Debian Codegraph @

rfc2170.html

raw
1 <pre>Network Working Group W. Almesberger
2 Request for Comments: 2170 J. Le Boudec
3 Category: Informational P. Oechslin
4 LRC, DI-EPFL, Switzerland
5 July 1997
6
7
8 <span class="h1">Application REQuested IP over ATM (AREQUIPA)</span>
9
10 Status of this Memo
11
12 This memo provides information for the Internet community. This memo
13 does not specify an Internet standard of any kind. Distribution of
14 this memo is unlimited.
15
16 IESG Note:
17
18 This RFC has not had the benefit of the rigorous peer review that is
19 part of the process in an IETF working group. The technology it
20 describes has been implemented and is now undergoing testing. It
21 would be wise to analyze the results of this testing as well as to
22 understand alternatives before committing to this approach for IP
23 over ATM with QoS guarantees.
24
25 Abstract
26
27 This document specifies a method for allowing ATM-attached hosts that
28 have direct ATM connectivity to set up end-to-end IP over ATM
29 connections within the reachable ATM cloud, on request from
30 applications, and for the exclusive use by the requesting
31 applications. This allows the requesting applications to benefit in a
32 straightforward way from ATM's inherent ability to guarantee the
33 quality of service (QoS).
34
35 Given a mapping from service classes, as defined by INTSERV[6], to
36 ATM traffic descriptors, Arequipa can be used to implement integrated
37 services over ATM link layers. Usage of Arequipa to provide
38 integrated services even if ATM is only available for intermediate
39 links is not discussed in this document but should be straight-
40 forward.
41
42 The major advantage of using an approach like Arequipa is that it
43 needs to be implemented only on the hosts using it. It requires no
44 extra service (eg. NHRP or RSVP) to be deployed on the switches or
45 routers of the ATM cloud.
46
47
48
49
50
51
52 <span class="grey">Almesberger, et. al. Informational [Page 1]</span></pre>
53 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-2" ></span>
54 <span class="grey"><a href="./rfc2170">RFC 2170</a> AREQUIPA July 1997</span>
55
56
57 We discuss the implementation of Arequipa for hosts running IPv4 and
58 IPv6. As an illustration, we also discuss how World-Wide-Web
59 applications can use Arequipa to deliver documents with a guaranteed
60 quality of service.
61
62 In particular we show how
63
64 - Arequipa can be implemented in IPv4 by slightly modifying the
65 - Arequipa can be implemented in IPv6[3] by the appropriate use of
66 flow labels and the extension of the neighbour cache,
67 - Arequipa can be used in the Web by adding extra information in
68 the headers of HTTP requests and responses.
69
70 Finally, we address safety and security implications.
71
72 <span class="h2"><a class="selflink" id="section-1" href="#section-1">1</a>. Introduction</span>
73
74 QoS guarantees are important for delivery of multi-media data and
75 commercial services on the Internet. When two applications on
76 machines running IP over ATM need to transfer data, all the necessary
77 gears to guarantee QoS can be found in the ATM layer. We consider
78 the case where it is desired to use end-to-end ATM connections
79 between applications residing on ATM hosts that have end-to-end ATM
80 connectivity.
81
82 Opening direct ATM connections between two applications is possible,
83 but then the already available transport protocols, like TCP, can not
84 be reused.
85
86 This is why we propose Application REQuested IP over ATM (AREQUIPA).
87 Arequipa allows applications to request that two machines be
88 connected by a direct ATM connection with given QoS at the link
89 level. Arequipa makes sure that only data from the applications that
90 requested the connection actually goes through that connection. After
91 setup of the Arequipa connection, the applications can use the
92 standard IP protocol suite to exchange data.
93
94 <span class="h2"><a class="selflink" id="section-2" href="#section-2">2</a>. API semantics</span>
95
96 We now define a semantical API for Arequipa. Note that an actual API
97 may perform additional functions (eg. mapping of a given service
98 specification to ATM traffic descriptors)
99
100 We define the three new API functions for the TCP/IP stack:
101
102 Arequipa_preset (socket_descriptor, destination IP address,
103 destination protid/port, destination ATM Address,
104 ATM service and QoS parameters)
105
106
107
108 <span class="grey">Almesberger, et. al. Informational [Page 2]</span></pre>
109 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-3" ></span>
110 <span class="grey"><a href="./rfc2170">RFC 2170</a> AREQUIPA July 1997</span>
111
112
113 Arequipa_preset establishes or prepares establishment of a new ATM
114 connection to the given address with the given ATM service and QoS.
115 It makes sure that any further data sent on the specified socket
116 will use the new ATM connection.
117
118 Arequipa_preset is invoked before a TCP/IP connection is
119 established or before sending data(grams), respectively. (Active
120 open.)
121
122 Arequipa_expect (socket_descriptor, allow)
123
124 Arequipa_expect prepares the system to use an expected incoming
125 Arequipa connection for outgoing traffic of a given socket. If
126 allow equals TRUE then, as soon as the socket receives data from an
127 incoming Arequipa connection, all its return traffic is sent over
128 that Arequipa connection. If allow equals FALSE the traffic from
129 that socket is always sent over the standard IP route. Note that
130 Arequipa_expect is only applicable to connection oriented sockets,
131 eg. TCP sockets or connected UDP sockets.
132
133 Arequipa_expect is invoked by the peer which is expecting
134 data(grams) or accepting connections. (Passive open.) It is
135 typically called immediately after a socket has been created. It
136 may also be called when a data transfer is already going on.
137
138 Arequipa_close (socket_descriptor)
139
140 Closes the corresponding ATM connection. Any further traffic
141 between the endpoints is routed like other traffic. Arequipa_close
142 is implied when closing the socket.
143
144 Note that the use of Arequipa_expect or _preset only reflects the
145 direction of the initial dialog in the Arequipa connection. Actual
146 data can flow in both directions.
147
148 An actual implementation may use less arguments for Arequipa_preset
149 if some of the information is already passed by other networking
150 operations.
151
152 <span class="h2"><a class="selflink" id="section-3" href="#section-3">3</a>. Implementation with IPv4</span>
153
154 To implement Arequipa with IPv4, ATMARP must be able not only to
155 handle associations of ATM addresses and IP addresses, but also
156 associations of one ATM address with an IP address plus endpoint
157 (socket). This allows to dedicate an ATM connection for the traffic
158 between two endpoints.
159
160
161
162
163
164 <span class="grey">Almesberger, et. al. Informational [Page 3]</span></pre>
165 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-4" ></span>
166 <span class="grey"><a href="./rfc2170">RFC 2170</a> AREQUIPA July 1997</span>
167
168
169 For the active open, a destination ATM address must be associated
170 with a socket. In systems using per-socket route and ARP caching,
171 this can be done by presetting the caches with the appropriate
172 values. Establishment of the SVC is delegated to ATMARP. Care must be
173 taken that routing and ARP information obtained through Arequipa does
174 not leak to other parts of the system.
175
176 For the passive open, an incoming SVC must be associated with the
177 socket that terminates the corresponding connection or data flow.
178 Most of this functionality is already available in the existing
179 protocol stack. To avoid an incoming Arequipa SVC to be mistaken for
180 an IP-over-ATM SVC, the setup message uses a specific Broadband High
181 Layer Identifier (BHLI), see below. Seeing the BHLI, ATMARP knows
182 that the SVC is of the dedicated type. The socket to which it has to
183 be associated is identified as soon as a datagram is received through
184 the SVC. If an Arequipa_expect has been done for that socket, then
185 the SVC is used for all return traffic of that socket.
186
187 If application A1 on host H1 wants a direct ATM connection to
188 application A2 on host H2 it does the following:
189
190 Both applications first get in contact using the standard IP over ATM
191 to exchange the ATM address of the receiver (atm2) and the endpoints
192 (S1, S2) (i.e. protocol and port number; we assume that a protocol
193 with ports, such as TCP or UDP, is used) at both hosts between which
194 communication will occur. How this is performed depends on the
195 application protocols. In <a href="#section-5">Section 5</a> we give an example for HTTP.
196
197 A2 invokes Arequipa_expect to indicate that it wants to make use of
198 an expected incoming ATM connection.
199
200 A1 invokes Arequipa_preset to open or prepare opening of an ATM
201 connection to H2 using ATM address atm2 and the QoS desired by A1 as
202 soon as data is sent through S1. The connection is associated with S1
203 such that no other traffic (e.g. generated by other applications)
204 uses the new ATM connection.
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220 <span class="grey">Almesberger, et. al. Informational [Page 4]</span></pre>
221 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-5" ></span>
222 <span class="grey"><a href="./rfc2170">RFC 2170</a> AREQUIPA July 1997</span>
223
224
225 An Arequipa connection shall be signaled by using the procedures and
226 codings described in <a href="./rfc1755">RFC1755</a> [<a href="#ref-7" title="D. Grossman">7</a>], with the addition that the BHLI
227 information element be included in the SETUP message, with the
228 following coding:
229
230 ------------------------------------------------------------------
231 | bb_high_layer_information |
232 ------------------------------------------------------------------
233 | high_layer_information_type 3 (vendor-specific |
234 | application id.) |
235 | high_layer_information 00-60-D7 (EPFL OUI) |
236 | 01-00-00-01 (Arequipa) |
237 ------------------------------------------------------------------
238
239 As soon as data arrives from H1:S1 at H2:S2, the ATM connection the
240 data has arrived on is identified as the dedicated connection for
241 this data flow and S2 is changed to exclusively send on that
242 connection.
243
244 From this point on all traffic exchanged between S1 of A1 and S2 of
245 A2 will use the new ATM connection with the desired QoS.
246
247 Note that it is possible for H1 and H2 to belong to the same LIS [<a href="#ref-2" title="C. Villamizar">2</a>]
248 and still decide to use an Arequipa connection between applications,
249 in addition to one or several other, non-Arequipa ATM connections
250 between hosts H1 and H2. There may also exist several Arequipa
251 connections between two hosts.
252
253 <span class="h2"><a class="selflink" id="section-4" href="#section-4">4</a>. Implementation with IPv6</span>
254
255 With IPv6, sources take advantage of the Flow Label field in the IPv6
256 header [<a href="#ref-3" title="R. and S. Deering">3</a>].
257
258 We assume as in [<a href="#ref-4" title="E. Nordmark and W.A. Simpson">4</a>] that the conceptual host model uses, among
259 others, a neighbour cache and a destination cache. The destination
260 cache holds entries about destinations to which traffic has been sent
261 recently, while the neighbour cache holds entries about neighbours to
262 which traffic has been sent recently. With the classical IP over ATM
263 model [<a href="#ref-1" title="RFC1577">1</a>], entries in the neighbour cache can only refer to systems
264 in the same LIS; we propose to go beyond this limitation and allow
265 systems beyond the LIS to appear there and be treated as neighbours,
266 in the case where a direct link level connection (here, an ATM
267 connection) can be established.
268
269 The destination is keyed in [<a href="#ref-4" title="E. Nordmark and W.A. Simpson">4</a>] by the IP (destination) address. We
270 replace this by the IP (destination) address and flow label.
271
272
273
274
275
276 <span class="grey">Almesberger, et. al. Informational [Page 5]</span></pre>
277 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-6" ></span>
278 <span class="grey"><a href="./rfc2170">RFC 2170</a> AREQUIPA July 1997</span>
279
280
281 We assume that with IPv6, a mechanism will be provided for
282 applications to request flow labels which, at the host, form a unique
283 flow-label/destination-address pair. This will prevent two different
284 flows which go to the same destination from accidentally using the
285 same flow label. Such a uniqueness requirement is also desirable for
286 other applications which rely on flow-label/destination-address
287 pairs, like for example RSVP.
288
289 A typical scenario is:
290
291 Application A1 on host H1 and application A2 on host H2 first get in
292 contact using the standard IP over ATM to exchange their ATM address
293 (atm1, atm2) and to define a protocol, port number pair (S1, S2) and
294 flow labels (L1, L2) for the communication over the ATM connection.
295 (We assume that a protocol with ports, such as TCP or UDP, is used).
296 How this is performed depends on the application protocols. In
297 <a href="#section-5">Section 5</a> we give an example for HTTP.
298
299 A2 tells its networking entity that it wants to send its outgoing
300 packets with flow label L2 over an expected incoming ATM connection.
301 A1 tells its data link entity to open an ATM connection to H2 using
302 ATM address atm2, with the QoS desired by A1. The connection is
303 associated with L1 and L2 as explained below so that no other traffic
304 generated by other applications uses the new ATM connection.
305
306 From this point on all traffic exchanged between applications A1 on
307 H1 and application A2 on H2 will use this ATM connection.
308
309 An example of destination and neighbour cache entries at H1 is given
310 below.
311
312 Destination Cache
313 IPAddr flowLabel neighbourCache pathMTU
314 H2 L1 ptr1 (1)
315 H2 * ptr2 (2)
316
317 Neighbour Cache
318 IPAddr linkLayerAddr isRouter reachabilityState invalidationTimer
319 H2 v2 no (3) t2
320 R3 v3 yes REACHABLE t3
321
322
323 In the example, the route to destination H2 for all traffic other
324 than the one using the ATM connection requested between application
325 A1 and A2 uses the default route (perhaps set up by the classical IP
326 model), with router R3 as the next hop; v2 is a pointer to an ATM
327 interface and a VPCI/VCI that identifies the Arequipa connection.
328 Similarly, v3 points to the ATM connection to router R3. ptr1 points
329
330
331
332 <span class="grey">Almesberger, et. al. Informational [Page 6]</span></pre>
333 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-7" ></span>
334 <span class="grey"><a href="./rfc2170">RFC 2170</a> AREQUIPA July 1997</span>
335
336
337 to the first line in Neighbour Cache, and ptr2 to the second one.
338 Path MTUs (1) and (2) are obtained by ATM signaling; they may be
339 different. Reachability state (3) is determined as usual by the
340 reachability protocol [<a href="#ref-4" title="E. Nordmark and W.A. Simpson">4</a>].
341
342 Host H1 must restrict the use of this ATM connection to datagrams
343 with flow label L1. Other traffic from H1 to H2 must use the generic
344 entry in the destination table (flow label = "*"). Host H1 must
345 restrict the use of flow label L1 for destination H2 to traffic
346 generated by application A1 on port S1. (The same holds by analogy
347 for host H2).
348
349 On the receiving side, host H2 may use label L1 for routing
350 internally the IP packets to the appropriate entity.
351
352 <span class="h2"><a class="selflink" id="section-5" href="#section-5">5</a>. Example: Arequipa for the Web</span>
353
354 This is a brief explanation of how Web [<a href="#ref-5" title="H. Nielsen">5</a>] servers and browsers can
355 use Arequipa to transmit documents with a guaranteed QoS.
356
357 What we describe below does not violate the standards of HTML and
358 HTTP but makes use of their built-in extensibility. The server and
359 client we describe can thus interact seamlessly with non-modified
360 servers or clients. A similar extension could be used if Web
361 documents were to be exchanged using RSVP.
362
363 Browsers add one extra field in all their requests or responses to
364 indicate their ATM address. Web documents are extended with meta
365 information to describe the ATM service and corresponding QoS needed
366 to transmit them. Note that this information could be in form of an
367 intserv flowspec and mapped to ATM traffic descriptors.
368
369 If a browser always wants documents with QoS meta-information to be
370 delivered using Arequipa, it adds an additional field in its request
371 to indicate the port on which it is expecting the data.
372
373 If a browser wants to decide whether Arequipa should be used or not,
374 it does not give the port on which the server should send the data.
375
376 When a server gets a request with an ATM address, it checks whether
377 the requested document has QoS meta-information. If this is not the
378 case, it delivers the document like a standard server. If the
379 document has QoS meta-information, the server looks for a port
380 information in the request. If it finds a port, it opens an Arequipa
381 socket (Arequipa_preset) to the ATM address of the client with the
382 QoS given in the document. It sends the reply through this new
383 connection. If the server finds no port information, it sends only
384 the header of the reply (which includes meta-information) over the
385
386
387
388 <span class="grey">Almesberger, et. al. Informational [Page 7]</span></pre>
389 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-8" ></span>
390 <span class="grey"><a href="./rfc2170">RFC 2170</a> AREQUIPA July 1997</span>
391
392
393 standard HTTP connection, as if the client had issued a HEAD or GET-
394 IF-MODIFIED request.
395
396 When a client receives the header of a document it can decide whether
397 it wants the document to be transmitted using Arequipa or not. A
398 client without a priori knowledge about the document, may therefore
399 always want to retrieve the header before requesting the full
400 document.
401
402 Illustration:
403
404 A client requests some documents but wants to decide if QoS sensitive
405 documents should be sent using Arequipa or not. Thus it adds to its
406 requests its ATM address but not the socket information.
407
408 GET batman.mpeg
409 UserAgent: MyAgent/1.0
410 ATM-address: my_public_address.my_private_address
411
412 The server checks batman.mpeg for QoS meta info. It finds the meta
413 info and sees an ATM address, but no socket pragma in the request. It
414 only returns the header of the document, which includes the meta-
415 information:
416
417
418 HTTP/1.0 200 OK
419 Server: MyAgent/1.0
420 ATM-Service: CBR
421 ATM-QoS-PCR: 2000
422 Content-type: video/mpeg
423
424
425 The client sees the QoS info and decides that it wants to download
426 the document using Arequipa. It opens a TCP socket for listening,
427 makes the Arequipa_expect call and sends the following request:
428
429 GET batman.mpeg
430 UserAgent: MyAgent/1.0
431 ATM-address: my_public_address.my_private_address
432 Pragma: socket=TCP.8090
433
434
435
436
437
438
439
440
441
442
443
444 <span class="grey">Almesberger, et. al. Informational [Page 8]</span></pre>
445 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-9" ></span>
446 <span class="grey"><a href="./rfc2170">RFC 2170</a> AREQUIPA July 1997</span>
447
448
449 Again the server checks batman.mpeg for QoS meta info. It finds the
450 meta info and sees the ATM address and the socket pragma in the
451 request. It creates a TCP socket, makes the Arequipa_preset call,
452 connects its TCP socket to the one of the client and sends the
453 response over the new TCP connection:
454
455 HTTP/1.0 200 OK
456 Server: MyAgent/1.0 ATM.address
457 ATM-Service: CBR
458 ATM-QoS-PCR: 2000
459 Content-type: video/mpeg
460
461 &lt;mpeg data&gt;
462
463 When the server sends the data over the new TCP connection it also
464 sends a copy of the response header over the TCP connection on which
465 the request was made. For example, this allows a browser to spawn a
466 viewer before requesting the data, to give the Arequipa connection to
467 the viewer and to still get the status of the request over the normal
468 TCP connection.
469
470 <span class="h2"><a class="selflink" id="section-6" href="#section-6">6</a>. Safety considerations (loops)</span>
471
472 A major concern about ATM shortcuts in IP networks are routing loops.
473 Arequipa is not prone to such dangers since it establishes
474 connections between applications and not between hosts. All datagrams
475 traveling through an Arequipa connection are destined for a given
476 socket on the machine at the end of the connection and don't need to
477 be forwarded by the IP layer. Therefore, neither hosts nor routers
478 implementing Arequipa as described in this document must ever forward
479 IP packets received over Arequipa connections.
480
481 <span class="h2"><a class="selflink" id="section-7" href="#section-7">7</a>. Security considerations</span>
482
483 The main security problem we see with Arequipa is that it could be
484 used to bypass IP firewalls.
485
486 IP firewalls are used to protect private networks connected to
487 untrusted IP networks. The network is configured such that all
488 traffic going into or coming from the protected network has to go
489 through the machine(s) acting as a firewall.
490
491 If hosts in a network protected by a firewall are able to establish
492 direct ATM connections to hosts outside the protected network, then
493 Arequipa could be used to bypass the firewall. To avoid this, hosts
494 inside a protected network should not be given direct connectivity to
495 the outside of the network.
496
497
498
499
500 <span class="grey">Almesberger, et. al. Informational [Page 9]</span></pre>
501 <hr class='noprint'/><!--NewPage--><pre class='newpage'><span id="page-10" ></span>
502 <span class="grey"><a href="./rfc2170">RFC 2170</a> AREQUIPA July 1997</span>
503
504
505 Arequipa can be used in a safe way by machines inside and outside a
506 protected network, if an application proxy is installed on the
507 firewall. In the Web, this is a typical scenario. Proxy HTTP servers
508 are often found on firewalls, not only for security reasons, but also
509 for caching. If an application proxy is used, each host can establish
510 an Arequipa connection to the proxy which can then relay and monitor
511 the traffic across the firewall.
512
513 Note that hosts can easily identify (and refuse) unsolicited Arequipa
514 connections by the BHLI identifier that is passed at connection
515 setup.
516
517 <span class="h2"><a class="selflink" id="section-8" href="#section-8">8</a>. References</span>
518
519 [<a id="ref-1">1</a>] Laubach, M., Classical IP and ARP over ATM, <a href="./rfc1577">RFC1577</a>,
520 January 1994.
521
522 [<a id="ref-2">2</a>] Cole, R. G., D. H. Shur, C. Villamizar, IP over ATM: A Framework
523 Document, <a href="./rfc1932">RFC1932</a>, April 1996.
524
525 [<a id="ref-3">3</a>] Hinden, R. and S. Deering, Internet Protocol Version (IPv6)
526 Addressing Architecture, <a href="./rfc1884">RFC1884</a>, December 1995.
527
528 [<a id="ref-4">4</a>] Narten, T., E. Nordmark and W.A. Simpson, Neighbour Discovery for
529 IPv6 (IPv6), <a href="./rfc1970">RFC1970</a>, August 1996.
530
531 [<a id="ref-5">5</a>] Berners-Lee, T., R. Fielding, H. Nielsen, Hypertext Transfer
532 Protocol -- HTTP/1.0, <a href="./rfc1945">RFC1945</a>, May 1996.
533
534 [<a id="ref-6">6</a>] Shenker, S./J. Wroclawski, Network Element Service Specification
535 Template, Work in Progess, November, 1995.
536
537 [<a id="ref-7">7</a>] Perez, M., F.-C. Liaw, A. Mankin, E. Hoffman, D. Grossman, A.
538 Malis, ATM Signaling Support for IP over ATM, <a href="./rfc1755">RFC1755</a>, February
539 1995.
540
541 <span class="h2"><a class="selflink" id="section-9" href="#section-9">9</a>. Authors' Address</span>
542
543 Werner Almesberger,
544 Jean-Yves Le Boudec,
545 Philippe Oechslin (contact author)
546
547 Laboratoire de Reseaux de Communication
548 Swiss Federal Institute of Technology (EPFL)
549 1015 Lausanne
550 Switzerland
551
552 email: {almesber, leboudec, oechslin}@di.epfl.ch
553
554
555
556 Almesberger, et. al. Informational [Page 10]
557 </pre>