer 7

switched circuit network with virtual circuits

Question 8

Which are the physical components of a MPLs network?

Answer 8

• Label Switching Routers LSRs: they perform packets switching with the label
• Label Edge Router: they insert or remove the label
• Label Switching Path = tunnel-> MPLS introduces a CONNECTION-ORIENTED PARADIGMN THE IP NET

Question 9

Which are the key components of MPLS?

Answer 9

• header (shim)
• enhanced routing protocols: they have to do the routing exchanging routing informations too so MPLS looks for the shortest path based on constraints
• label distribution protocols: done with signaling

Question 10

How a MPLS header is made?

Answer 10

• shim header: it can have multiple protocols
• generally a module has the label field (20), the Experimental bits field (3), the S (BOTTOMOF STACK, IF IT IS SET TO 1 IT MEANS THAT THATIS. THE LAST MODULE) (1) and the TTL (8). S is a stack so new modules are added and removed from the head of the stack

Sometimes Layer 2 header has a Label field that can be used for the MPLS header without adding a separated MPLS header. When it is used the ATM and FRAME RELAY switches become MPLS switches

Question 11

Which are the phases of a LSP setup? (list)

Answer 11

1. label binding
2. label mapping
3. label distribution

Question 12

LSP setup: label binding

Answer 12

it’s the phase in which the label are chosen by the LSRs.
The label is chosen by the downstream node that communicates its decision to the upstream one

It can be done in two ways, ONE UNSOLICITED AND ONE ON-DEMAND:
- static label binding: the network manager decides and imposes the chosen labels it it a permanent circuit as ATM, not scalable AND NOT INTEROPERABILITY AMON DIFFERENT MANAGING SYSTEMS, IT IS NOT POSSIBLE TO HAVE LSPS THAT PASS THROUGH DIFFERENT NETWORKS

• dynamic label binding: at two levels:
  
  1. data/trafic DRIVEN: decided when the packet arrives
  2. control DRIVEN: decided when a control msg arrives. this can originate a Topology based LSP (UNSOLICITED, THE LSP CREATION IS LINKED TO THE DISCOVERY OF ROUTERS TOWARDS DESTINATIONS) and an Explicit LSP (ON-DEMAND, THE LER STARTS A SIGNALING PROCESS)

Question 13

LSP setup: label mapping

Answer 13

It’s the phase in which there is the association between input label, output label and next-hop label.
- input label: the considerate LSR decides the label and then communicate it to its upstream node
- output label: it is the one communicated by the downstream router
- next hop: it is decided by the considerate LSR and it can be done in two modes: hop-by-hop ROUTING and explicitROUTING

Question 14

LSP setup: label distribution

Answer 14

it is the phase in which router communicate the chosen label FOR A FEC

It cane be done whit three different types of protocol for label distribution:
1. routing protocols: BGP -> only with topology based LSP (DYNAMIC. LABEL BINDING, CONTROL DRIVEN LSP CREATION)
2. LDP (Label distribution protocol): deprecated, comes from CISCO routers, designated FOR THE purpose
3. RSVP (Resource RESERVATION Protocol): DESIGNED FOR ALLOCATION IN INTEGRATED SERVICE NETWORKS, IT CAN SUPPORT QOS, MOST USED AND CREATED BEFORE MPLS

2 and 3 become CR-LDP and RSVP-TE for MPLS

Question 15

What is a FEC?

Answer 15

It is a class at which are associated labels and each packet belonging to a FEC is treated similarly: same path and same label

CoS and QoS can be associated to a FEC

Question 16

CoS and QoS for a FEC

Question 17

Which are the routing protocols for MPLS?

Answer 17

THE ROUTING TABLES ARE CREATED AND THEN USED FOR THE MAPPING.
The used routing protocols are OSPF, IS-IS, BGP-4 (not RIP because it is not good when the network is large).
MPLS USES THESE PROTOCOLS TO EXCHANGE BOTH TOPOLOGY INFOS AND INFORMATIONS TO CONSTRAINT ROUTING DECISIONS

they are modified for Traffic Engineering so that they can carry network informations too.

BGP IS. ONLY USED FOR TOPOLOGY BASED INFOS

Question 18

Which are the routing modes for MPLS?

Answer 18

• Hop-by-Hop ROUTING: EACH ROUTER CHOOSES THE LABEL AND THEN IT ANNOUNCES IT AND THERE IS THE MAPPING
• explicit ROUTING: THE ROUTING IS DONE ON THE CONTROL PLANE AND THE INGRES LSR DECIDES THE WHOLE PATH AND. THEN COMMUNICATES IT. THIS DECISION IS MADE WITH SOME CONSTRAINTS -> THIS ROUTING MODE NEEDS A PROTOCOL TO EXCHANGE DYNAMIC INFOS EVERY TIME SOMETHING CHANGES => IT. USES THE ROUTING PROTOCOLS AND ALSO CR-LDP AND RSVP-TE

Question 19

MPLS protocols are enhanced protocols. What these modifications permit?

Answer 19

• Traffic Engineering
• Synergy with Differentiated Services
• Fast fault recovery
• ASSISTANCE TO HAVE QOS

Question 20

Traffic engineering with IP

Answer 20

it cannot be done because it would be stable. That is due to the fact that the routing is done with the routing tables -> control plane and data plane are directly connected

Question 21

Traffic engineering with MPLS

Answer 21

it can be done because control plane and data plane are separated: the routing is done with the forwarding table.

CONTROLPLANE: OPERATES ON IPS
DATA PLANE: OPERATES ON LABELS

Question 22

MPLS: fast fault recovery. How it is done?

Answer 22

A different LSP between two routers is performed a priori so when there is a fault the router applies another label to not pass through the fault and when it is surpassed the label is removed

IT IS CALLED LINK RE-ROUTING

Question 23

MPLS: label stack hierarchy and scalability

Answer 23

• larger networks
• shorter routing tables
• a label for each network

Question 24

PHP complete name

Answer 24

PENULTIMATE hop Popping

Question 25

PHP

Answer 25

When the last but one router instead of the last one removes the last label. the last router has to be notified that this happened and this can be done in two ways:
- explicit PHP: THE MOST EXTERNAL LABEL IS REMOVED
- implicit PHP: 0 IS WRITTEN INSIDE THE LABEL (NOT JUST S=1???)

Question 26

list of MPLS-based solution

Answer 26

VPN connections based on Layer 2 or Layer 3

Question 27

Layer-2 based MPLS

Answer 27

Both for Ethernet and IP

TO DO

Question 28

Layer-3 based MPLS

Answer 28

Only for Ip

TO DO

Question 29

Which was the network schema before MPLS?

Answer 29

onion

1. optical switching: it was a high-speed backbone (Gbit/s)
2. switched circuit based on sonet/sdh: to lower the speed.
3. ATM: with the first two components resources were used even when not needed so ATM provided a way to manage them. ATM is expensive.
4. Frame Relay: it is difficult do connect end users to ATM so this layer was added