Shamim Which Routing N49
Shamim Which Routing N49
Shamim Which Routing N49
IS-IS was not designed from the start as an IP routing protocol Adjacency is reported once two-way connectivity has been ensured IS-IS essentially uses its regular flooding techniques to synchronize neighbors Coarse database granularity makes this easy (just a few CSNPs)
Intended to minimize transient routing problems by ensuring that a newborn router has nearly complete routing information before it begins carrying traffic Accounts for a significant portion of OSPFs implementation complexity
Transient routing issues can be reduced (albeit non deterministically) by judicious use of the overload bit
Encapsulation
IS-IS runs directly over L2 (next to IP) Sort of makes sense (ISIS was originally designed for CLNS)
Relies on IP fragmentation for large LSAs Subject to spoofing and DoS attacks (use of authentication is strongly advised)
Which Routing Protocol? 2010 Cisco Systems, Inc. All rights reserved.
Terminology
Terminology
OSPF:
Host Link Router Packet
ISIS:
Intermediate System (IS) Protocol Data Unit (PDU) Designated IS (DIS) N/A (no BDIS is used) Link-state PDU (LSP) IIH PDU
Designated router (DR) Backup DR (BDR) Hello packet Link-state advertisement (LSA) Database Description (DBD)
Terminology (cont.)
OSPF:
LS update Area
ISIS:
LS acknowledgement Non-backbone area Backbone area Virtual link Router ID Area Border Router (ABR)
LSP (ISIS runs over layer-2) Subdomain (area) Level-1 area Level-2 area L1L2 router
Advertising router ID
2010 Cisco Systems, Inc. All rights reserved.
Packets
Packets
OSPF basic header is fixed 20 bytes
Version Type Packet Length
1 1 1 1 1 1 1 1
PDU Type
Authentication
Authentication
2010 Cisco Systems, Inc. All rights reserved.
Packets
Packet Encoding
OSPF is efficiently encoded Holy 32-bit alignment provides tidy packet pictures, but not much else Positional fields
Packet Encoding
Extensible from the start (unknown types ignored but still flooded) All packet types are extensible Nested TLVs provide structure for more granular extension (though base spec does not use them; OSPF is starting to do so)
9
No particular alignment
Unrecognized LSA types not flooded (though opaque LSAs can suffice, if implemented universally)
Which Routing Protocol? 2010 Cisco Systems, Inc. All rights reserved.
Packets OSPF
1. Hello 2. DBD 5 type of basic packets
ISIS
1. Hello (3 types L1 LAN, L2 LAN, Point-to-point) 3. Sequence number packet (CSNP, PSNP) 3 types of basic packets granularity within 2. Link state packet (L1,L2)
10
Hello
OSPF:
Fixed format Sent every 10 sec by default.
ISIS:
TLVs (extendable) Sent every 10 secs by default
11
OSPF LSAs
Type
1 2 3 4 5 6 7 8 911
LSA
Router Network Summary Network Summary ASBR External Group Membership NSSA External Attributes Opaque
ISIS LSPs
Purpose
Neighbor announcement Authentication Extended neighbor info(TE) Internal IP Routing info NLPID announcement (IP) External IP Routing info IP Interface addresses Wide scale metrics
12
Adjacency Establishment
OSPF:
LSDB synchronisation is performed before a neighbor is reported in the router-LSA Adjacency is reported once twoway connectivity has been ensured Point-to-point links are treated the same way as in OSPF
ISIS:
On point-to-point links adjacencies are established between every pair of neighbors that can see each other On LAN segments adjacencies are established with the DR and BDR MTU mismatch is detected
On LAN segments, adjacencies are established with the DIS (no BDIS is elected) MTU mismatch is detected
13
Database Granularity
LSAs are mostly numerous and small (one external per LSA, one summary per LSA) Network and Router LSAs can become large
LSAs grouped into LSUpdates during flooding LSUpdates are built individually at each hop
Always flooded intact, unchanged across all flooding hops (so LSP MTU is an architectural constant--it must fit across all links) Small topology changes always yield entire LSPs (though packet size turns out to be much less of an issue than packet count) Implementations can attempt clever packing
Small changes can yield small packets (but Router, Network LSAs can be large)
14
Designated Routers
Both protocols elect a designated router on multiaccess networks to remove O(N^2) link problem (by creating a pseudonode) and to reduce flooding traffic (DR ensures flooding reliability)
In IS-IS all routers are adjacent (but adjacency is far less stateful) If DR dies, new DR must be elected, with short connectivity loss (synchronization is fast)
OSPF elects both a DR and a Backup DR, each of which becomes adjacent with all other routers
BDR takes over if DR fails DRship is sticky, not deterministic Complex algorithm
DRship is deterministic (highest priority, highest MAC address always wins) DRship can be made sticky by cool priority hack (DR increases its DR priority)
15
DR Election
Every LAN interface goes through the Waiting state to listen if the DR and BDR are already elected, if so, the new router does not try to pre-empt
OSPF:
Interfaces also go through a delay (3 seconds), but this is just an attempt to collect as much info for DR election as possible
ISIS:
New router attached to a DR/BDR re-election segment may cause DR happens only when current switch-over DR/BDR goes down (stability)
16
LAN Flooding
OSPF uses multicast send, unicast ack from DR Reduces flood traffic by 50% (uninteresting) IS-IS uses multicast LSP from all routers, CSNP from DR Periodic CSNPs ensure databases are synced
Requires per-neighbor state (for retransmissions) Interesting (but complex) acknowledgement suppression Flood traffic grows as O(N)
17
Multiple areas
If backbone is attached, it In ISIS multi-area has is an ABR and attracts been added - multiple inter-area traffic ISIS processes If no backbone is attached, the router is internal to more than one area and does not attract inter-area traffic
One of the processes will be L1L2 to advertise all area addresses from all processes into L2 Designed to use for CLNS, not for IP
This is Cisco-specific, OSPF standard says more than one area, youre an ABR See RFC 3509 for more details
Which Routing Protocol? 2010 Cisco Systems, Inc. All rights reserved.
18
19
Area types
OSPF has ordinary, stub, totally-stub, NSSA (with and without summaries)
ISIS originally supported areas with no inter-area routes (NSSA, no-summary), now it allows for route leaking (more like NSSA)
20
21
Standard specifies aggregation to be done only when summaries are created based on intra-area routes Inter-area routes can further be aggregated by ABRs when reannounced from the backbone (CSCXXXX)
22
Route leaking was added to ISIS to solve the problem--good filtering capability
23
Type-5 LSAs are used to TLV 130 is used to announce announce external routes by external routing information, several externals share the ASBRs, one LSA per one external same LSP fragment route Every L1L2 router re-announces ABRs announce location of it to L2 (and back to L1 if route ASBRs in type-4 LSAs leaking is configured) Only one copy of LSA per domain Remote areas have as many (type-5s are flooded throughout copies of a TLV as many L1L2 the whole domain except for stub routers are leaking it from L2 and NSSA areas) into these areas Administrative tags may be set in OSPF when an external route is injected into the OSPF domain External routes are differentiated with internal ones No administrative tags External routes look just like internal in the routing table, only L1 and L2 are differentiated
External routing
Number of neighbors
Both protocols can maintain hundreds of neighbors (whether its a good idea is a different question)
ISIS has been deployed with more neighbors in the field (people didnt want areas)
25
Scalability Issues
26
Scalability Issues
Database Size
OSPF topologies limited by Network and Router LSA size (max 64KB) to O(5000) links IS-IS topologies limited by LSP count (256 fragments * 1470 bytes) for all route types
External and Interarea routes are essentially unbounded
27
Scalability Issues
Database Churn
Both protocols have time-limited database entries and therefore require refreshing OSPF age (counts up) has an architectural lifetime limit of 1 hour (80,000 LSAs yield a refresh every 23 milliseconds) Do-not-age LSAs are not backward compatible Dont inject zillions of routes into your IGP IS-IS lifetime field is 16 bits, giving 18.7-hour lifetimes (with refresh times close to this)
28
Scalability Issues
Flooding load--the only serious issue
Link failure: information Full-mesh topologies are worst-case for both
N^2 copies of each update (each of which is O(N) in size) Router failure: information
IS-IS mesh group hack provides backward-compatible way of pruning flooding topology OSPF has interface blocking
29
OSPF v3
30
Explicit support for multiple instances per link Use of IPv6 link-local addresses Authentication method changes
Packet format, LSAs header format changes Handling of unknown LSA types
31
IPv6 uses the term "link" instead of network or subnet to indicate communication Multiple IPv6 subnets can be assigned to a single link, and two nodes can talk directly over a single link, even if they do not share a common IPv6 subnet
Change affects the receiving of OSPF protocol packets, and the contents of Hello Packets and Network-LSAs
32
Mechanisms for neighbor discovery and adjacency formation Interface types LSA flooding and aging
P2P, P2MP, Broadcast, NBMA, Virtual
OSPFv3 has the same 5 packet type but some fields have been changed.
33
The high-order three bits of LS type {1 bit (U) for handling unrecognized LSA and two bits (S2, S1) for flooding scope} encode generic properties of the LSA, while the remainder, (called LSA function code) indicate the LSA's specific functionality OSPFv2 had two flooding scope, AS wide and area wide. OSPFv3 has three flooding scope:
AS scope, LSA is flooded throughout the AS Area scope, LSA is flooded only within an area Link-local scope, LSA is flooded only on the local link.
Which Routing Protocol? 2010 Cisco Systems, Inc. All rights reserved.
34
S1 0 1 0 1
Flooding scope Link-Local flooding scope Area flodding scope AS flooding scope Reserved
35
ISIS extension
36
X: External origin bit S: Sub-TLV present Prefix length: Length of prefix 8 bits
37
For hello PDU interface address must use link local IPv6 address assigned to the interface For LSP non-link local address must be used
38
All interfaces configured with IS-IS for both protocols must support both of them
Otherwise, consider Multi-Topology IS-IS (separate SPF)
Which Routing Protocol? 2010 Cisco Systems, Inc. All rights reserved.
Cant be configured on MPLS/TE since IS-ISv6 extensions for TE are not yet defined
IPv6 configured tunnel wont work, GRE should be used in this configuration
39
Mechanism that allows IS-IS, used within a single domain, to maintain a set of independent IP topologies
IPv4
Introduction
Multicast
40
The problem
Current IS-IS spec and implementation forces all protocols carried by IS-IS to agree on a common Shortest Path Tree Single SPT means congruent topologies
Single SPF run for all protocols
Single SPT means all links need to understand all address families present in the domain
41
42
Two methods
Multi-Topology
Single ISIS domain with set of independent IP topologies
Common flooding and resource associated with both router and network Multiple SPF Large Database Multiple instance of protocol on a given link
Multi-instance
Enhances the ability to isolate the resources associated with both router and network
43
Two methods
OSPF currently is based on multi-instance
Adding multi topology is very easy for OSPFv3 Multiple address family support is already there just minor extension for multi-topology needs to be added Multi-topology support has been there for a while Multi-instance draft is there for ISIS now Depends who you talk to
ISIS
44
Convergence
45
Convergence
Convergence depends on several factors: - failure detection - change propagation - initial wait for SPF computation - time to run SPF
46
Convergence Considerations
The IGPs Will Compete over Processor Cycles Based on Their Relative Tuning If you configure the IPv4 and IPv6 IGPs the same way (aggressively tuned for fast convergence), naturally expect a doubling of their stand alone operation convergence time If the IPv6 IGP is operating under default settings, the convergence time for the optimally tuned IPv4 IGP is not significantly affected
47
48
49
Conclusion
50
Conclusions
OSPF is much more widely understood
Broadly deployed in enterprise market Many books of varying quality available Preserves our investment in terminology
Folks who build very large, very visible networks are comfortable with it
51
Conclusions
For all but extreme cases (large full-mesh networks), protocols are pretty much equivalent in scalability and functionality Stability and scalability are largely artifacts of implementation, not protocol design
Familiarity and comfort in both engineering and operations is probably the biggest factor in choosing
52
Conclusions
Does the world really need two protocols?
Nearly complete overlap in functionality means (ironically) that few people are motivated to switch Entrenched constituencies (large ISPs; everyone else) ensure that installed bases will continue to exist
As long as there are two, people will never agree on only one
53
Questions?
54