rfc1245.ps

著名的RFC文档,其中有一些文档是已经翻译成中文的的.

7 X
V
0 X
([Moy]) 72 73 T
([Page 9]) 499.7 73 T
72 108 540 684 R
7 X
V
3 F
0 X
(3.2  Link bandwidth) 72 674.67 T
0 F
-0.02 (In this section we attempt to calculate how much link bandwidth is consumed by the OSPF \337ood-) 72 648 P
(ing process. The amount of link bandwidth consumed increases linearly with the number of ) 72 634 T
(advertisements present in the OSPF database.W) 72 620 T
(e assume that the majority of advertisements in ) 300.88 620 T
(the database will be AS external LSAs \050operationally this is true, see [1]\051.) 72 606 T
(From the statistics presented in Section 3.1, any particular advertisement is \337ooded \050on average\051 ) 72 580 T
(every 30 minutes. In addition, three advertisements \336t in a single packet. \050This packet could be ) 72 566 T
(either a Link State Update packet or a Link State Acknowledgment packet; in this analysis we ) 72 552 T
(select the Link State Update packet, which is the lar) 72 538 T
(ger\051. An AS external LSA is 36 bytes long. ) 320.93 538 T
(Adding one third of a packet header \050IP header plus OSPF Update packet\051 yields 52 bytes. T) 72 524 T
(rans-) 515.59 524 T
(mitting this amount of data every 30 minutes gives an average rate of 23/100 bits/second.) 72 510 T
-0.05 (If you want to limit your routing traf) 72 484 P
-0.05 (\336c to 5% of the link\325) 247.03 484 P
-0.05 (s total bandwidth, you get the following ) 345.75 484 P
(maximums for database size:) 72 470 T
72 434.01 540 442 C
72 439.98 540 439.98 2 L
0.5 H
0 Z
0 X
0 K
N
0 0 612 792 C
5 F
0 X
0 K
(T) 72 445.33 T
(ABLE 2. Database size as a function of link speed \0505% utilization\051) 77.93 445.33 T
(Speed) 180 423.34 T
(# external advertisements) 288 423.34 T
1 F
(9.6 Kb) 180 406.34 T
(2087) 288 406.34 T
(56 Kb) 180 390.34 T
(12,174) 288 390.34 T
0 F
-0.46 (Higher line speeds have not been included, because other factors will then limit database size \050like ) 72 365.01 P
-0.12 (router memory\051 before line speed becomes a factor) 72 351.01 P
-0.12 (. Note that in the above calculation, the size of ) 315.32 351.01 P
-0.06 (the data link header was not taken into account. Also, note that while the OSPF database is likely ) 72 337.01 P
(to be mostly external LSAs, other LSAs have a size also. As a ballpark estimate, router links and ) 72 323.01 T
-0.01 (network links are generally three times as lar) 72 309.01 P
-0.01 (ge as an AS external link, with summary link adver-) 287.18 309.01 P
(tisements being the same size as external link LSAs.) 72 295.01 T
(OSPF consumes considerably less link bandwidth than RIP) 72 269.01 T
(. This has been shown experimentally ) 355.51 269.01 T
(in the NSI network. See Jef) 72 255.01 T
(frey Bur) 203.69 255.01 T
(gan\325) 243.77 255.01 T
(s \322NASA Sciences Internet\323 report in [3].) 264.42 255.01 T
3 F
(3.3  Router memory) 72 221.67 T
0 F
-0.1 (Memory requirements in OSPF are dominated by the size of the link state database. As in the pre-) 72 195.01 P
(vious section, it is probably safe to assume that most of the advertisements in the database are ) 72 181.01 T
(external LSAs. While an external LSA is 36 bytes long, it is generally stored by an OSPF imple-) 72 167.01 T
-0.34 (mentation together with some support data. So a good estimate of router memory consumed by an ) 72 153.01 P
(external LSA is probably 64 bytes. So a database having 10,000 external LSAs will consume ) 72 139.01 T
(640K bytes of router memory) 72 125.01 T
(. OSPF de\336nitely requires more memory than RIP) 213.79 125.01 T
(.) 452.98 125.01 T
FMENDPAGE
%%EndPage: "9" 10
%%Page: "10" 10
612 792 0 FMBEGINPAGE
72 702 540 720 R
7 X
0 K
V
0 F
0 X
(RFC 1245) 72 712 T
(OSPF protocol analysis) 249.36 712 T
(July 1991) 493.02 712 T
72 69.05 540 81 R
7 X
V
0 X
([Moy]) 72 73 T
([Page 10]) 493.7 73 T
72 108 540 684 R
7 X
V
0 X
-0.35 (Using the Proteon P4200 implementation as an example, the P4200 has 2Mbytes of memory) 72 676 P
-0.35 (. This ) 510.38 676 P
-0.02 (is shared between instruction, data and packet buf) 72 662 P
-0.02 (fer memory) 310.78 662 P
-0.02 (. The P4200 has enough memory to ) 366.26 662 P
(store 10, 000 external LSAs, and still have enough packet buf) 72 648 T
(fer memory available to run a rea-) 367.58 648 T
(sonable number of interfaces.) 72 634 T
(Also, note that while the OSPF database is likely to be mostly external LSAs, other LSAs have a ) 72 608 T
-0.06 (size also. As a ballpark estimate, router links and network links consume generally three times as ) 72 594 P
(much memory as an AS external link, with summary link advertisements being the same size as ) 72 580 T
(external link LSAs.) 72 566 T
3 F
(3.4  Router CPU) 72 532.67 T
0 F
(Assume that, as the size of the OSPF routing domain grows, the number of interfaces per router ) 72 506 T
(stays bounded. Then the Dijkstra calculation is of order \050n * log \050n\051\051, where n is the number of ) 72 492 T
(routers in the routing domain. \050This is the complexity of the Dijkstra algorithm in a sparse net-) 72 478 T
(work\051. Of course, it is implementation speci\336c as to how expensive the Dijkstra really is.) 72 464 T
(W) 72 438 T
(e have no experimental numbers for the cost of the Dijkstra calculation in a real OSPF imple-) 82.36 438 T
(mentation. However) 72 424 T
(, Steve Deering presented results for the Dijkstra calculation in the \322MOSPF ) 169.45 424 T
(meeting report\323 in [3]. Steve\325) 72 410 T
(s calculation was done on a DEC 5000 \05010 mips processor\051, using ) 212.9 410 T
(the Stanford internet as a model. His graphs are based on numbers of networks, not number of ) 72 396 T
(routers. However) 72 382 T
(, if we extrapolate that the ratio of routers to networks remains the same, the ) 154.78 382 T
(time to run Dijkstra for 200 routers in Steve\325) 72 368 T
(s implementation was around 15 milliseconds.) 285.87 368 T
-0.46 (This seems a reasonable cost, particularly when you notice that the Dijkstra calculation is run very ) 72 342 P
(infrequently in operational deployments. In the three networks presented in Section 3.1, Dijkstra ) 72 328 T
-0.35 (was run on average only every 13 to 50 minutes. Since the Dijkstra is run so infrequently) 72 314 P
-0.35 (, it seems ) 493.06 314 P
-0.02 (likely that OSPF overall consumes less CPU than RIP \050because of RIP\325) 72 300 P
-0.02 (s frequent updates, requir-) 413.95 300 P
(ing routing table lookups\051.) 72 286 T
(As another example, the routing algorithm in MILNET is SPF-based. MILNET\325) 72 260 T
(s current size is ) 456.42 260 T
-0.02 (230 nodes, and the routing calculation still consumes less than 5% of the MILNET switches\325 pro-) 72 246 P
(cessor bandwidth [4]. Because the routing algorithm in the MILNET adapts to network load, it ) 72 232 T
(runs the Dijkstra process quite frequently \050on the order of seconds as compared to OSPF\325) 72 218 T
(s min-) 499.7 218 T
(utes\051. However) 72 204 T
(, it should be noted that the routing algorithm in MILNET incrementally updates ) 144.79 204 T
(the SPF-tree, while OSPF rebuilds it from scratch at each Dijkstra calculation) 72 190 T
(OSPF\325) 72 164 T
(s Area capability provides a way to reduce Dijkstra overhead, if it becomes a burden. The ) 104 164 T
-0 (routing domain can be split into areas. The extent of the Dijkstra calculation \050and its complexity\051 ) 72 150 P
(is limited to a single area at a time.) 72 136 T
FMENDPAGE
%%EndPage: "10" 11
%%Page: "11" 11
612 792 0 FMBEGINPAGE
72 702 540 720 R
7 X
0 K
V
0 F
0 X
(RFC 1245) 72 712 T
(OSPF protocol analysis) 249.36 712 T
(July 1991) 493.02 712 T
72 69.05 540 81 R
7 X
V
0 X
([Moy]) 72 73 T
([Page 11]) 493.7 73 T
72 108 540 684 R
7 X
V
3 F
0 X
(3.5  Role of Designated Router) 72 674.67 T
0 F
(This section explores the number of routers that can be attached to a single network. As the num-) 72 648 T
-0.36 (ber of routers attached to a network grows, so does the amount of OSPF routing traf) 72 634 P
-0.36 (\336c seen on the ) 469.48 634 P
(network. Some of this is Hello traf) 72 620 T
(\336c, which is generally multicast by each router every 10 sec-) 238.01 620 T
-0.07 (onds. This burden is borne by all routers attached to the network. However) 72 606 P
-0.07 (, because of its special ) 429.77 606 P
-0.08 (role in the \337ooding process, the Designated router ends up sending more Link State Updates than ) 72 592 P
(the other routers on the network. Also, the Designated Router receives Link State Acknowledg-) 72 578 T
-0.15 (ments from all attached routers, while the other routers just receive them from the DR. \050Although ) 72 564 P
(it is important to note that the rate of Link State Acknowledgments will generally be limited to ) 72 550 T
(one per second from each router) 72 536 T
(, because acknowledgments are generally delayed.\051) 226.38 536 T
-0.22 (So, if the amount of protocol traf) 72 510 P
-0.22 (\336c on the LAN becomes a limiting factor) 228.71 510 P
-0.22 (, the limit is likely to be ) 424.24 510 P
(detected in the Designated Router \336rst. However) 72 496 T
(, such a limit is not expected to be reached in ) 305.68 496 T
(practice. The amount of routing protocol traf) 72 482 T
(\336c generated by OSPF has been shown to be small ) 286.62 482 T
-0.11 (\050see Section 3.2\051. Also, if need be OSPF\325) 72 468 P
-0.11 (s hello timers can be con\336gured to reduce the amount of ) 268.43 468 P
(protocol traf) 72 454 T
(\336c on the network. Note that more than 50 routers have been simulated attached to a ) 131.4 454 T
(single LAN \050see [1]\051. Also, in interoperability testing 13 routers have been attached to a single ) 72 440 T
(ethernet with no problems encountered.) 72 426 T
-0.02 (Another factor in the number of routers attached to a single network is the cutover time when the ) 72 400 P
-0.17 (Designated Router fails. OSPF has a Backup Designated Router so that the cutover does not have ) 72 386 P
-0.31 (to wait for the new DR to synchronize \050the adjacency bring-up process mentioned earlier\051 with all ) 72 372 P
-0.43 (the other routers on the LAN; as a Backup DR it had already synchronized. However) 72 358 P
-0.43 (, in those rare ) 473.46 358 P
-0.33 (cases when both DR and Backup DR crash at the same time, the new DR will have to synchronize ) 72 344 P
(\050via the adjacency bring-up process\051 with all other routers before becoming functional. Field ) 72 330 T
-0.44 (experience show that this synchronization process takes place in a timely fashion \050see the OARnet ) 72 316 P
(report in [1]\051. However) 72 302 T
(, this may be an issue in systems that have many routers attached to a sin-) 183.42 302 T
(gle network.) 72 288 T
-0.15 (In the unlikely event that the number of routers attached to a LAN becomes a problem, either due ) 72 262 P
(to the amount of routing protocol traf) 72 248 T
(\336c or the cutover time, the LAN can be split into separate ) 251 248 T
(pieces \050similar to splitting up the AS into separate areas\051.) 72 234 T
3 F
(3.6  Summary) 72 200.67 T
0 F
(In summary) 72 174 T
(, it seems like the most likely limitation to the size of an OSPF system is available ) 128.85 174 T
-0.4 (router memory) 72 160 P
-0.4 (. W) 142.43 160 P
-0.4 (e have given as 10,000 as the number of external LSAs that can be supported by ) 158.39 160 P
(the memory available in one con\336guration of a particular implementation \050the Proteon P4200\051. ) 72 146 T
-0.09 (Other implementations may vary; nowadays routers are being built with more and more memory) 72 132 P
-0.09 (. ) 534.09 132 P
FMENDPAGE
%%EndPage: "11" 12
%%Page: "12" 12
612 792 0 FMBEGINPAGE
72 702 540 720 R
7 X
0 K
V
0 F
0 X
(RFC 1245) 72 712 T
(OSPF protocol analysis) 249.36 712 T
(July 1991) 493.02 712 T
72 69.05 540 81 R
7 X
V
0 X
([Moy]) 72 73 T
([Page 12]) 493.7 73 T
72 108 540 684 R
7 X
V
0 X
(Note that 10,000 routes is considerably lar) 72 676 T
(ger than the lar) 275.31 676 T
(gest \336eld implementation \050BARRNet; ) 347.37 676 T
(which at 1816 external LSAs is still very lar) 72 662 T
(ge\051.) 283.65 662 T
(Note that there may be ways to reduce database size in a routing domain. First, the domain can ) 72 636 T
-0.19 (make use of default routing, reducing the number of external routes that need to be imported. Sec-) 72 622 P
(ondly) 72 608 T
(, an EGP can be used that will transport its own information through the AS instead of rely-) 98.54 608 T
-0.21 (ing on the IGP \050OSPF in this case\051 to do transfer the information for it \050the EGP\051. Thirdly) 72 594 P
-0.21 (, routers ) 498.11 594 P
(having insuf) 72 580 T
(\336cient memory may be able to be assigned to stub areas \050whose databases are drasti-) 131.41 580 T
(cally smaller\051. Lastly) 72 566 T
(, if the Internet went away from a \337at address space the amount of external ) 172.82 566 T
(information imported into an OSPF domain could be reduced drastically) 72 552 T
(.) 418.67 552 T
(While not as likely) 72 526 T
(, there could be other issues that would limit the size of an OSPF routing ) 162.17 526 T
(domain. If there are slow lines \050like 9600 baud\051, the size of the database will be limited \050see Sec-) 72 512 T
(tion 3.2\051. Dijkstra may get to be expensive when there are hundreds of routers in the OSPF ) 72 498 T
(domain; although at this point the domain can be split into areas. Finally) 72 484 T
(, when there are many ) 418.69 484 T
(routers attached to a single network, there may be undue burden imposed upon the Designated ) 72 470 T
(Router; although at that point a LAN can be split into separate LANs.) 72 456 T
FMENDPAGE
%%EndPage: "12" 13
%%Page: "13" 13
612 792 0 FMBEGINPAGE
72 702 540 720 R
7 X
0 K
V
0 F
0 X
(RFC 1245) 72 712 T
(OSPF protocol analysis) 249.36 712 T
(July 1991) 493.02 712 T
72 69.05 540 81 R
7 X
V
0 X
([Moy]) 72 73 T
([Page 13]) 493.7 73 T
72 108 540 684 R
7 X
V
2 F
0 X
(4.0  Suitable envir) 72 673.33 T
(onments) 195.21 673.33 T
0 F
-0.14 (Suitable environments for the OSPF protocol range from lar) 72 646 P
-0.14 (ge to small. OSPF is particular suited ) 359.11 646 P
(for transit Autonomous Systems for the following reasons. OSPF can accommodate a lar) 72 632 T
(ge num-) 497.84 632 T
(ber of external routes. In OSPF the import of external information is very \337exible, having provi-) 72 618 T
-0.39 (sions for a forwarding address, two levels of external metrics, and the ability to tag external routes ) 72 604 P
-0.29 (with their AS number for easy management. Also OSPF\325) 72 590 P
-0.29 (s ability to do partial updates when exter-) 343.17 590 P
(nal information changes is very useful on these networks.) 72 576 T
(OSPF is also suited for smaller) 72 550 T
(, either stand alone or stub Autonomous Systems, because of its ) 220.44 550 T
(wide array of features: fast conver) 72 536 T
(gence, equal-cost-multipath, T) 235.96 536 T
(OS routing, areas, etc.) 382.3 536 T
2 F
(5.0  Unsuitable envir) 72 469.33 T
(onments) 212.98 469.33 T
0 F
-0.22 (OSPF has a very limited ability to express policy) 72 442 P
-0.22 (. Basically) 304.62 442 P
-0.22 (, its only policy mechanisms are in the ) 354.25 442 P
(establishment of a four level routing hierarchy: intra-area, inter) 72 428 T
(-area, type 1 and type 2 external ) 374.52 428 T
(routes. A system wanting more sophisticated policies would have to be split up into separate ) 72 414 T
(ASes, running a policy-based EGP between them.) 72 400 T
FMENDPAGE
%%EndPage: "13" 14
%%Page: "14" 14
612 792 0 FMBEGINPAGE
72 702 540 720 R
7 X
0 K
V
0 F
0 X
(RFC 1245) 72 712 T
(OSPF protocol analysis) 249.36 712 T
(July 1991) 493.02 712 T
72 69.05 540 81 R
7 X
V
0 X
([Moy]) 72 73 T
([Page 14]) 493.7 73 T
72 108 540 684 R
7 X
V
2 F
0 X
(6.0  Refer) 72 673.33 T
(ence Documents) 137.87 673.33 T
0 F
(The following documents have been referenced by this report:) 72 646 T
([1]) 72 626 T
(Moy) 108 626 T
(, J., \322Experience with the OSPF protocol\323, RFC 1246, July 1991.) 129.88 626 T
([2]) 72 608 T
(Moy) 108 608 T
(, J., \322OSPF V) 129.88 608 T
(ersion 2\323, RFC 1247, July 1991.) 193.85 608 T
([3]) 72 590 T
(Corporation for National Research Initiatives, \322Proceedings of the Eighteenth Internet ) 108 590 T
(Engineering T) 108 576 T
(ask Force\323, University of British Columbia, July 30-August 3, 1990.) 176.11 576 T
FMENDPAGE
%%EndPage: "14" 15
%%Page: "15" 15
612 792 0 FMBEGINPAGE
72 702 540 720 R
7 X
0 K
V
0 F
0 X
(RFC 1245) 72 712 T
(OSPF protocol analysis) 249.36 712 T
(July 1991) 493.02 712 T
72 69.05 540 81 R
7 X
V
0 X
([Moy]) 72 73 T
([Page 15]) 493.7 73 T
72 108 540 684 R
7 X
V
2 F
0 X
(Security Considerations) 72 673.33 T
0 F
(Security issues are not discussed in this memo.) 72 646 T
2 F
(Author) 72 617.33 T
(\325) 122.04 617.33 T
(s Addr) 126.77 617.33 T
(ess) 173.13 617.33 T
0 F
(John Moy) 72 590 T
(Proteon Inc.) 72 576 T
(2 T) 72 562 T
(echnology Drive) 87.48 562 T
(W) 72 548 T
(estborough, MA 01581) 82.36 548 T
(Phone: \050508\051 898-2800) 72 522 T
(Email: [e-mail]) 72 508 T
FMENDPAGE
%%EndPage: "15" 16
%%Trailer
%%BoundingBox: 0 0 612 792
%%Pages: 15 1
%%DocumentFonts: Times-Roman
%%+ Times-Bold