PNNI: How it Works - WINLAB
35
PNNI: How it Works
Transcript of PNNI: How it Works - WINLAB
PNNI: How it Works
This presenta t ion has been generated by theATM Forum for the purpose o f educat ing thepub l i c on ATM Techno logy and the ATMForum’s act iv i t ies.
This presentat ion is the property o f the ATMForum and can on l y be g i ven to ex te rna laud i ences by an au tho r i zed ATM Fo rumAmbassado r . ATM Fo rum Ambassado r ’ sregu lar l y a t tend ATM Forum meet ings , sothey can re late f i rs t hand about Forumactiv i t ies. There are no restr ict ions on howthis mater ia l is used for educat ional purposesto internal audiences.
To request an ATM Forum Ambassador to present th is mater ia l , p lease v is i t our web s i te ,http://www.atmforum.com, contact us via e-mail at info@ a tmforum.com or contact one o four regional of f ices: Headquarters +1.650.949.6700, Asia-Paci f ic Of f ice +81.3.3438.3694,Europe Of f ice +32.2.732.8505.
Topicsl Introductionl Background Conceptsl Key Phase 1 Requirementsl PNNI Routingl PNNI Signallingl Configuring the PNNI Hierarchyl Future Capabilitiesl Summary
What Is a “PNNI” ?
l PNNI: Private “Network-Network” or “Network-Node” Interface
l UNI: User Network Interface
U N I P N N I U N I
U N I P N N I U N I
A T MN e t w o r k
A T MN e t w o r k
- O R -
PNNI and Other ATM Interfaces
l UNI User Network Interface
l NNI Network Node Interfacel B-ICI BISDN Inter-Carrier Interfacel DXI Data eXchange Interface
P r i v a t e
U N I
P r i v a t e
N N I
P u b l i c
U N I
A T M
D X I
B - I C I
M e t r o p o l i s D a t a S e r v i c e s
I n c .
C o u n t r y W i d e C a r r i e r
S e r v i c e s
P u b l i c
N N I
ATM Layer SpecificationNNI Cell Format
UNI and NNI cell header difference:
-NNI: 12-bit VPI field, no GFC
-UNI: 8-bit VPI field, 4-bit GFC
L e g e n d
C L P -
G F C -
V P I -
Ce l l L o s s P r i o r i t y
G e n e r i c F l o w C o n t r o l
V i r tua l Pa th Ident i f i e r
5 By tes
48 Bytes
V i r t u a l P a t h I d e n t i f i e r
V i r t u a l C h a n n e l
I d e n t i f i e r
7 6 5 4 3 2 1 0
P a y l o a d T y p e
I d e n t i f i e rC L P
H e a d e r E r r o r
C h e c k
P a y l o a d
( 4 8 b y t e s )
V i r t u a l P a t h
I d e n t i f i e r
V i r t u a l C h a n n e l I d e n t i f i e r
V i r t u a l C h a n n e l
I d e n t i f i e r
Simple PNNI Solution:Interim Inter-switch Signalling Protocol(IISP)
Purpose: Enables Switched Virtual Circuit (SVC) interoperability in small, static environments.
l Uses UNI 3.0/3.1 Signalling between switches
l Requires manual configuration of static topology and resource tables
l Specification released: 3/95
I I S PI I S P
I I S PI I S PI I S PI I S P
V E N D O R BV E N D O R E
V E N D O R D
V E N D O R C
V E N D O R A
A
Advanced Solution:PNNI 1.0
Purpose : Enables extremely scalable, full function, dynamic,multi-vendor ATM networks.
l Designed to support a multi-vendor, global ATM internet.
l Two key protocols:– PNNI Routing:
– PNNI Signalling:
l Specification released: 3/96
Hierarchical, state-of-the-art routing protocol
Based on Q .2931 , e x t ended as necessa r y .
X.1 X.2
X .1 .1X .1 .2 X .2 .1
X .2 .2
U s e r A(A d d r : X . 1 . 1 . A )
U s e r B(A d d r : X . 2 . 2 . B )
Concept of “Topology StateRouting”
Each node periodically:(1) Exchanges “Hello” packets with directly neighboring nodes.
(2) Constructs “PNNI Topology State Elements” (PTSEs), describing the node and listing links to direct neighbors, as shown above.(3) Floods PTSEs to all other nodes.
Each node uses own view of global topology to compute routes.A hierarchical topology state algorithm is the key to PNNI Routing.
A
S 2
U s e r A U s e r B
S 1 S 2
S 3
S 4
S 5
S 1 S 2
S 1
S 3
S 4
S 4
S 2
S 3
S 5
S 5
B
S 4
S 3
S 3
S 2
S 4
S 5
Concept of “Source Routes”
l Ingress nodes choose a complete path to the destinationl Ingress node then adds full path to the message itselfl Transit nodes simply follow the given path
PNNI Signalling uses a hierarchical version of this concept.
U s e r A U s e r B
S 1 S 2
S 3
S 4
S 5
A d d s
S o u r c e
R o u t e
B
B S 1 S 2 S 4 S 5
Key Phase 1 Requirementsl Support UNI 3.1 capabilities
– Point-to-point and point-to-multipoint connections– QOS classes
l Support some UNI 4.0 (TM and Signalling) capabilities– ABR– Individual QoS parameters
l Support massive scalability
l Support intra- and inter-domain functionality in a single,integrated protocol.
l Leverage existing work– PNNI Signalling:– PNNI Routing: Use Q.2931 as base
Borrow / extend concepts fromOSPF, IS - IS , IDRP
PNNI RoutingKey Concepts
l Addressing– ATM End System Address (AESA) format– ATM Address Prefixes
– Reachability Information
l Hierarchical Aggregation of Information– PNNI Routing Hierarchy– Aggregating Information “Up” the Hierarchy
– Passing Summarized Information Back Down
– Result: How a Switch Views the Global Topology
l Enabling QOS-sensitive routing– Advertising Topology State Parameters
– Call Admission Control (CAC)
Address Format Used in PNNIl ATM End System Addresses (AESAs)
– Based on ISO NSAP
– Come in multiple varieties, most notably DCC (Data Country Code)
ICD (International Code Designator)
E.164 (E.164 address contained in AESA)
– Other AESA formats may be supported
l No direct support of native E.164 numbers– May translate to E.164 AESA in PNNI network
Initial Domain Identifier - Identifies the ‘authority’ responsible for allocating structure of DSP
20 octets
A F I ID I HO - DSP E S I S E L
7 octets
Domain Specific Part - Structured by the authority identified by IDI
D S P
H O - D S P = H i g h e r O r d e r D S PESI = End Sys tem Iden t i f i e r (MAC address )SEL = Se lec to r ( f o r end sys tem use on l y )
Authority and Format Identifier - Identifies what scheme is to follow
AESA Format
ATM Address Prefixes
l Address prefix of length p is first p bits of address
l Prefixes with length 0 <= p <= 152 used tosummarize portions of addressing domain– Represents reachability to all addresses that begin with
stated prefix– Selector (20th) octet has only local significance to end
system, ignored by PNNI routing
l Strictly hierarchical (left-to-right) interpretationl Boundaries can be put at any bit position p
– Allows variable levels of hierarchy
A F I ID I HO - DSP E S I S E L
pp
47.0091.8000.1122.0001.0F01.0001.00000C.5BD9A5.00C5.0079.0000.0000.0000.0000.0000.00A03E.000001.00 (LECS)
47.0091.8000.1122.0001.0F01.0001.001011.BAD524.00
47.0091.8000.1122.0001.0F01.0001.00603E.7B190A.00
47.0091.8000.1122.0001.0F01.0001.001011.BAE247.00
Advertised reachable address prefixes: 47.0091.8000.1122.0001.0F01.0001 summary C5.0079.0000.0000.0000.0000.0000.00A03E.000001
matches summary address
Engineering 1
Reachability Informationl In addition to topology information, each node
advertises list of reachable address prefixes
l PNNI routes to nodes advertising the longest matchingprefix for the destination address
PNNI Routing Hierarchy
l Each switch is (manually) initialized with a full 20-byte addressl Routing hierarchy is then defined recursively:
– Neighboring nodes form Peer Groups based on their longest prefix incommon
– Each peer group then behaves as a Logical Group Node (LGN) , toform (next level) peer group, etc.
CB.1 B.2 B.3
B.3.2
B.3 .1
B.1 .1B.1 .2
B.1 .3
B.2 .1
B.2 .2
C .1
C .2
A.1A.2
A .1 .1 A .1 .2
A .1 .3
A .2 .1
A .2 .2
AB
Aggregation of TopologyInformation
l Within each peer group, nodes (or LGNs) exchange PNNITopology State Packets (PTSPs)
l A Peer Group Leader (PGL) is elected to represent its peergroup as a single LGN at next level
l SVCs are set up as routing control channels between LGNs atthe same level
= P e e r G r o u p
L e a d e r
A B CA
A . 1 A . 2 B .1 B .2 B .3 C
Passing Information “Down”the Hierarchy
l Problem:With PTSPs sent among peer group members alone, a switch
would not know how to route to remote peer groups.
⁺ Solution:
Peer group leaders feed higher-level topology down into peergroup.
⁺ Solution:
Peer group leaders feed higher-level topology down into peergroup.
A B CA
A . 1 A . 2 B .1 B .2 B .3 C
Single Node’s Topology View
l Global topology as seen by node A.1.1
A.1 A
CB.1 B .2 B .3B.3 .2
B.3 .1
B.1 .1B.1 .2
B.1 .3
B.2 .1
B.2 .2
C . 1
C . 2
A . 1A . 2
A .1 .1 A .1 .2
A .1 .3
A .2 .1
A .2 .2
AB
A .1 .1 A .1 .2
A .1 .3
A.2 B C
Advertising QOS-relatedTopology State Parameters
Resource information (e.g. available bandwidth)also needs to be advertised in PTSPs.
l Purpose:– Allows better paths to be chosen
– Makes call admission more efficient
U s e r A U s e r B
S 1 S 2
S 3
S 4
S 5
Se tup(New Cal l )
Wh i chPath?
Ava i lab le Bandwidth?
Ava i lab le Bandwidth?
QOS-Related Information:How Much / How Often to Advertise?
N o Q O S -re la ted s ta te exposed
No adver t i sements sen t
All Q O S -re la ted s ta te exposed
Frequent adver t i sements sent
Pro:• simple• no overhead
Con :• high call blocking rate• “blind” retries waste bandwidth• low network utilization
Pro:• near zero call blocking• highest network utilization• optimal routes possible
Con :• most complex• high overhead
Topology State ParametersAdvertised in PNNI 1.0
l Administrative Weight (AW)
l Maximum Cell Transfer Delay (maxCTD)
l Peak-to-Peak Cell Delay Variation (CDV)
l Cell Loss Ratio for CLP=0 traffic (CLR0)
l Cell Loss Ratio for CLP=0+1 traffic (CLR0+1)l Available Cell Rate (AvCR)
l Cell Rate Margin (CRM)
l Variance Factor (VF)
Parameters adver t ised on per serv ice category bas is(may be di f ferent for CBR, rt- V B R , nrt- V B R , A B R , U B R )
O ptional G C A C p a rameters :
Call Admission Control
l Generic Call Admission Control (GCAC)– Run by a switch choosing a source route
– Determines if a path can (probably) support the call
l Actual Call Admission Control (ACAC)– Run by each switch in the chosen path
– Determines whether or not the switch can support the call
U s e r A U s e r B
S 1 S 2
S 3
S 4
S 5
Se tup
R u n s A C A C
R u n s G C A C
C h o o s e s P a t h
R u n s A C A C
R u n s A C A C
R u n s A C A C
PNNI SignallingKey Conceptsl Complete Source Routing across each level
of hierarchy
l Use of Designated Transit Lists
l “Crankback” and Alternate Path routing
Complete Source RoutingAcross Each Peer Group
⁺ Answer:
Š A t the ingress entry switch of each peer group, a completesource route across the peer group is specified.
Š Switch specifying the source route may or may not be the peergroup leader.
l Question: How is a call setup progressed across thenetwork?
B.1
B.2
B.3
A . 1
A . 2
A .2 .2
A .2 .3
A .2 .1
A .1 .2
A .1 .1
ABU s e r X
U s e r Y
Designated Transit Lists(DTLs)l New Information Element (IE) appended to
SETUP and ADD PARTY messages (to carrySource Route)
⁺ Actua l l y implemented as a “push-down / pop-of f stack”
U s e r A U s e r B
S 1 S 2
S 3
S 4
S 5
S E T U P
D T L
S 1 S 2 S 4 S 5 S E T U P
B.1
B.2
B.3
A . 1
A .2 .2
A .2 .3
A .2 .1
A .1 .2
A .1 .1
U s e r X U s e r Y
A .1 .2A .1 .1
A . 1A . 2
AB
A .2 .3A .2 .1
A . 1A . 2
AB
A .2 .3A .2 .1
A . 1A . 2
AB
AB
B.2B .1
B .3
AB
B.2B .1
B .3
A .1 .2A .1 .1
A . 1A . 2
AB
Designated Transit Lists (Example)
Crankback and Alternate PathRoutingl Question: What happens when a call fails along a specified
source route (e.g. due to insufficient bandwidth)?
l Answer : Call is “cranked back” to originator of the top SourceRoute DTL
– Originator then generates a new Source Route, or
– Originator cranks back call to generator of higher-level source route.
B.1
B.2
B.3
A . 1
A . 2
A .2 .2
A .2 .3
A .2 .1
A .1 .2
A .1 .1
ABU s e r X
U s e r Y
1
2
Configuring the PNNIHierarchyl Hierarchical address assignment is the key
– Minimize configuration Use first level bits of switch address to generate peer
group ID
Use prefix of switch address to generate defaultsummary address prefix
– Minimize number of routes
l Planning ahead is important to avoid renumbering– Can run with fewer levels in the beginning
– Can add levels of hierarchy without renumbering
l PGL/LGN configuration also required
l Lots of other things may be configured
Hierarchical AddressAssignment Example
. . .
C o r p o r a t e A d d r e s s P r e f i x
47 .0091 .8000 .1122
L o n d o n M a i n
4 7 . 0 0 9 1 . 8 0 0 0 . 1 1 2 2 . 0 0 0 1
L o n d o n S o u t h
4 7 . 0 0 9 1 . 8 0 0 0 . 1 1 2 2 . 0 0 0 2
B e r l i n
4 7 . 0 0 9 1 . 8 0 0 0 . 1 1 2 2 . 0 1 0 1
B u i l d i n g A
4 7 . 0 0 9 1 . 8 0 0 0 . 1 1 2 2 . 0 0 0 1 . 0 1
B u i l d i n g O
4 7 . 0 0 9 1 . 8 0 0 0 . 1 1 2 2 . 0 0 0 1 . 0 F
B u i l d i n g O , 1 s t F l o o r
4 7 . 0 0 9 1 . 8 0 0 0 . 1 1 2 2 . 0 0 0 1 . 0 F . 0 1
B u i l d i n g O , 2 n d F l o o r
4 7 . 0 0 9 1 . 8 0 0 0 . 1 1 2 2 . 0 0 0 1 . 0 F . 0 2
l e v e l 5 6
l e v e l 7 2
l e v e l 8 0
l e v e l 7 2l e v e l 7 2
l e v e l 8 8
l e v e l 8 8
l e v e l 8 0
PNNI 1.0 Addenda - Work InProgress
l Leaf Initiated Joinl Multipoint-to-Point Connectionsl Closed User Groupsl Network Call Correlation Identifierl Reroutingl Path and Connection Tracel PNNI Routing Authenticationl Transported Address Stackl Vendor-Specific Signalling Informationl Soft PVC Interworking with Frame Relayl Mobility Extension
– Mobile ATM network linked to fixed ATM infrastructuree.g., an airplane
B
PNNI Augmented Routing(PAR)l Allows distribution of information about non-ATM
services via PNNI
l Examples– Automatic establishment of router overlay, e.g.
Discover routers running OSPF in same subnet Establish full mesh of SVCs (in small networks)
– Server detection (Classical IP, NHRP, MARS, DNS)
l Edge devices not running PNNI can use Proxy PARprotocol to interact with PAR capable switches.
Summaryl Interim Inter-switch Signalling Protocol (IISP)
– Minimal functionality; Small, static environments only
l PNNI 1.0– Full functionality; Any-size, static / dynamic environments
l Key PNNI Routing Concepts– Addressing– Routing Hierarchy
– Topology advertisement and aggregation– Resource advertisement and QOS-sensitive routing
l Key PNNI Signalling Concepts– Full source routing across peer groups (via DTLs)
– Crankback and Alternate Path routing
This concludes thepresentation provided by
The ATM Forum
