Network Working Group P. Saint-Andre Internet-Draft XMPP Standards Foundation Intended status: Informational January 16, 2008 Expires: July 19, 2008 Interdomain Presence Scaling Analysis for the Extensible Messaging and Presence Protocol (XMPP) draft-saintandre-xmpp-presence-analysis-03 Status of this Memo By submitting this Internet-Draft, each author represents that any applicable patent or other IPR claims of which he or she is aware have been or will be disclosed, and any of which he or she becomes aware will be disclosed, in accordance with Section 6 of BCP 79. Internet-Drafts are working documents of the Internet Engineering Task Force (IETF), its areas, and its working groups. Note that other groups may also distribute working documents as Internet- Drafts. Internet-Drafts are draft documents valid for a maximum of six months and may be updated, replaced, or obsoleted by other documents at any time. It is inappropriate to use Internet-Drafts as reference material or to cite them other than as "work in progress." The list of current Internet-Drafts can be accessed at http://www.ietf.org/ietf/1id-abstracts.txt. The list of Internet-Draft Shadow Directories can be accessed at http://www.ietf.org/shadow.html. This Internet-Draft will expire on July 19, 2008. Copyright Notice Copyright (C) The IETF Trust (2008). Abstract This document analyzes the network impact of presence sharing between domains that federate using the Extensible Messaging and Presence Protocol (XMPP). Saint-Andre Expires July 19, 2008 [Page 1] Internet-Draft XMPP Presence Analysis January 2008 Table of Contents 1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . 3 2. Assumptions . . . . . . . . . . . . . . . . . . . . . . . . . 3 3. Protocol Flows . . . . . . . . . . . . . . . . . . . . . . . . 4 4. Analysis . . . . . . . . . . . . . . . . . . . . . . . . . . . 6 4.1. Constants . . . . . . . . . . . . . . . . . . . . . . . . 6 4.2. Initial Stanzas . . . . . . . . . . . . . . . . . . . . . 7 4.3. State-Change Stanzas . . . . . . . . . . . . . . . . . . . 7 4.4. Termination Stanzas . . . . . . . . . . . . . . . . . . . 7 4.5. Bottom Line . . . . . . . . . . . . . . . . . . . . . . . 7 5. Scenarios . . . . . . . . . . . . . . . . . . . . . . . . . . 8 5.1. Enterprises in Different Industries . . . . . . . . . . . 8 5.2. Enterprises in the Same Industry . . . . . . . . . . . . . 9 5.3. Medium-Sized Service Providers . . . . . . . . . . . . . . 10 5.4. Large Service Providers . . . . . . . . . . . . . . . . . 12 5.5. Very Large Service Providers . . . . . . . . . . . . . . . 13 6. Optimizations . . . . . . . . . . . . . . . . . . . . . . . . 14 7. Comparison With Other Presence Technologies . . . . . . . . . 14 8. Security Considerations . . . . . . . . . . . . . . . . . . . 14 9. Informative References . . . . . . . . . . . . . . . . . . . . 14 Appendix A. Copying Conditions . . . . . . . . . . . . . . . . . 15 Author's Address . . . . . . . . . . . . . . . . . . . . . . . . . 16 Intellectual Property and Copyright Statements . . . . . . . . . . 17 Saint-Andre Expires July 19, 2008 [Page 2] Internet-Draft XMPP Presence Analysis January 2008 1. Introduction Presence is information about the network availability of an individual (or, more precisely, of a presence address of the kind that is often but not necessarily associated with an individual). As typically designed and deployed, presence is shared only with authorized entities, where the authorization takes the form of a subscription. (In this document, we employ the term "user" to signify an account that generates presence information and the term "contact" to signify an annount that is subscribed to the user's presence.) The sharing of basic presence information can result in a large volume of traffic as users log on or off throughout the life of a presence session, especially for users with large numbers of contacts (e.g., the author of this document has over 1,700 contacts in his presence-enabled contact list). The volume is increased by communication of information beyond basic on-off network availability, such as availability substates (e.g., "away" and "do not disturb"). The volume is further increased if the presence "transport" is used to communicate information such as device capabilities, geolocation, mood, activity, even the music to which a user is listening. Such traffic may be a concern even in a standalone presence domain. However, when presence is shared across domain boundaries through presence "federation", then such traffic may introduce a more significant impact on the functioning of the Internet as a whole. Therefore it is important to analyze the traffic generated during interdomain communication of presence information. This document provides such an analysis regarding the Extensible Messaging and Presence Protocol (XMPP) as defined in [XMPP-CORE] and [XMPP-IM]. 2. Assumptions The XMPP presence model is based on the following assumptions: 1. A user shares presence only with a contact whom the user has explicitly authorized via a presence subscription. 2. Presence subscriptions are long-lived: they last across presence sessions and indeed as long as the user and contact maintain their XMPP accounts (until and unless the subscription is cancelled by one of the parties). 3. The typical subscription state is a bidirectional subscription from the contact to the user and from the user to the contact (so that both entities can view each other's presence). Saint-Andre Expires July 19, 2008 [Page 3] Internet-Draft XMPP Presence Analysis January 2008 4. Users have presence sessions, i.e., times in which they advertise their network availability to their contacts. 5. Not all registered users have an active presence session at any one time. In typical usage patterns, the number of online users is some percentage of the number of registered users. Within an organization, the precentage might be as high as 50%. At a consumer-oriented provider of presence-enabled communication services, the percentage might be 10% or less. 6. A presence session is initiated when a user's client sends an initial presence notification to its server, expressing network availability. 7. Upon receiving the initial presence notification, a user's server broadcasts that presence notification to all of the user's contacts and also sends a presence probe to all of the user's contacts. 8. Upon receiving a presence probe, a contact's server checks the contact's authorization policies and, if the user is authorized and the contact is online, sends a presence notification to the probing user. 9. During the life of the user's presence session, any subsequent changes to the user's presence information are broadcasted via presence notifications sent by the user's server to the user's online contacts. 10. Presence notifications are not acknowledged by the recipient. 11. Presence notifications are generated by a user's client only to advertise on-off network availability, availability sub-states (e.g., "away" or "do not disturb") as defined in [XMPP-IM], or information related to the communication capabilities of the user's XMPP client (see [CAPS]). Any other real-time notifications (a user's activity or mood, the music to which a user is listening, the games a user is playing, etc.) are not sent via the XMPP presence "transport" but instead are published using non-presence technologies such as the XMPP Publish- Subscribe extension [PUBSUB], in particular the Personal Eventing profile thereof (see [PEP]). 12. A presence session is terminated when a user's client sends an unavailable presence notification to its server or the server detects that the client is no longer online; in either case the user's server broadcasts an unavailable presence notification to all of the user's online contacts. 3. Protocol Flows A user (in these examples romeo@example.net) initiates a presence session by sending an initial presence notification. To provide a realistic example, this presence notification includes entity capabilities information as defined in [CAPS]. Saint-Andre Expires July 19, 2008 [Page 4] Internet-Draft XMPP Presence Analysis January 2008 User sends initial presence notification (200 bytes): 5 Upon receiving the initial presence notification, the user's server sends an XMPP presence stanza of type "probe" to the contact (in these examples juliet@example.com). User's server sends presence probe to contact (82 bytes): If the contact's server determines that the user is authorized to see the contact's presence, the contact's server returns the contact's current presence state to the user. Here again the presence notification includes entity capabilities information. Contact's server sends presence notification to user (311 bytes): away be right back 0 If the contact subsequently changes her presence, the contact's server sends an updated presence notification to the user. Saint-Andre Expires July 19, 2008 [Page 5] Internet-Draft XMPP Presence Analysis January 2008 Contact's server sends updated presence to user (301 bytes): xa bbiab 0 A presence session can include any number of subsequent presence changes. When the user goes offline, the user's server sends a presence stanza of type "unavailable" to the contact. User's server sends unavailable presence to contact (96 bytes): Naturally, similar protocol flows are generated by the contact during her presence session. 4. Analysis Traffic calculations are based on the following inputs and formulae. 4.1. Constants o (C1) Presence session lifetime in hours -- assumed to be 8 hours in all scenarios. o (C2) Presence state changes per hour -- assumed to be 3 times per hour in most scenarios. o (C3) Total federated contacts per user -- varies according to the scenario. o (C4) Number of online contacts -- varies according to the scenario. Saint-Andre Expires July 19, 2008 [Page 6] Internet-Draft XMPP Presence Analysis January 2008 o (C5) Number of federated users -- varies according to the scenario. o (C6) Number of online users -- varies according to the scenario. o (C7) Size of presence probe sent by user's server upon receipt of initial outbound presence notification -- 100 bytes. o (C8) Size of presence notifications sent by users and contacts -- 300 bytes. o (C9) Size of unavailable presence notifications -- 100 bytes. 4.2. Initial Stanzas When a user initiates a presence session, the following presence stanzas are exchanged. o (I1) Number of outbound presence notifications = 1 per federated contact (C3). o (I2) Size of outbound presence notifications = (C3*C8). o (I3) Number of presence probes = one per federated contact (C3). o (I4) Size of presence probes = (C3*C7). o (I5) Number of inbound presence notifications = 1 per online contact (C4). o (I6) Size of inbound presence notifications = (C4*C8). o (I7) Total number of initial stanzas = (I1+I3+I5). o (I8) Total size of initial stanzas = (I2+I4+I6). 4.3. State-Change Stanzas During the life of a user's presence session, the following presence stanzas are exchanged as a result of changes in the user's availability. o (S1) Number of presence state changes per user = (C1*C2)-2). o (S2) Number of outbound presence notifications = (S1*C4). o (S3) Size of presence notifications = (S2*C8). 4.4. Termination Stanzas When a user terminates a presence session, the following presence stanzas are exchanged. o (T1) Number of unavailable presence notifications = 1 per online contact (C4). o (T2) Size of unavailable presence notifications = (C4*C9). 4.5. Bottom Line The foregoing assumptions result in the following number and size of stanzas exchanged per user per presence session. Saint-Andre Expires July 19, 2008 [Page 7] Internet-Draft XMPP Presence Analysis January 2008 o (B1) Number of stanzas exchanged per presence session = (I7+S2+T1). o (B2) Size of stanzas exchanged per presence session = (I8+S3+T2). Therefore the total number and size of stanzas exchanged between two federated domains is as follows (i.e., summed for all online users). o (B3) Total number of stanzas exchanged = (B1*C6). o (B4) Total size of stanzas exchanged = (B2*C6). o (B5) Total stanzas exchanged per second = (B3/(C1*3600)). o (B6) Total bytes exchanged per second = (B4/(C1*3600)). 5. Scenarios 5.1. Enterprises in Different Industries This scenario assumes two domains, each with 20,000 users, where each user has 4 contacts in the other domain, each user changes presence 3 times per hour during an 8-hour presence session, and 50% of the users are online at any one time. Such a scenario might be applicable to presence federation between two medium-sized enterprises in different industries. Saint-Andre Expires July 19, 2008 [Page 8] Internet-Draft XMPP Presence Analysis January 2008 CONSTANTS (C1) Presence session lifetime (hours) ....................... 8 (C2) Presence state changes per hour ......................... 3 (C3) Total federated contacts per user ....................... 4 (C4) Number of online contacts per user ...................... 2 (C5) Number of federated users .......................... 40,000 (C6) Number of online users ............................. 20,000 (C7) Size of presence probes ............................... 100 (C8) Size of presence notifications ........................ 300 (C9) Size of unavailable presence notification ............. 100 INITIAL STANZAS (I1) Number of outbound presence notifications ............... 4 (I2) Size of outbound presence notifications ............. 1,200 (I3) Number of presence probes per user ...................... 4 (I4) Size of presence probes per user ...................... 400 (I5) Number of inbound presence notifications ................ 2 (I6) Size of inbound presence notifications ................ 600 (I7) Total number of initial stanzas ........................ 10 (I8) Total size of initial stanzas ....................... 2,200 STATE CHANGE STANZAS (S1) Number of state changes per user ....................... 22 (S2) Number of outbound presence notifications .............. 44 (S3) Size of outbound presence notifications ............ 13,200 TERMINATION MESSAGES (T1) Number of unavailable presence notifications ............ 2 (T2) Size of unavailable presence notifications ............ 200 BOTTOM LINE (B1) Number of stanzas per presence session ................. 56 (B2) Size of stanzas per presence session ............... 15,600 (B3) Total number of stanzas exchanged ............... 1,120,000 (B4) Total size of stanzas exchanged ............... 312,000,000 (B5) Total stanzas exchanged per second ..................... 39 (B6) Total bytes exchanged per second ................... 10,833 With compression as described under Section 6, the bytes per second might be as low as 1,083. 5.2. Enterprises in the Same Industry This scenario assumes two domains, each with 20,000 users, where each user has 20 contacts in the other domain, each user changes presence 3 times per hour during an 8-hour presence session, and 50% of the users are online at any one time. Such a scenario might be applicable to presence federation between two medium-sized Saint-Andre Expires July 19, 2008 [Page 9] Internet-Draft XMPP Presence Analysis January 2008 enterprises in the same industry. CONSTANTS (C1) Presence session lifetime (hours) ....................... 8 (C2) Presence state changes per hour ......................... 3 (C3) Total federated contacts per user ...................... 20 (C4) Number of online contacts per user ..................... 10 (C5) Number of federated users .......................... 40,000 (C6) Number of online users ............................. 20,000 (C7) Size of presence probes ............................... 100 (C8) Size of presence notifications ........................ 300 (C9) Size of unavailable presence notification ............. 100 INITIAL STANZAS (I1) Number of outbound presence notifications .............. 20 (I2) Size of outbound presence notifications ............. 6,000 (I3) Number of presence probes per user ..................... 20 (I4) Size of presence probes per user .................... 2,000 (I5) Number of inbound presence notifications ............... 10 (I6) Size of inbound presence notifications .............. 3,000 (I7) Total number of initial stanzas ........................ 50 (I8) Total size of initial stanzas ...................... 11,000 STATE CHANGE STANZAS (S1) Number of state changes per user ....................... 22 (S2) Number of outbound presence notifications ............. 220 (S3) Size of outbound presence notifications ............ 66,000 TERMINATION MESSAGES (T1) Number of unavailable presence notifications ........... 10 (T2) Size of unavailable presence notifications .......... 1,000 BOTTOM LINE (B1) Number of stanzas per presence session ................ 280 (B2) Size of stanzas per presence session ............... 78,000 (B3) Total number of stanzas exchanged ............... 5,600,000 (B4) Total size of stanzas exchanged ............. 1,560,000,000 (B5) Total stanzas exchanged per second .................... 194 (B6) Total bytes exchanged per second ................... 54,167 With compression as described under Section 6, the bytes per second might be as low as 5,417. 5.3. Medium-Sized Service Providers This scenario assumes two domains, each with 60,000 users, where each user has 10 contacts in the other domain, each user changes presence 3 times per hour during an 8-hour presence session, and 10% of the Saint-Andre Expires July 19, 2008 [Page 10] Internet-Draft XMPP Presence Analysis January 2008 users are online at any one time. Such a scenario might be applicable to presence federation between two medium-sized service providers. CONSTANTS (C1) Presence session lifetime (hours) ....................... 8 (C2) Presence state changes per hour ......................... 3 (C3) Total federated contacts per user ...................... 10 (C4) Number of online contacts per user ...................... 1 (C5) Number of federated users ......................... 120,000 (C6) Number of online users ............................. 60,000 (C7) Size of presence probes ............................... 100 (C8) Size of presence notifications ........................ 300 (C9) Size of unavailable presence notification ............. 100 INITIAL STANZAS (I1) Number of outbound presence notifications .............. 10 (I2) Size of outbound presence notifications ............. 3,000 (I3) Number of presence probes per user ..................... 10 (I4) Size of presence probes per user .................... 1,000 (I5) Number of inbound presence notifications ................ 1 (I6) Size of inbound presence notifications ................ 300 (I7) Total number of initial stanzas ........................ 21 (I8) Total size of initial stanzas ....................... 4,300 STATE CHANGE STANZAS (S1) Number of state changes per user ....................... 22 (S2) Number of outbound presence notifications .............. 22 (S3) Size of outbound presence notifications ............. 6,600 TERMINATION MESSAGES (T1) Number of unavailable presence notifications ............ 1 (T2) Size of unavailable presence notifications ............ 100 BOTTOM LINE (B1) Number of stanzas per presence session ................. 44 (B2) Size of stanzas per presence session ............... 11,000 (B3) Total number of stanzas exchanged ............... 2,640,000 (B4) Total size of stanzas exchanged ............... 660,000,000 (B5) Total stanzas exchanged per second ..................... 92 (B6) Total bytes exchanged per second ................... 22,917 With compression as described under Section 6, the bytes per second might be as low as 2,292. Saint-Andre Expires July 19, 2008 [Page 11] Internet-Draft XMPP Presence Analysis January 2008 5.4. Large Service Providers This scenario assumes two domains, each with 300,000 users, where each user has 20 contacts in the other domain, each user changes presence 3 times per hour during an 8-hour presence session, and 10% of the users are online at any one time. Such a scenario might be applicable to presence federation between two large service providers. CONSTANTS (C1) Presence session lifetime (hours) ....................... 8 (C2) Presence state changes per hour ......................... 3 (C3) Total federated contacts per user ...................... 20 (C4) Number of online contacts per user ...................... 2 (C5) Number of federated users ......................... 600,000 (C6) Number of online users ............................ 300,000 (C7) Size of presence probes ............................... 100 (C8) Size of presence notifications ........................ 300 (C9) Size of unavailable presence notification ............. 100 INITIAL STANZAS (I1) Number of outbound presence notifications .............. 20 (I2) Size of outbound presence notifications ............. 6,000 (I3) Number of presence probes per user ..................... 20 (I4) Size of presence probes per user .................... 2,000 (I5) Number of inbound presence notifications ................ 2 (I6) Size of inbound presence notifications ................ 600 (I7) Total number of initial stanzas ........................ 42 (I8) Total size of initial stanzas ....................... 8,600 STATE CHANGE STANZAS (S1) Number of state changes per user ....................... 22 (S2) Number of outbound presence notifications .............. 44 (S3) Size of outbound presence notifications ............ 13,200 TERMINATION MESSAGES (T1) Number of unavailable presence notifications ............ 2 (T2) Size of unavailable presence notifications ............ 200 BOTTOM LINE (B1) Number of stanzas per presence session ................. 88 (B2) Size of stanzas per presence session ............... 22,000 (B3) Total number of stanzas exchanged .............. 26,400,000 (B4) Total size of stanzas exchanged ............. 6,600,000,000 (B5) Total stanzas exchanged per second .................... 917 (B6) Total bytes exchanged per second .................. 229,167 With compression as described under Section 6, the bytes per second Saint-Andre Expires July 19, 2008 [Page 12] Internet-Draft XMPP Presence Analysis January 2008 might be as low as 22,917. 5.5. Very Large Service Providers This scenario assumes two domains, each with 10,000,000 users, where each user has 100 contacts in the other domain, each user changes presence 6 times per hour during an 8-hour presence session, and 20% of the users are online at any one time. Such a scenario might be applicable to presence federation between two very large service providers. CONSTANTS (C1) Presence session lifetime (hours) ....................... 8 (C2) Presence state changes per hour ......................... 6 (C3) Total federated contacts per user ..................... 100 (C4) Number of online contacts per user ..................... 20 (C5) Number of federated users ...................... 20,000,000 (C6) Number of online users .......................... 4,000,000 (C7) Size of presence probes ............................... 100 (C8) Size of presence notifications ........................ 300 (C9) Size of unavailable presence notification ............. 100 INITIAL STANZAS (I1) Number of outbound presence notifications ............. 100 (I2) Size of outbound presence notifications ............ 30,000 (I3) Number of presence probes per user .................... 100 (I4) Size of presence probes per user ................... 10,000 (I5) Number of inbound presence notifications ............... 20 (I6) Size of inbound presence notifications .............. 6,000 (I7) Total number of initial stanzas ....................... 220 (I8) Total size of initial stanzas ...................... 46,000 STATE CHANGE STANZAS (S1) Number of state changes per user ....................... 46 (S2) Number of outbound presence notifications ............. 920 (S3) Size of outbound presence notifications ........... 276,000 TERMINATION MESSAGES (T1) Number of unavailable presence notifications ........... 20 (T2) Size of unavailable presence notifications .......... 2,000 BOTTOM LINE (B1) Number of stanzas per presence session .............. 1,160 (B2) Size of stanzas per presence session .............. 324,000 (B3) Total number of stanzas exchanged ........... 4,640,000,000 (B4) Total size of stanzas exchanged ......... 1,296,000,000,000 (B5) Total stanzas exchanged per second ................ 161,111 (B6) Total bytes exchanged per second ............... 45,000,000 Saint-Andre Expires July 19, 2008 [Page 13] Internet-Draft XMPP Presence Analysis January 2008 With compression as described under Section 6, the bytes per second might be as low as 4,500,000. 6. Optimizations This document does not focus on optimizations that can be applied to XMPP traffic. The main such optimization is compression of XML streams. There are several reasons why stream compression can yield significant reductions in the network impact of XMPP traffic, especially in the case of interdomain federation: 1. XMPP uses long-lived TCP connections in which (typically) a single XML parser instance is used to parse the incoming and outgoing XML stanzas. 2. The fact that XMPP stanzas are XML means that the same strings are repeatedly communicated over the stream (e.g., the string "