idnits 2.17.1 draft-ietf-ippm-6man-pdm-option-08.txt: Checking boilerplate required by RFC 5378 and the IETF Trust (see https://trustee.ietf.org/license-info): ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/1id-guidelines.txt: ---------------------------------------------------------------------------- No issues found here. Checking nits according to https://www.ietf.org/id-info/checklist : ---------------------------------------------------------------------------- No issues found here. Miscellaneous warnings: ---------------------------------------------------------------------------- == The copyright year in the IETF Trust and authors Copyright Line does not match the current year -- The document date (February 28, 2017) is 2613 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) ** Obsolete normative reference: RFC 793 (Obsoleted by RFC 9293) ** Obsolete normative reference: RFC 2460 (Obsoleted by RFC 8200) Summary: 2 errors (**), 0 flaws (~~), 1 warning (==), 1 comment (--). Run idnits with the --verbose option for more detailed information about the items above. -------------------------------------------------------------------------------- 2 INTERNET-DRAFT N. Elkins 3 Inside Products 4 R. Hamilton 5 Chemical Abstracts Service 6 M. Ackermann 7 Intended Status: Proposed Standard BCBS Michigan 8 Expires: September 1, 2017 February 28, 2017 10 IPv6 Performance and Diagnostic Metrics (PDM) Destination Option 11 draft-ietf-ippm-6man-pdm-option-08 13 Abstract 15 To assess performance problems, measurements based on optional 16 sequence numbers and timing may be embedded in each packet. Such 17 measurements may be interpreted in real-time or after the fact. An 18 implementation of the existing IPv6 Destination Options extension 19 header, the Performance and Diagnostic Metrics (PDM) Destination 20 Options extension header as well as the field limits, calculations, 21 and usage of the PDM in measurement are included in this document. 23 Status of this Memo 25 This Internet-Draft is submitted to IETF in full conformance with the 26 provisions of BCP 78 and BCP 79. 28 Internet-Drafts are working documents of the Internet Engineering 29 Task Force (IETF), its areas, and its working groups. Note that 30 other groups may also distribute working documents as 31 Internet-Drafts. 33 Internet-Drafts are draft documents valid for a maximum of six months 34 and may be updated, replaced, or obsoleted by other documents at any 35 time. It is inappropriate to use Internet-Drafts as reference 36 material or to cite them other than as "work in progress." 38 The list of current Internet-Drafts can be accessed at 39 http://www.ietf.org/1id-abstracts.html 41 The list of Internet-Draft Shadow Directories can be accessed at 42 http://www.ietf.org/shadow.html 44 Copyright and License Notice 46 Copyright (c) 2017 IETF Trust and the persons identified as the 47 document authors. All rights reserved. 49 IETF Trust Legal Provisions of 28-dec-2009, Section 6.b(i), paragraph 50 3: This document is subject to BCP 78 and the IETF Trust's Legal 51 Provisions Relating to IETF Documents 52 (http://trustee.ietf.org/license-info) in effect on the date of 53 publication of this document. Please review these documents 54 carefully, as they describe your rights and restrictions with respect 55 to this document. Code Components extracted from this document must 56 include Simplified BSD License text as described in Section 4.e of 57 the Trust Legal Provisions and are provided without warranty as 58 described in the Simplified BSD License. 60 Table of Contents 62 1 Background . . . . . . . . . . . . . . . . . . . . . . . . . . 4 63 1.1 Terminology . . . . . . . . . . . . . . . . . . . . . . . . 4 64 1.2 End User Quality of Service (QoS) . . . . . . . . . . . . . 4 65 1.3 Need for a Packet Sequence Number (PSN) . . . . . . . . . . 5 66 1.4 Rationale for defined solution . . . . . . . . . . . . . . . 5 67 1.5 PDM Works in Collaboration with Other Headers . . . . . . . 6 68 1.6 IPv6 Transition Technologies . . . . . . . . . . . . . . . . 6 69 2 Measurement Information Derived from PDM . . . . . . . . . . . . 6 70 2.1 Round-Trip Delay . . . . . . . . . . . . . . . . . . . . . . 7 71 2.2 Server Delay . . . . . . . . . . . . . . . . . . . . . . . . 7 72 3 Performance and Diagnostic Metrics Destination Option Layout . . 7 73 3.1 Destination Options Header . . . . . . . . . . . . . . . . . 7 74 3.2 Performance and Diagnostic Metrics Destination Option . . . 7 75 3.2.1 PDM Layout . . . . . . . . . . . . . . . . . . . . . . . 7 76 3.2.2 Base Unit for Time Measurement . . . . . . . . . . . . . 9 77 3.2.3 Considerations of this time-differential 78 representation . . . . . . . . . . . . . . . . . . . . . 10 79 3.2.3.1 Limitations with this encoding method . . . . . . . 10 80 3.2.3.2 Loss of precision induced by timer value 81 truncation . . . . . . . . . . . . . . . . . . . . . 11 82 3.3 Header Placement . . . . . . . . . . . . . . . . . . . . . . 12 83 3.4 Header Placement Using IPSec ESP Mode . . . . . . . . . . . 12 84 3.4.1 Using ESP Transport Mode . . . . . . . . . . . . . . . . 12 85 3.4.2 Using ESP Tunnel Mode . . . . . . . . . . . . . . . . . 13 86 3.5 Implementation Considerations . . . . . . . . . . . . . . . 14 87 3.6 Dynamic Configuration Options . . . . . . . . . . . . . . . 14 88 3.6 5-tuple Aging . . . . . . . . . . . . . . . . . . . . . . . 14 89 4 Security Considerations . . . . . . . . . . . . . . . . . . . . 14 90 4.1. SYN Flood and Resource Consumption Attacks . . . . . . . . 15 91 4.2 Pervasive monitoring . . . . . . . . . . . . . . . . . . . 15 92 4.3 PDM as a Covert Channel . . . . . . . . . . . . . . . . . . 16 93 4.4 Timing Attacks . . . . . . . . . . . . . . . . . . . . . . . 16 94 5 IANA Considerations . . . . . . . . . . . . . . . . . . . . . . 17 95 6 References . . . . . . . . . . . . . . . . . . . . . . . . . . . 17 96 6.1 Normative References . . . . . . . . . . . . . . . . . . . . 17 97 6.2 Informative References . . . . . . . . . . . . . . . . . . . 18 98 Appendix A : Timing Time Differential Calculations . . . . . . . . 18 99 Appendix B: Sample Packet Flows . . . . . . . . . . . . . . . . . 19 100 B.1 PDM Flow - Simple Client Server . . . . . . . . . . . . . . 19 101 B.1.1 Step 1 . . . . . . . . . . . . . . . . . . . . . . . . . 20 102 B.1.2 Step 2 . . . . . . . . . . . . . . . . . . . . . . . . . 20 103 B.1.3 Step 3 . . . . . . . . . . . . . . . . . . . . . . . . . 21 104 B.1.4 Step 4 . . . . . . . . . . . . . . . . . . . . . . . . . 22 105 B.1.5 Step 5 . . . . . . . . . . . . . . . . . . . . . . . . . 23 106 B.2 Other Flows . . . . . . . . . . . . . . . . . . . . . . . . 23 107 B.2.1 PDM Flow - One Way Traffic . . . . . . . . . . . . . . . 23 108 B.2.2 PDM Flow - Multiple Send Traffic . . . . . . . . . . . . 24 109 B.2.3 PDM Flow - Multiple Send with Errors . . . . . . . . . . 25 110 Appendix C: Potential Overhead Considerations . . . . . . . . . . 27 111 Acknowledgments . . . . . . . . . . . . . . . . . . . . . . . . . 28 112 Authors' Addresses . . . . . . . . . . . . . . . . . . . . . . . . 28 114 1 Background 116 To assess performance problems, measurements based on optional 117 sequence numbers and timing may be embedded in each packet. Such 118 measurements may be interpreted in real-time or after the fact. 120 As defined in RFC2460 [RFC2460], destination options are carried by 121 the IPv6 Destination Options extension header. Destination options 122 include optional information that need be examined only by the IPv6 123 node given as the destination address in the IPv6 header, not by 124 routers or other "middle boxes". This document specifies a new 125 destination option, the Performance and Diagnostic Metrics (PDM) 126 destination option. This document specifies the layout, field 127 limits, calculations, and usage of the PDM in measurement. 129 1.1 Terminology 131 The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", 132 "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this 133 document are to be interpreted as described in RFC 2119 [RFC2119]. 135 1.2 End User Quality of Service (QoS) 137 The timing values in the PDM embedded in the packet will be used to 138 estimate QoS as experienced by an end user device. 140 For many applications, the key user performance indicator is response 141 time. When the end user is an individual, he is generally 142 indifferent to what is happening along the network; what he really 143 cares about is how long it takes to get a response back. But this is 144 not just a matter of individuals' personal convenience. In many 145 cases, rapid response is critical to the business being conducted. 147 When the end user is a device (e.g. with the Internet of Things), 148 what matters is the speed with which requested data can be 149 transferred -- specifically, whether the requested data can be 150 transferred in time to accomplish the desired actions. This can be 151 important when the relevant external conditions are subject to rapid 152 change. 154 Low, reliable and acceptable response times are not just "nice to 155 have". On many networks, the impact can be financial hardship or can 156 endanger human life. In some cities, the emergency police contact 157 system operates over IP; law enforcement, at all levels, use IP 158 networks; transactions on our stock exchanges are settled using IP 159 networks. The critical nature of such activities to our daily lives 160 and financial well-being demand a simple solution to support response 161 time measurements. 163 1.3 Need for a Packet Sequence Number (PSN) 165 While performing network diagnostics of an end-to-end connection, it 166 often becomes necessary to isolate the factors along the network path 167 responsible for problems. Diagnostic data may be collected at 168 multiple places along the path (if possible), or at the source and 169 destination. Then, in post-collection processing, the diagnostic 170 data corresponding to each packet at different observation points 171 must be matched for proper measurements. A sequence number in each 172 packet provides sufficient basis for the matching process. If need 173 be, the timing fields may be used along with the sequence number to 174 ensure uniqueness. 176 This method of data collection along the path is of special use to 177 determine where packet loss or packet corruption is happening. 179 The packet sequence number needs to be unique in the context of the 180 session (5-tuple). See section 2 for a definition of 5-tuple. 182 1.4 Rationale for defined solution 184 The current IPv6 specification does not provide timing nor a similar 185 field in the IPv6 main header or in any extension header. So, we 186 define the IPv6 Performance and Diagnostic Metrics destination option 187 (PDM). 189 Advantages include: 191 1. Real measure of actual transactions 192 2. Independence from transport layer protocols 193 3. Ability to span organizational boundaries with consistent 194 instrumentation 195 4. No time synchronization needed between session partners 196 5. Ability to handle all transport protocols (TCP, UDP, SCTP, etc) 197 in a uniform way 199 The PDM provides the ability to determine quickly if the (latency) 200 problem is in the network or in the server (application). More 201 intermediate measurements may be needed if the host or network 202 discrimination is not sufficient. At the client, TCP/IP stack time 203 vs. application time may still need to be broken out by client 204 software. 206 1.5 PDM Works in Collaboration with Other Headers 208 The purpose of the PDM is not to supplant all the variables present 209 in all other headers but to provide data which is not available or 210 very difficult to get. The way PDM would be used is by a technician 211 (or tool) looking at a packet capture. Within the packet capture, 212 they would have available to them the layer 2 header, IP header (v6 213 or v4), TCP, UCP, ICMP, SCTP or other headers. All information 214 would be looked at together to make sense of the packet flow. The 215 technician or processing tool could analyze, report or ignore the 216 data from PDM, as necessary. 218 For an example of how PDM can help with TCP retransmit problems, 219 please look at section 8. 221 1.6 IPv6 Transition Technologies 223 In the path to full implementation of IPv6, transition technologies 224 such as translation or tunneling may be employed. The PDM header is 225 not expected to work in such scenarios. It is likely that an IPv6 226 packet containing PDM will be dropped if using IPv6 transition 227 technologies. 229 2 Measurement Information Derived from PDM 231 Each packet contains information about the sender and receiver. In IP 232 protocol, the identifying information is called a "5-tuple". 234 The 5-tuple consists of: 236 SADDR : IP address of the sender 237 SPORT : Port for sender 238 DADDR : IP address of the destination 239 DPORT : Port for destination 240 PROTC : Protocol for upper layer (ex. TCP, UDP, ICMP, etc.) 242 The PDM contains the following base fields: 244 PSNTP : Packet Sequence Number This Packet 245 PSNLR : Packet Sequence Number Last Received 246 DELTATLR : Delta Time Last Received 247 DELTATLS : Delta Time Last Sent 249 Other fields for storing time scaling factors are also in the PDM and 250 will be described in section 3. 252 This information, combined with the 5-tuple, allows the measurement 253 of the following metrics: 255 1. Round-trip delay 256 2. Server delay 258 2.1 Round-Trip Delay 260 Round-trip *Network* delay is the delay for packet transfer from a 261 source host to a destination host and then back to the source host. 262 This measurement has been defined, and the advantages and 263 disadvantages discussed in "A Round-trip Delay Metric for IPPM" 264 [RFC2681]. 266 2.2 Server Delay 268 Server delay is the interval between when a packet is received by a 269 device and the first corresponding packet is sent back in response. 270 This may be "Server Processing Time". It may also be a delay caused 271 by acknowledgements. Server processing time includes the time taken 272 by the combination of the stack and application to return the 273 response. The stack delay may be related to network performance. If 274 this aggregate time is seen as a problem, and there is a need to make 275 a clear distinction between application processing time and stack 276 delay, including that caused by the network, then more client based 277 measurements are needed. 279 3 Performance and Diagnostic Metrics Destination Option Layout 281 3.1 Destination Options Header 283 The IPv6 Destination Options Header is used to carry optional 284 information that needs to be examined only by a packet's destination 285 node(s). The Destination Options Header is identified by a Next 286 Header value of 60 in the immediately preceding header and is defined 287 in RFC2460 [RFC2460]. The IPv6 Performance and Diagnostic Metrics 288 Destination Option (PDM) is an implementation of the Destination 289 Options Header. The PDM does not require time synchronization. 291 3.2 Performance and Diagnostic Metrics Destination Option 293 3.2.1 PDM Layout 295 The IPv6 Performance and Diagnostic Metrics Destination Option (PDM) 296 contains the following fields: 298 SCALEDTLR: Scale for Delta Time Last Received 299 SCALEDTLS: Scale for Delta Time Last Sent 300 PSNTP : Packet Sequence Number This Packet 301 PSNLR : Packet Sequence Number Last Received 302 DELTATLR : Delta Time Last Received 303 DELTATLS : Delta Time Last Sent 305 The PDM destination option is encoded in type-length-value (TLV) 306 format as follows: 308 0 1 2 3 309 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 2 3 4 5 6 7 8 9 0 1 310 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 311 | Option Type | Option Length | ScaleDTLR | ScaleDTLS | 312 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 313 | PSN This Packet | PSN Last Received | 314 |-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 315 | Delta Time Last Received | Delta Time Last Sent | 316 +-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+-+ 318 Option Type 320 TBD = 0xXX (TBD) [To be assigned by IANA] [RFC2780] 322 Option Length 324 8-bit unsigned integer. Length of the option, in octets, excluding 325 the Option Type and Option Length fields. This field MUST be set to 326 16. 328 Scale Delta Time Last Received (SCALEDTLR) 330 8-bit unsigned integer. This is the scaling value for the Delta Time 331 Last Received (DELTATLR) field. The possible values are from 0-255. 332 See Section 4 for further discussion on Timing Considerations and 333 formatting of the scaling values. 335 Scale Delta Time Last Sent (SCALEDTLS) 337 8-bit signed integer. This is the scaling value for the Delta Time 338 Last Sent (DELTATLS) field. The possible values are from 0 to 255. 340 Packet Sequence Number This Packet (PSNTP) 342 16-bit unsigned integer. This field will wrap. It is intended for 343 use while analyzing packet traces. 345 Initialized at a random number and incremented monotonically for each 346 packet of the session flow of the 5-tuple. The random number 347 initialization is intended to make it harder to spoof and insert such 348 packets. 350 Operating systems MUST implement a separate packet sequence number 351 counter per 5-tuple. 353 Packet Sequence Number Last Received (PSNLR) 355 16-bit unsigned integer. This is the PSNTP of the packet last 356 received on the 5-tuple. 358 Delta Time Last Received (DELTATLR) 360 A 16-bit unsigned integer field. The value is set according to the 361 scale in SCALEDTLR. 363 Delta Time Last Received = (Send time packet 2 - Receive time packet 364 1) 366 Delta Time Last Sent (DELTATLS) 368 A 16-bit unsigned integer field. The value is set according to the 369 scale in SCALEDTLS. 371 Delta Time Last Sent = (Receive time packet 2 - Send time packet 1) 373 Option Type 375 In keeping with RFC2460[RFC2460], the two high order bits of the 376 Option Type field are encoded to indicate specific processing of the 377 option; for the PDM destination option, these two bits MUST be set to 378 00. 380 The third high order bit of the Option Type specifies whether or not 381 the Option Data of that option can change en-route to the packet's 382 final destination. 384 In the PDM, the value of the third high order bit MUST be 0. 385 3.2.2 Base Unit for Time Measurement 387 A time differential is always a whole number in a CPU; it is the unit 388 specification -- hours, seconds, nanoseconds -- that determine what 389 the numeric value means. For PDM, we establish the base time unit as 390 1 attosecond (asec). This allows for a common unit and scaling of the 391 time differential among all IP stacks and hardware implementations. 393 Note that we are trying to provide the ability to measure both time 394 differentials that are extremely small, and time differentials in a 395 DTN-type environment where the delays may be very great. To store a 396 time differential in just 16 bits that must range in this way will 397 require some scaling of the time differential value. 399 One issue is the conversion from the native time base in the CPU 400 hardware of whatever device is in use to some number of attoseconds. 401 It might seem this will be an astronomical number, but the conversion 402 is straightforward. It involves multiplication by an appropriate 403 power of 10 to change the value into a number of attoseconds. Then, 404 to scale the value so that it fits into DELTATLR or DELTATLS, the 405 value is shifted by of a number of bits, retaining the 16 high-order 406 or most significant bits. The number of bits shifted becomes the 407 scaling factor, stored as SCALEDTLR or SCALEDTLS, respectively. For a 408 full description of this process, including examples, please see 409 Appendix A. 411 3.2.3 Considerations of this time-differential representation 413 There are a few considerations to be taken into account with this 414 representation of a time differential. The first is whether there are 415 any limitations on the maximum or minimum time differential that can 416 be expressed using method of a delta value and a scaling factor. The 417 second is the amount of imprecision introduced by this method. 419 3.2.3.1 Limitations with this encoding method 421 The DELTATLS and DELTATLR fields store only the 16 most-significant 422 bits of the time differential value. Thus the range, excluding the 423 scaling factor, is from 0 to 65535, or a maximum of 2**16-1. This 424 method is further described in [TRAM-TCPM]. 426 The actual magnitude of the time differential is determined by the 427 scaling factor. SCALEDTLR and SCALEDTLS are 8-bit unsigned integers, 428 so the scaling factor ranges from 0 to 255. The smallest number that 429 can be represented would have a value of 1 in the delta field and a 430 value of 0 in the associated scale field. This is the representation 431 for 1 attosecond. Clearly this allows PDM to measure extremely small 432 time differentials. 434 On the other end of the scale, the maximum delta value is 65535, or 435 FFFF in hexadecimal. If the maximum scale value of 255 is used, the 436 time differential represented is 65535*2**255, which is over 3*10**55 437 years, essentially, forever. So there appears to be no real 438 limitation to the time differential that can be represented. 440 3.2.3.2 Loss of precision induced by timer value truncation 442 As PDM specifies the DELTATLR and DELTATLS values as 16-bit unsigned 443 integers, any time the precision is greater than those 16 bits, there 444 will be truncation of the trailing bits, with an accompanying loss of 445 precision in the value. 447 Any time differential value smaller than 65536 asec can be stored 448 exactly in DELTATLR or DELTATLS, because the representation of this 449 value requires at most 16 bits. 451 Since the time differential values in PDM are measured in 452 attoseconds, the range of values that would be truncated to the same 453 encoded value is 2**(Scale)-1 asec. 455 For example, the smallest time differential that would be truncated 456 to fit into a delta field is 458 1 0000 0000 0000 0000 = 65536 asec 460 This value would be encoded as a delta value of 8000 (hexadecimal) 461 with a scaling factor of 1. The value 463 1 0000 0000 0000 0001 = 65537 asec 465 would also be encoded as a delta value of 8000 with a scaling factor 466 of 1. This actually is the largest value that would be truncated to 467 that same encoded value. When the scale value is 1, the value range 468 is calculated as 2**1 - 1, or 1 asec, which you can see is the 469 difference between these minimum and maximum values. 471 The scaling factor is defined as the number of low-order bits 472 truncated to reduce the size of the resulting value so it fits into a 473 16-bit delta field. If, for example, you had to truncate 12 bits, the 474 loss of precision would depend on the bits you truncated. The range 475 of these values would be 477 0000 0000 0000 = 0 asec 478 to 479 1111 1111 1111 = 4095 asec 481 So the minimum loss of precision would be 0 asec, where the delta 482 value exactly represents the time differential, and the maximum loss 483 of precision would be 4095 asec. As stated above, the scaling factor 484 of 12 means the maximum loss of precision is 2**12-1 asec, or 4095 485 asec. 487 Compare this loss of precision to the actual time differential. The 488 range of actual time differential values that would incur this loss 489 of precision is from 491 1000 0000 0000 0000 0000 0000 0000 = 2**27 asec or 134217728 asec 492 to 493 1111 1111 1111 1111 1111 1111 1111 = 2**28-1 asec or 268435455 asec 495 Granted, these are small values, but the point is, any value between 496 these two values will have a maximum loss of precision of 4095 asec, 497 or about 0.00305% for the first value, as encoded, and at most 498 0.001526% for the second. These maximum-loss percentages are 499 consistent for all scaling values. 501 3.3 Header Placement 503 The PDM Destination Option is placed as defined in RFC2460 [RFC2460]. 504 There may be a choice of where to place the Destination Options 505 header. If using ESP mode, please see section 3.4 of this document 506 for placement of the PDM Destination Options header. 508 For each IPv6 packet header, the PDM MUST NOT appear more than once. 509 However, an encapsulated packet MAY contain a separate PDM associated 510 with each encapsulated IPv6 header. 512 3.4 Header Placement Using IPSec ESP Mode 514 IPSec Encapsulating Security Payload (ESP) is defined in [RFC4303] 515 and is widely used. Section 3.1.1 of [RFC4303] discusses placement 516 of Destination Options Headers. 518 The placement of PDM is different depending on if ESP is used in 519 tunnel or transport mode. 521 3.4.1 Using ESP Transport Mode 523 Below is the diagram from [RFC4303] discussing placement of headers. 524 Note that Destination Options MAY be placed before or after ESP or 525 both. If using PDM in ESP transport mode, PDM MUST be placed after 526 the ESP header so as not to leak information. 528 BEFORE APPLYING ESP 529 --------------------------------------- 530 IPv6 | | ext hdrs | | | 531 | orig IP hdr |if present| TCP | Data | 532 --------------------------------------- 533 AFTER APPLYING ESP 534 --------------------------------------------------------- 535 IPv6 | orig |hop-by-hop,dest*,| |dest| | | ESP | ESP| 536 |IP hdr|routing,fragment.|ESP|opt*|TCP|Data|Trailer| ICV| 537 --------------------------------------------------------- 538 |<--- encryption ---->| 539 |<------ integrity ------>| 541 * = if present, could be before ESP, after ESP, or both 543 3.4.2 Using ESP Tunnel Mode 545 Below is the diagram from [RFC4303] discussing placement of headers. 547 Note that Destination Options MAY be placed before or after ESP or 548 both in both the outer set of IP headers and the inner set of IP 549 headers. 551 In ESP tunnel mode, PDM MAY be placed before or after the ESP header 552 or both. 554 BEFORE APPLYING ESP 556 --------------------------------------- 557 IPv6 | | ext hdrs | | | 558 | orig IP hdr |if present| TCP | Data | 559 --------------------------------------- 561 AFTER APPLYING ESP 563 ------------------------------------------------------------ 564 IPv6 | new* |new ext | | orig*|orig ext | | | ESP | ESP| 565 |IP hdr| hdrs* |ESP|IP hdr| hdrs * |TCP|Data|Trailer| ICV| 566 ------------------------------------------------------------ 567 |<--------- encryption ---------->| 568 |<------------ integrity ------------>| 570 * = if present, construction of outer IP hdr/extensions and 571 modification of inner IP hdr/extensions is discussed in 572 the Security Architecture document. 574 As a completely new IP packet will be made, it means that PDM 575 information for that packet does not contain any information from the 576 inner packet, i.e. the PDM information will NOT be based on the 577 transport layer (TCP, UDP, etc) ports etc in the inner header, but 578 will be specific to the ESP flow. 580 If PDM information for the inner packet is desired, the original host 581 sending the inner packet needs to put PDM header in the tunneled 582 packet, and then the PDM information will be specific for that 583 stream. 585 3.5 Implementation Considerations 587 The PDM destination options extension header MUST be explicitly 588 turned on by each stack on a host node by administrative action. The 589 default value of PDM is off. 591 PDM MUST NOT be turned on merely if a packet is received with a PDM 592 header. The received packet could be spoofed by another device. 594 3.6 Dynamic Configuration Options 596 If implemented, each operating system MUST have a default 597 configuration parameter, e.g. diag_header_sys_default_value=yes/no. 598 The operating system MAY also have a dynamic configuration option to 599 change the configuration setting as needed. 601 If the PDM destination options extension header is used, then it MAY 602 be turned on for all packets flowing through the host, applied to an 603 upper-layer protocol (TCP, UDP, SCTP, etc), a local port, or IP 604 address only. These are at the discretion of the implementation. 606 3.6 5-tuple Aging 608 Within the operating system, metrics must be kept on a 5-tuple basis. 610 The question comes of when to stop keeping data or restarting the 611 numbering for a 5-tuple. For example, in the case of TCP, at some 612 point, the connection will terminate. Keeping data in control blocks 613 forever, will have unfortunate consequences for the operating system. 615 So, the recommendation is to use a known aging parameter such as Max 616 Segment Lifetime (MSL) as defined in Transmission Control Protocol 617 [RFC0793] to reuse or drop the control block. The choice of aging 618 parameter is left up to the implementation. 620 4 Security Considerations 622 PDM may introduce some new security weaknesses. 624 4.1. SYN Flood and Resource Consumption Attacks 626 PDM needs to calculate the deltas for time and keep track of the 627 sequence numbers. This means that control blocks must be kept at the 628 end hosts per 5-tuple. Any time a control block is kept, an 629 attacker can try to mis-use the control blocks such that there is a 630 compromise of the end host. 632 PDM is used only at the end hosts and the control blocks are only 633 kept at the end host and not at routers or middle boxes. Remember, 634 PDM is an implementation of the Destination Option extension header. 636 A "SYN flood" type of attack succeeds because a TCP SYN packet is 637 small but it causes the end host to start creating a place holder for 638 the session such that quite a bit of control block and other storage 639 is used. This is an asynchronous type of attack in that a small 640 amount of work by the attacker creates a large amount of work by the 641 resource attacked. 643 For PDM, the amount of data to be kept is quite small. That is, the 644 control block is quite lightweight. Concerns about SYN Flood and 645 other type of resource consumption attacks (memory, processing power, 646 etc) can be alleviated by having a limit on the number of control 647 block entries. 649 We recommend that implementation of PDM SHOULD have a limit on the 650 number of control block entries. 652 4.2 Pervasive monitoring 654 Since PDM passes in the clear, a concern arises as to whether the 655 data can be used to fingerprint the system or somehow obtain 656 information about the contents of the payload. 658 Let us discuss fingerprinting of the end host first. It is possible 659 that seeing the pattern of deltas or the absolute values could give 660 some information as to the speed of the end host - that is, if it is 661 a very fast system or an older, slow device. This may be useful to 662 the attacker. However, if the attacker has access to PDM, the 663 attacker also has access to the entire packet and could make such a 664 deduction based merely on the time frames elapsed between packets 665 WITHOUT PDM. 667 As far as deducing the content of the payload, it appears to us that 668 PDM is quite unhelpful in this regard. 670 4.3 PDM as a Covert Channel 672 PDM provides a set of fields in the packet which could be used to 673 leak data. But, there is no real reason to suspect that PDM would 674 be chosen rather than another part of the payload or another 675 Extension Header. 677 A firewall or another device could sanity check the fields within the 678 PDM but randomly assigned sequence numbers and delta times might be 679 expected to vary widely. The biggest problem though is how an 680 attacker would get access to PDM in the first place to leak data. 681 The attacker would have to either compromise the end host or have Man 682 in the Middle (MitM). It is possible that either one could change 683 the fields. But, then the other end host would get sequence numbers 684 and deltas that don't make any sense. Presumably, one is using PDM 685 and doing packet tracing for diagnostic purposes, so the changes 686 would be obvious. It is conceivable that someone could compromise 687 an end host and make it start sending packets with PDM without the 688 knowledge of the host. But, again, the bigger problem is the 689 compromise of the end host. Once that is done, the attacker 690 probably has better ways to leak data. 692 Having said that, an implementation SHOULD stop using PDM if it gets 693 some number of "nonsensical" sequence numbers. 695 4.4 Timing Attacks 697 The fact that PDM can help in the separation of node processing time 698 from network latency brings value to performance monitoring. Yet, it 699 is this very characteristic of PDM which may be misused to make 700 certain new type of timing attacks against protocols and 701 implementations possible. 703 Depending on the nature of the cryptographic protocol used, it may be 704 possible to leak the long term credentials of the device. For 705 example, if an attacker is able to create an attack which causes the 706 enterprise to turn on PDM to diagnose the attack, then the attacker 707 might use PDM during that debugging time to launch a timing attack 708 against the long term keying material used by the cryptographic 709 protocol. 711 An implementation may want to be sure that PDM is enabled only for 712 certain ip addresses, or only for some ports. Additionally, we 713 recommend that the implementation SHOULD require an explicit restart 714 of monitoring after a certain timeperiod (for example for 1 hour), to 715 make sure that PDM is not accidently left on after debugging has been 716 done etc. 718 Even so, if using PDM, we introduce the concept of user "Consent to 719 be Measured" as a pre-requisite for using PDM. Consent is common in 720 enterprises and with some subscription services. So, if with PDM, we 721 recommend that the user SHOULD consent to its use. 723 5 IANA Considerations 725 This draft requests an Option Type assignment in the Destination 726 Options and Hop-by-Hop Options sub-registry of Internet Protocol 727 Version 6 (IPv6) Parameters [ref to RFCs and URL below]. 729 http://www.iana.org/assignments/ipv6-parameters/ipv6- 730 parameters.xhtml#ipv6-parameters-2 732 Hex Value Binary Value Description Reference 733 act chg rest 734 ------------------------------------------------------------------- 735 TBD TBD Performance and [This draft] 736 Diagnostic Metrics 737 (PDM) 739 6 References 741 6.1 Normative References 743 [RFC0793] Postel, J., "Transmission Control Protocol", STD 7, RFC 744 793, September 1981. 746 [RFC1122] Braden, R., "Requirements for Internet Hosts -- 747 Communication Layers", RFC 1122, October 1989. 749 [RFC2119] Bradner, S., "Key words for use in RFCs to Indicate 750 Requirement Levels", BCP 14, RFC 2119, March 1997. 752 [RFC2460] Deering, S. and R. Hinden, "Internet Protocol, Version 6 753 (IPv6) Specification", RFC 2460, December 1998. 755 [RFC2681] Almes, G., Kalidindi, S., and M. Zekauskas, "A Round-trip 756 Delay Metric for IPPM", RFC 2681, September 1999. 758 [RFC2780] Bradner, S. and V. Paxson, "IANA Allocation Guidelines For 759 Values In the Internet Protocol and Related Headers", BCP 37, RFC 760 2780, March 2000. 762 [RFC4303] Kent, S, "IP Encapsulating Security Payload (ESP)", RFC 763 4303, December 2005. 765 6.2 Informative References 767 [TRAM-TCPM] Trammel, B., "Encoding of Time Intervals for the TCP 768 Timestamp Option-01", Internet Draft, July 2013. [Work in Progress] 770 Appendix A : Timing Time Differential Calculations 772 The time counter in a CPU is a binary whole number, representing a 773 number of milliseconds (msec), microseconds (usec) or even 774 picoseconds (psec). Representing one of these values as attoseconds 775 (asec) means multiplying by 10 raised to some exponent. Refer to this 776 table of equalities: 778 Base value = # of sec = # of asec 1000s of asec 779 --------------- ------------- ------------- ------------- 780 1 second 1 sec 10**18 asec 1000**6 asec 781 1 millisecond 10**-3 sec 10**15 asec 1000**5 asec 782 1 microsecond 10**-6 sec 10**12 asec 1000**4 asec 783 1 nanosecond 10**-9 sec 10**9 asec 1000**3 asec 784 1 picosecond 10**-12 sec 10**6 asec 1000**2 asec 785 1 femtosecond 10**-15 sec 10**3 asec 1000**1 asec 787 For example, if you have a time differential expressed in 788 microseconds, since each microsecond is 10**12 asec, you would 789 multiply your time value by 10**12 to obtain the number of 790 attoseconds. If you time differential is expressed in nanoseconds, 791 you would multiply by 10**9 to get the number of attoseconds. 793 The result is a binary value that will need to be shortened by a 794 number of bits so it will fit into the 16-bit PDM DELTA field. 796 The next step is to divide by 2 until the value is contained in just 797 16 significant bits. The exponent of the value in the last column of 798 of the table is useful here; the initial scaling factor is that 799 exponent multiplied by 10. This is the minimum number of low-order 800 bits to be shifted-out or discarded. It represents dividing the time 801 value by 1024 raised to that exponent. 803 The resulting value may still be too large to fit into 16 bits, but 804 can be normalized by shifting out more bits (dividing by 2) until the 805 value fits into the 16-bit DELTA field. The number of extra bits 806 shifted out is then added to the scaling factor. The scaling factor, 807 the total number of low-order bits dropped, is the SCALEDTL value. 809 For example: say an application has these start and finish timer 810 values (hexadecimal values, in microseconds): 812 Finish: 27C849234 usec (02:57:58.997556) 813 -Start: 27C83F696 usec (02:57:58.957718) 814 ========== ========= =============== 815 Difference 9B9E usec 00.039838 sec or 39838 usec 817 To convert this differential value to binary attoseconds, multiply 818 the number of microseconds by 10**12. Divide by 1024**4, or simply 819 discard 40 bits from the right. The result is 36232, or 8D88 in hex, 820 with a scaling factor or SCALEDTL value of 40. 822 For another example, presume the time differential is larger, say 823 32.311072 seconds, which is 32311072 usec. Each microsecond is 10**12 824 asec, so multiply by 10**12, giving the hexadecimal value 825 1C067FCCAE8120000. Using the initial scaling factor of 40, drop the 826 last 10 characters (40 bits) from that string, giving 1C067FC. This 827 will not fit into a DELTA field, as it is 25 bits long. Shifting the 828 value to the right another 9 bits results in a DELTA value of E033, 829 with a resulting scaling factor of 49. 831 When the time differential value is a small number, regardless of the 832 time unit, the exponent trick given above is not useful in 833 determining the proper scaling value. For example, if the time 834 differential is 3 seconds and you want to convert that directly, you 835 would follow this path: 837 3 seconds = 3*10**18 asec (decimal) 838 = 29A2241AF62C0000 asec (hexadecimal) 840 If you just truncate the last 60 bits, you end up with a delta value 841 of 2 and a scaling factor of 60, when what you really wanted was a 842 delta value with more significant digits. The most precision with 843 which you can store this value in 16 bits is A688, with a scaling 844 factor of 46. 846 Appendix B: Sample Packet Flows 848 B.1 PDM Flow - Simple Client Server 850 Following is a sample simple flow for the PDM with one packet sent 851 from Host A and one packet received by Host B. The PDM does not 852 require time synchronization between Host A and Host B. The 853 calculations to derive meaningful metrics for network diagnostics are 854 shown below each packet sent or received. 856 B.1.1 Step 1 858 Packet 1 is sent from Host A to Host B. The time for Host A is set 859 initially to 10:00AM. 861 The time and packet sequence number are saved by the sender 862 internally. The packet sequence number and delta times are sent in 863 the packet. 865 Packet 1 867 +----------+ +----------+ 868 | | | | 869 | Host | ----------> | Host | 870 | A | | B | 871 | | | | 872 +----------+ +----------+ 874 PDM Contents: 876 PSNTP : Packet Sequence Number This Packet: 25 877 PSNLR : Packet Sequence Number Last Received: - 878 DELTATLR : Delta Time Last Received: - 879 SCALEDTLR: Scale of Delta Time Last Received: 0 880 DELTATLS : Delta Time Last Sent: - 881 SCALEDTLS: Scale of Delta Time Last Sent: 0 883 Internally, within the sender, Host A, it must keep: 885 Packet Sequence Number of the last packet sent: 25 886 Time the last packet was sent: 10:00:00 888 Note, the initial PSNTP from Host A starts at a random number. In 889 this case, 25. The time in these examples is shown in seconds for 890 the sake of simplicity. 892 B.1.2 Step 2 894 Packet 1 is received at Host B. Its time is set to one hour later 895 than Host A. In this case, 11:00AM 897 Internally, within the receiver, Host B, it must note: 899 Packet Sequence Number of the last packet received: 25 900 Time the last packet was received : 11:00:03 902 Note, this timestamp is in Host B time. It has nothing whatsoever to 903 do with Host A time. The Packet Sequence Number of the last packet 904 received will become PSNLR which will be sent out in the packet sent 905 by Host B in the next step. The time last received will be used to 906 calculate the DELTALR value to be sent out in the packet sent by Host 907 B in the next step. 909 B.1.3 Step 3 911 Packet 2 is sent by Host B to Host A. Note, the initial packet 912 sequence number (PSNTP) from Host B starts at a random number. In 913 this case, 12. Before sending the packet, Host B does a calculation 914 of deltas. Since Host B knows when it is sending the packet, and it 915 knows when it received the previous packet, it can do the following 916 calculation: 918 Sending time : packet 2 - receive time : packet 1 920 We will call the result of this calculation: Delta Time Last Received 921 (DELTATLR) 923 Note, both sending time and receive time are saved internally in Host 924 B. They do not travel in the packet. Only the Delta is in the 925 packet. 927 Assume that within Host B is the following: 929 Packet Sequence Number of the last packet received: 25 930 Time the last packet was received: 11:00:03 931 Packet Sequence Number of this packet: 12 932 Time this packet is being sent: 11:00:07 934 We can now calculate a delta value to be sent out in the packet. 935 DELTATLR becomes: 937 4 seconds = 11:00:07 - 11:00:03 = 3782DACE9D900000 asec 939 This is the derived metric: Server Delay. The time and scaling 940 factor must be converted; in this case, the time differential is 941 DE0B, and the scaling factor is 2E, or 46 in decimal. Then, these 942 values, along with the packet sequence numbers will be sent to Host A 943 as follows: 945 Packet 2 947 +----------+ +----------+ 948 | | | | 949 | Host | <---------- | Host | 950 | A | | B | 951 | | | | 952 +----------+ +----------+ 954 PDM Contents: 956 PSNTP : Packet Sequence Number This Packet: 12 957 PSNLR : Packet Sequence Number Last Received: 25 958 DELTATLR : Delta Time Last Received: DE0B (4 seconds) 959 SCALEDTLR: Scale of Delta Time Last Received: 2E (46 decimal) 960 DELTATLS : Delta Time Last Sent: - 961 SCALEDTLS: Scale of Delta Time Last Sent: 0 963 The metric left to be calculated is the Round-Trip Delay. This will 964 be calculated by Host A when it receives Packet 2. 966 B.1.4 Step 4 968 Packet 2 is received at Host A. Remember, its time is set to one 969 hour earlier than Host B. Internally, it must note: 971 Packet Sequence Number of the last packet received: 12 972 Time the last packet was received : 10:00:12 974 Note, this timestamp is in Host A time. It has nothing whatsoever to 975 do with Host B time. 977 So, now, Host A can calculate total end-to-end time. That is: 979 End-to-End Time = Time Last Received - Time Last Sent 981 For example, packet 25 was sent by Host A at 10:00:00. Packet 12 was 982 received by Host A at 10:00:12 so: 984 End-to-End time = 10:00:12 - 10:00:00 or 12 (Server and Network RT 985 delay combined). This time may also be called total Overall Round- 986 Trip Time (RTT) which includes Network RTT and Host Response Time. 988 This derived metric we will call Delta Time Last Sent (DELTATLS) 990 We can now also calculate round trip delay. The formula is: 992 Round trip delay = (Delta Time Last Sent - Delta Time Last Received) 993 Or: 995 Round trip delay = 12 - 4 or 8 997 Now, the only problem is that at this point all metrics are in Host A 998 only and not exposed in a packet. To do that, we need a third packet. 1000 Note: this simple example assumes one send and one receive. That 1001 is done only for purposes of explaining the function of the PDM. In 1002 cases where there are multiple packets returned, one would take the 1003 time in the last packet in the sequence. The calculations of such 1004 timings and intelligent processing is the function of post-processing 1005 of the data. 1007 B.1.5 Step 5 1009 Packet 3 is sent from Host A to Host B. 1011 +----------+ +----------+ 1012 | | | | 1013 | Host | ----------> | Host | 1014 | A | | B | 1015 | | | | 1016 +----------+ +----------+ 1018 PDM Contents: 1020 PSNTP : Packet Sequence Number This Packet: 26 1021 PSNLR : Packet Sequence Number Last Received: 12 1022 DELTATLR : Delta Time Last Received: 0 1023 SCALEDTLS: Scale of Delta Time Last Received 0 1024 DELTATLS : Delta Time Last Sent: A688 (scaled value) 1025 SCALEDTLR: Scale of Delta Time Last Received: 30 (48 decimal) 1027 To calculate Two-Way Delay, any packet capture device may look at 1028 these packets and do what is necessary. 1030 B.2 Other Flows 1032 What we have discussed so far is a simple flow with one packet sent 1033 and one returned. Let's look at how PDM may be useful in other 1034 types of flows. 1036 B.2.1 PDM Flow - One Way Traffic 1038 The flow on a particular session may not be a send-receive paradigm. 1039 Let us consider some other situations. In the case of a one-way 1040 flow, one might see the following: 1042 Note: The time is expressed in generic units for simplicity. That 1043 is, these values do not represent a number of attoseconds, 1044 microseconds or any other real units of time. 1046 Packet Sender PSN PSN Delta Time Delta Time 1047 This Packet Last Recvd Last Recvd Last Sent 1048 ===================================================================== 1049 1 Server 1 0 0 0 1050 2 Server 2 0 0 5 1051 3 Server 3 0 0 12 1052 4 Server 4 0 0 20 1054 What does this mean and how is it useful? 1056 In a one-way flow, only the Delta Time Last Sent will be seen as 1057 used. Recall, Delta Time Last Sent is the difference between the 1058 send of one packet from a device and the next. This is a measure of 1059 throughput for the sender - according to the sender's point of view. 1060 That is, it is a measure of how fast is the application itself (with 1061 stack time included) able to send packets. 1063 How might this be useful? If one is having a performance issue at 1064 the client and sees that packet 2, for example, is sent after 5 time 1065 units from the server but takes 10 times that long to arrive at the 1066 destination, then one may safely conclude that there are delays in 1067 the path other than at the server which may be causing the delivery 1068 issue of that packet. Such delays may include the network links, 1069 middle-boxes, etc. 1071 Now, true one-way traffic is quite rare. What people often mean by 1072 "one-way" traffic is an application such as FTP where a group of 1073 packets (for example, a TCP window size worth) is sent, then the 1074 sender waits for acknowledgment. This type of flow would actually 1075 fall into the "multiple-send" traffic model. 1077 B.2.2 PDM Flow - Multiple Send Traffic 1079 Assume that two packets are sent for each ACK from the server. For 1080 example, a TCP flow will do this, per RFC1122 [RFC1122] Section- 1081 4.2.3. 1083 Packet Sender PSN PSN Delta Time Delta Time 1084 This Packet Last Recvd Last Recvd Last Sent 1085 ===================================================================== 1086 1 Server 1 0 0 0 1087 2 Server 2 0 0 5 1088 3 Client 1 2 20 0 1089 4 Server 3 1 10 15 1090 How might this be used? 1092 Notice that in packet 3, the client has a value of Delta Time Last 1093 received of 20. Recall that Delta Time Last Received is the Send 1094 time of packet 3 - receive time of packet 2. So, what does one know 1095 now? In this case, Delta Time Last Received is the processing time 1096 for the Client to send the next packet. 1098 How to interpret this depends on what is actually being sent. 1099 Remember, PDM is not being used in isolation, but to supplement the 1100 fields found in other headers. Let's take some examples: 1102 1. Client is sending a standalone TCP ACK. One would find this by 1103 looking at the payload length in the IPv6 header and the TCP 1104 Acknowledgement field in the TCP header. So, in this case, the 1105 client is taking 20 units to send back the ACK. This may or may not 1106 be interesting. 1108 2. Client is sending data with the packet. Again, one would find 1109 this by looking at the payload length in the IPv6 header and the TCP 1110 Acknowledgement field in the TCP header. So, in this case, the 1111 client is taking 20 units to send back data. This may represent 1112 "User Think Time". Again, this may or may not be interesting, in 1113 isolation. But, if there is a performance problem receiving data at 1114 the server, then taken in conjunction with RTT or other packet timing 1115 information, this information may be quite interesting. 1117 Of course, one also needs to look at the PSN Last Received field to 1118 make sure of the interpretation of this data. That is, to make 1119 sure that the Delta Last Received corresponds to the packet of 1120 interest. 1122 The benefits of PDM are that we have such information available in a 1123 uniform manner for all applications and all protocols without 1124 extensive changes required to applications. 1126 B.2.3 PDM Flow - Multiple Send with Errors 1128 Let us now look at a case of how PDM may be able to help in a case of 1129 TCP retransmission and add to the information that is sent in the TCP 1130 header. 1132 Assume that three packets are sent with each send from the server. 1134 From the server, this is what is seen. 1136 Pkt Sender PSN PSN Delta Time Delta Time TCP Data 1137 This Pkt LastRecvd LastRecvd LastSent SEQ Bytes 1138 ===================================================================== 1139 1 Server 1 0 0 0 123 100 1140 2 Server 2 0 0 5 223 100 1141 3 Server 3 0 0 5 333 100 1143 The client, however, does not receive all the packets. From the 1144 client, this is what is seen for the packets sent from the server. 1146 Pkt Sender PSN PSN Delta Time Delta Time TCP Data 1147 This Pkt LastRecvd LastRecvd LastSent SEQ Bytes 1148 ===================================================================== 1149 1 Server 1 0 0 0 123 100 1150 2 Server 3 0 0 5 333 100 1152 Let's assume that the server now retransmits the packet. (Obviously, 1153 a duplicate acknowledgment sequence for fast retransmit or a 1154 retransmit timeout would occur. To illustrate the point, these 1155 packets are being left out.) 1157 So, then if a TCP retransmission is done, then from the client, this 1158 is what is seen for the packets sent from the server. 1160 Pkt Sender PSN PSN Delta Time Delta Time TCP Data 1161 This Pkt LastRecvd LastRecvd LastSent SEQ Bytes 1162 ===================================================================== 1163 1 Server 4 0 0 30 223 100 1165 The server has resent the old packet 2 with TCP sequence number of 1166 223. The retransmitted packet now has a PSN This Packet value of 4. 1168 The Delta Last Sent is 30 - the time between sending the packet with 1169 PSN of 3 and this current packet. 1171 Let's say that packet 4 is lost again. Then, after some amount of 1172 time (RTO) then the packet with TCP sequence number of 223 is resent. 1174 From the client, this is what is seen for the packets sent from the 1175 server. 1177 Pkt Sender PSN PSN Delta Time Delta Time TCP Data 1178 This Pkt LastRecvd LastRecvd LastSent SEQ Bytes 1179 ===================================================================== 1180 1 Server 5 0 0 60 223 100 1181 If now, this packet arrives at the destination, one has a very good 1182 idea that packets exist which are being sent from the server as 1183 retransmissions and not arriving at the client. This is because the 1184 PSN of the resent packet from the server is 5 rather than 4. If we 1185 had used TCP sequence number alone, we would never have seen this 1186 situation. The TCP sequence number in all situations is 223. 1188 This situation would be experienced by the user of the application 1189 (the human being actually sitting somewhere) as a "hangs" or long 1190 delay between packets. On large networks, to diagnose problems such 1191 as these where packets are lost somewhere on the network, one has to 1192 take multiple traces to find out exactly where. 1194 The first thing is to start with doing a trace at the client and the 1195 server. So, we can see if the server sent a particular packet and 1196 the client received it. If the client did not receive it, then we 1197 start tracking back to trace points at the router right after the 1198 server and the router right before the client. Did they get these 1199 packets which the server has sent? This is a time consuming 1200 activity. 1202 With PDM, we can speed up the diagnostic time because we may be able 1203 to use only the trace taken at the client to see what the server is 1204 sending. 1206 Appendix C: Potential Overhead Considerations 1208 One might wonder as to the potential overhead of PDM. First, PDM is 1209 entirely optional. That is, a site may choose to implement PDM or 1210 not as they wish. If they are happy with the costs of PDM vs. the 1211 benefits, then the choice should be theirs. 1213 Below is a table outlining the potential overhead in terms of 1214 additional time to deliver the response to the end user for various 1215 assumed RTTs. 1217 Bytes RTT Bytes Bytes New Overhead 1218 in Packet Per Millisec in PDM RTT 1219 ===================================================================== 1220 1000 1000 milli 1 16 1016.000 16.000 milli 1221 1000 100 milli 10 16 101.600 1.600 milli 1222 1000 10 milli 100 16 10.160 .160 milli 1223 1000 1 milli 1000 16 1.016 .016 milli 1225 Below are some examples of actual RTTs for packets traversing large 1226 enterprise networks. The first example is for packets going to 1227 multiple business partners. 1229 Bytes RTT Bytes Bytes New Overhead 1230 in Packet Per Millisec in PDM RTT 1231 ===================================================================== 1232 1000 17 milli 58 16 17.360 .360 milli 1234 The second example is for packets at a large enterprise customer 1235 within a data center. Notice that the scale is now in microseconds 1236 rather than milliseconds. 1238 Bytes RTT Bytes Bytes New Overhead 1239 in Packet Per Microsec in PDM RTT 1240 ===================================================================== 1241 1000 20 micro 50 16 20.320 .320 micro 1243 Acknowledgments 1245 The authors would like to thank Keven Haining, Al Morton, Brian 1246 Trammel, David Boyes, Bill Jouris, Richard Scheffenegger, and Rick 1247 Troth for their comments and assistance. We would also like to thank 1248 Tero Kivinen for his detailed and perceptive review. 1250 Authors' Addresses 1252 Nalini Elkins 1253 Inside Products, Inc. 1254 36A Upper Circle 1255 Carmel Valley, CA 93924 1256 United States 1257 Phone: +1 831 659 8360 1258 Email: nalini.elkins@insidethestack.com 1259 http://www.insidethestack.com 1261 Robert M. Hamilton 1262 Chemical Abstracts Service 1263 A Division of the American Chemical Society 1264 2540 Olentangy River Road 1265 Columbus, Ohio 43202 1266 United States 1267 Phone: +1 614 447 3600 x2517 1268 Email: rhamilton@cas.org 1269 http://www.cas.org 1271 Michael S. Ackermann 1272 Blue Cross Blue Shield of Michigan 1273 P.O. Box 2888 1274 Detroit, Michigan 48231 1275 United States 1276 Phone: +1 310 460 4080 1277 Email: mackermann@bcbsm.com 1278 http://www.bcbsm.com