idnits 2.17.1 draft-ietf-bmwg-ospfconv-term-02.txt: ** The Abstract section seems to be numbered Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- ** Looks like you're using RFC 2026 boilerplate. This must be updated to follow RFC 3978/3979, as updated by RFC 4748. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- ** The document seems to lack a 1id_guidelines paragraph about Internet-Drafts being working documents. ** The document seems to lack a 1id_guidelines paragraph about 6 months document validity. ** The document seems to lack a 1id_guidelines paragraph about the list of current Internet-Drafts -- however, there's a paragraph with a matching beginning. Boilerplate error? ** The document seems to lack a 1id_guidelines paragraph about the list of Shadow Directories -- however, there's a paragraph with a matching beginning. Boilerplate error? == No 'Intended status' indicated for this document; assuming Proposed Standard == The page length should not exceed 58 lines per page, but there was 12 longer pages, the longest (page 2) being 60 lines == It seems as if not all pages are separated by form feeds - found 0 form feeds but 13 pages Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- ** The document seems to lack an Introduction section. ** The document seems to lack a Security Considerations section. ** The document seems to lack an IANA Considerations section. (See Section 2.2 of https://www.ietf.org/id-info/checklist for how to handle the case when there are no actions for IANA.) ** The abstract seems to contain references ([2]), which it shouldn't. Please replace those with straight textual mentions of the documents in question. Miscellaneous warnings: ---------------------------------------------------------------------------- == The document doesn't use any RFC 2119 keywords, yet seems to have RFC 2119 boilerplate text. -- The document seems to lack a disclaimer for pre-RFC5378 work, but may have content which was first submitted before 10 November 2008. If you have contacted all the original authors and they are all willing to grant the BCP78 rights to the IETF Trust, then this is fine, and you can ignore this comment. If not, you may need to add the pre-RFC5378 disclaimer. (See the Legal Provisions document at https://trustee.ietf.org/license-info for more information.) -- The document date (January 2003) is 7772 days in the past. Is this intentional? Checking references for intended status: Proposed Standard ---------------------------------------------------------------------------- (See RFCs 3967 and 4897 for information about using normative references to lower-maturity documents in RFCs) -- Missing reference section? '2' on line 472 looks like a reference -- Missing reference section? '1' on line 469 looks like a reference -- Missing reference section? '3' on line 475 looks like a reference -- Missing reference section? '4' on line 477 looks like a reference -- Missing reference section? '5' on line 480 looks like a reference Summary: 10 errors (**), 0 flaws (~~), 4 warnings (==), 7 comments (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 Network Working Group Vishwas Manral 3 Internet Draft Netplane Systems 4 Russ White 5 Cisco Systems 6 Aman Shaikh 7 Expiration Date: June 2003 University of California 8 File Name: draft-ietf-bmwg-ospfconv-term-02.txt January 2003 10 OSPF Benchmarking Terminology and Concepts 11 draft-ietf-bmwg-ospfconv-term-02.txt 13 1. Status of this Memo 15 This document is an Internet-Draft and is in full conformance with 16 all provisions of Section 10 of RFC2026. 18 Internet Drafts are working documents of the Internet Engineering 19 Task Force (IETF), its Areas, and its Working Groups. Note that other 20 groups may also distribute working documents as Internet Drafts. 22 Internet Drafts are draft documents valid for a maximum of six 23 months. Internet Drafts may be updated, replaced, or obsoleted by 24 other documents at any time. It is not appropriate to use Internet 25 Drafts as reference material or to cite them other than as a "working 26 draft" or "work in progress". 28 The list of current Internet-Drafts can be accessed at 29 http//www.ietf.org/ietf/1id-abstracts.txt 31 The list of Internet-Draft Shadow Directories can be accessed at 32 http//www.ietf.org/shadow.html. 34 2. Abstract 36 This draft explains the terminology and concepts used in [2] and 37 future OSPF benchmarking drafts. 39 3. Conventions used in this document 41 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 42 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 43 document are to be interpreted as described in [1]. 45 4. Motivation 47 This draft is a companion to [2], which describes basic Open Shortest 48 Path First (OSPF [3]) testing methods. This draft explains 49 terminology and concepts used in OSPF Testing Framework Drafts, such 50 as [2]. 52 5. Definitions 54 o Internal Measurements 56 - Definition 58 Internal measurements are measurements taken on the Device 59 Under Test (DUT) itself. 61 - Discussion 63 These measurement rely on output and event recording, 64 along with the clocking and timestamping available on the 65 DUT itself. Taking measurements on the DUT may impact the 66 actual outcome of the test, since it can increase proces- 67 sor loading, memory utilization, and timing factors. Some 68 devices may not have the required output readily available 69 for taking internal measurements, as well. 71 Note: Internal measurements can be influenced by the 72 vendor's implementation of the various timers and process- 73 ing models. Whenever possible, internal measurements 74 should be compared to external measurements to verify and 75 validate them. 77 o External Measurements 79 - Definition 80 External measurements infer the performance of the DUT 81 through observation of its communications with other dev- 82 ices. 84 - Discussion 86 One example of an external measurement is when a down- 87 stream device receives complete routing information from 88 the DUT, it can be inferred that the DUT has transmitted 89 all the routing information available. External measure- 90 ments suffer in that they include not just the protocol 91 action times, but also propagation delays, queuing delays, 92 and other such factors. 94 For the purposes of this paper, external techniques are 95 more readily applicable. 97 o Multi-device Measurements 99 - Definition 101 Multi-device measurements require the measurement of 102 events occurring on multiple devices within the testbed. 104 - Discussion 106 For instance, the timestamp on a device generating an 107 event could be used as the marker for the beginning of a 108 test, while the timestamp on the DUT or some other device 109 might be used to determine when the DUT has finished pro- 110 cessing the event. 112 These sorts of measurements are the most problematic, and 113 are to be avoided where possible, since the timestamps of 114 the devices in the test bed must be synchronized within 115 milliseconds for the test results to be meaningful. Given 116 the state of network time protocol implementation, expect- 117 ing the timestamps on several devices to be within mil- 118 liseconds of each other is highly optimistic. 120 o Point-to-Point links 121 - Definition 123 A network that joins a single pair of routers is called a 124 point-to-point link. For OSPF [3], point-to-point links 125 are those on which a designated router are not elected. 127 - Discussion 129 A point-to-point link can take lesser time to converge 130 than a broadcast link of the same speed because it does 131 not have the overhead of DR election. Point-to-point links 132 can be either numbered or unnumbered. However in the con- 133 text of [2] and [3], the two can be regarded the same. 135 o Broadcast Link 137 - Definition 139 Networks supporting many (more than two) attached routers, 140 together with the capability to address a single physical 141 message to all of the attached routers (broadcast). In the 142 context of [2] and [3], broadcast links are taken as those 143 on which a designated router is elected. 145 - Discussion 147 The adjacency formation time on a broadcast link can be 148 more than that on a point-to-point link of the same speed, 149 because DR election has to take place. All routers on a 150 broadcast network form adjacency with the DR and BDR. 152 Async flooding also takes place thru the DR. In context of 153 convergence, it may take more time for an LSU to be 154 flooded from one DR-other router to another DR-other 155 router, because the LSA has to be first processed at the 156 DR. 158 o Shortest Path First Time 160 - Definition 162 The time taken by a router to complete the SPF process. 164 - Discussion 166 This does not include the time taken by the router to give 167 routes to the forwarding engine. 169 o Measurement Units 171 The SPF time is generally measured in milliseconds. 173 o Hello Interval 175 - Definition 177 The length of time, in seconds, between the Hello Packets 178 that the router sends hello packets on the interface. The 179 typical hello interval is 10 seconds on broadcast net- 180 works, and 30 seconds for point-to-multipoint and point- 181 to-point networks. On multicast capable media, hellos are 182 sent to a multicast address; on non-multicast capable 183 media, they are sent unicast to each neighbor. 185 - Discussion 187 The hello interval should be the same for all routers on a 188 network. 190 Decreasing the hello interval can allow the router dead 191 interval (below) to be reduced, thus reducing convergence 192 times in those situations where the router dead interval 193 timing out causes an OSPF process to notice an adjacency 194 failure. Very small router dead intervals accompanied by 195 very small hello intervals can produce more problems than 196 they resolve, as described in [4] & [5]. 198 o Router Dead interval 200 - Definition 202 After ceasing to hear a router's Hello Packets, the number 203 of seconds before its neighbors declare the router down. 204 The default dead interval is four times the hello inter- 205 val; 40 seconds on broadcast networks, and 120 seconds on 206 non-broadcast networks. 208 - Discussion 210 This is advertised in the router's Hello Packets in the 211 RouterDeadInterval field. The router dead interval should 212 be some multiple of the HelloInterval (say 4 times the 213 hello interval), and must be the same for all routers 214 attached to a common network. 216 o Incremental SPF 218 - Definition 220 The ability to recalculate a small portion of the SPF 221 tree, rather than the entire SPF tree, on receiving notif- 222 ication of a change in the network topology. At worst, 223 incremental SPF should perform no worse than a full SPF. 224 In better situations, an incremental SPF run will rebuild 225 the SPF tree in much shorter time than a full SPF run. 227 6. Concepts 229 6.1. The Meaning of Convergence 231 A network is termed to be converged when all of the devices within 232 the network have a loop free path to each possible destination. Since 233 we are not testing network convergence, but performance for a partic- 234 ular device within a network, however, this definition needs to be 235 narrowed somewhat to fit within a single device view. 237 In this case, convergence will mean the point in time when the DUT 238 has performed all actions needed to react to the change in topology 239 represented by the test condition; for instance, an OSPF device must 240 flood any new information it has received, rebuild its shortest path 241 first (SPF) tree, and install any new paths or destinations in the 242 local routing information base (RIB, or routing table). 244 Note that the word convergence has two distinct meanings; the process 245 of a group of individuals meeting the same place, and the process of 246 a single individual meeting in the same place as an existing group. 247 This work focuses on the second meaning of the word, so we consider 248 the time required for a single device to adapt to a network change to 249 be SR-Convergence, or Single Router Convergence. 251 6.2. Measuring Convergence 253 Obviously, there are several elements to convergence, even under the 254 definition given above for a single device. We will try to provide 255 tests to measure each of these: 257 o The time it takes for the DUT to pass the information about a 258 network event on to its neighbors. 260 o The time it takes for the DUT to process information about a 261 network event and calculate a new Shortest Path Tree (SPT). 263 o The time it takes for the DUT to make changes in its local 264 rib reflecting the new shortest path tree. 266 6.3. Types of Network Events 268 A network event is an event which causes a change in the network 269 topology. 271 o Link or Neighbor Device Up 273 The time needed for an OSPF implementation to recoginize a 274 new link coming up on the device, build any necessarily adja- 275 cencies, synchronize its database, and perform all other 276 needed actions to converge. 278 o Initialization 280 The time needed for an OSPF implementation to be initialized, 281 recognize any links across which OSPF must run, build any 282 needed adjacencies, synchronize its database, and perform 283 other actions needed to converge. 285 o Adjacency Down 287 The time needed for an OSPF implementation to recognize a 288 link down/adjacency loss based on hello timers alone, 289 propogate any information as necessary to its remaining adja- 290 cencies, and perform other actions needed to converge. 292 o Link Down 294 The time needed for an OSPF implementation to recognize a 295 link down based on layer 2 provided information, propogate 296 any information as needed to its remaining adjacencies, and 297 perform other actions needed to converge. 299 6.4. LSA and Destination mix 301 In many OSPF benchmark tests, a generator injecting a number of LSAs 302 is called for. There are several areas in which injected LSAs can be 303 varied in testing: 305 o The number of destinations represented by the injected LSAs 307 Each destination represents a single reachable IP network; 308 these will be leaf nodes on the shortest path tree. The pri- 309 mary impact to performance should be the time required to 310 insert destinations in the local routing table and handling 311 the memory required to store the data. 313 o The types of LSAs injected 315 There are several types of LSAs which would be acceptable 316 under different situations; within an area, for instance, 317 type 1, 2, 3, 4, and 5 are likely to be received by a router. 318 Within a not-so-stubby area, however, type 7 LSAs would 319 replace the type 5 LSAs received. These sorts of characteri- 320 zations are important to note in any test results. 322 o The Number of LSAs injected 324 Within any injected set of information, the number of each 325 type of LSA injected is also important. This will impact the 326 shortest path algorithms ability to handle large numbers of 327 nodes, large shortest path first trees, etc. 329 o The Order of LSA Injection 330 The order in which LSAs are injected should not favor any 331 given data structure used for storing the LSA database on the 332 device under test. For instance, AS-External LSA's have AS 333 wide flooding scope; any Type-5 LSA originated is immediately 334 flooded to all neighbors. However the Type-4 LSA which 335 announces the ASBR as a border router is originated in an 336 area at SPF time (by ABR's on the edge of the area in which 337 the ASBR is). If SPF isn't scheduled immediately on the ABRs 338 originating the type 4 LSA, the type-4 LSA is sent after the 339 type-5 LSA's reach a router in the adjacent area. So routes 340 to the external destinations aren't immediately added to the 341 routers in the other areas. When the routers which already 342 have the type 5's receive the type-4 LSA, all the external 343 routes are added to the tree at the same time. This timing 344 could produce different results than a router receiving a 345 type 4 indicating the presence of a border router, followed 346 by the type 5's originated by that border router. 348 The ordering can be changed in various tests to provide 349 insight on the efficiency of storage within the DUT. Any such 350 changes in ordering should be noted in test results. 352 6.5. Tree Shape and the SPF Algorithm 354 The complexity of Dijkstra's algo depends on the data structure used 355 for storing vertices with their current minimum distances from the 356 source. The simplest structure is a list of vertices currently reach- 357 able from the source. Finding the minimum cost vertex then would take 358 O(size of the list). There will be O(n) such operations if we assume 359 that all the vertices are ultimately reachable from the source. More- 360 over, after the vertex with min cost is found, the algo iterates thru 361 all the edges of the vertex and updates cost of other vertices. With 362 an adjacency list representation, this step when iterated over all 363 the vertices, would take O(E) time. Thus, overall running time is: 365 O(sum(i:1, n)(size(list at level i) + E). 367 So, everything boils down to the size(list at level i). 369 If the graph is linear: 371 root 372 | 373 1 374 | 375 2 376 | 377 3 378 | 379 4 380 | 381 5 382 | 383 6 385 and source is a vertex on the end, then size(list at level i) 386 = 1 for all i. Moreover, E = n - 1. Therefore, running time 387 is O(n). 389 If graph is a balanced binary tree: 391 root 392 / \ 393 1 2 394 / \ / \ 395 3 4 5 6 397 size(list at level i) is a little complicated. First it 398 increases by 1 at each level upto a certain number, and then 399 goes down by 1. If we assumed that tree is a complete tree 400 (like the one in the draft) with k levels (1 to k), then 401 size(list) goes on like this: 1, 2, 3, 403 Then the number of edges E is still n - 1. It then turns out 404 that the run-time is O(n^2) for such a tree. 406 If graph is a complete graph (fully-connected mesh), then 407 size(list at level i) = n - i. Number of edges E = O(n^2). 408 Therefore, run-time is O(n^2). 410 shortest path first algorithm to compute the best paths 411 through the network need to be aware that the construction of 412 the tree may impact the performance of the algorithm. Best 413 practice would be to try and make any emulated network look 414 as much like a real network as possible, especially in the 415 area of the tree depth, the meshiness of the network, the 416 number of stub links verses transit links, and the number of 417 connections and nodes to process at each level within the 418 original tree. 420 7. Topology Generation 422 As the size of networks grows, it becomes more and more difficult to 423 actually create a large scale network on which to test the properties 424 of routing protocols and their implementations. In general, network 425 emulators are used to provide emulated topologies which can be adver- 426 tised to a device with varying conditions. Route generators either 427 tend to be a specialized device, a piece of software which runs on a 428 router, or a process that runs on another operating system, such as 429 Linux or another variant of Unix. 431 Some of the characteristics of this device should be: 433 o The ability to connect to the several devices using both point- 434 to-point and broadcast high speed media. Point-to-point links can 435 be emulated with high speed Ethernet as long as there is no hub or 436 other device in between the DUT and the route generator, and the 437 link is configured as a point-to-point link within OSPF. 439 o The ability to create a set of LSAs which appear to be a logical, 440 realistic topology. For instance, the generator should be able to 441 mix the number of point-to-point and broadcast links within the 442 emulated topology, and should be able to inject varying numbers of 443 externally reachable destinations. 445 o The ability to withdraw and add routing information into and from 446 the emulated topology to emulate links flapping. 448 o The ability to randomly order the LSAs representing the emulated 449 topology as they are advertised. 451 o The ability to log or otherwise measure the time between packets 452 transmitted and received. 454 o The ability to change the rate at which OSPF LSAs are transmitted. 456 o The generator and the collector should be fast enough so that they 457 are not bottle necks. The devices should also have a degree of 458 granularity of measurement atleast as small as desired from the 459 test results. 461 8. Acknowedgements 463 The authors would like to thank Howard Berkowitz (hcb@clark.net), 464 Kevin Dubray, (kdubray@juniper.net), and Randy Bush (randy@psg.com) 465 for their discussion, ideas, and support. 467 9. References 469 [1] Bradner, S., "Key words for use in RFCs to Indicate Requirement 470 Levels", RFC2119, March 1997. 472 [2] Manral, V., "Benchmarking Methodology for Basic OSPF Convergence", 473 draft-ietf-bmwg-ospfconv-intraarea, January 2003 475 [3] Moy, J., "OSPF Version 2", RFC 2328, April 1998. 477 [4] Ash, J., "Proposed Mechanisms for Congestion Control/Failure 478 Recovery in OSPF & ISIS Networks", October, 2001 480 [5] Choudhury, G., et al, "Explicit Marking and Prioritized Treatment 481 of Specific IGP Packets for Faster IGP Convergence and Improved 482 Network Scalability and Stability", draft-ietf-ospf-scalability, 483 April 2002 485 10. Authors' Addresses 487 Vishwas Manral, 488 Netplane Systems, 489 189 Prashasan Nagar, 490 Road number 72, 491 Jubilee Hills, 492 Hyderabad. 494 vmanral@netplane.com 496 Russ White 497 Cisco Systems, Inc. 498 7025 Kit Creek Rd. 499 Research Triangle Park, NC 27709 501 riw@cisco.com 502 Aman Shaikh 503 University of California 504 School of Engineering 505 1156 High Street 506 Santa Cruz, CA 95064 508 aman@soe.ucsc.edu