The Common Log Format (CLF) for the Session Initiation Protocol (SIP)
DOCNAME

Status of this Memo

This Internet-Draft is submitted to IETF in full conformance with the provisions of BCP 78 and 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 April 23, 2010.

Copyright Notice

Copyright (c) 2009 IETF Trust and the persons identified as the document authors. All rights reserved.

This document is subject to BCP 78 and the IETF Trust's Legal Provisions Relating to IETF Documents in effect on the date of publication of this document (http://trustee.ietf.org/license-info). Please review these documents carefully, as they describe your rights and restrictions with respect to this document.

Abstract

Well-known web servers such as Apache and web proxies like Squid support event logging using a common log format. The logs produced using these de-facto standard formats are invaluable to system administrators for trouble-shooting a server and tool writers to craft tools that mine the log files and produce reports and trends. Furthermore, these log files can also be used to train anomaly detection systems and feed events into a security event management system. The Session Initiation Protocol does not have a common log format, and as a result, each server supports a distinct log format that makes it unnecessarily complex to produce tools to do trend analysis and security detection. We propose a common log file format for SIP servers that can be used uniformly for proxies, registrars, redirect servers as well as back-to-back user agents.

Table of Contents

1.  Terminology
2.  Introduction
3.  Motivation and use cases
4.  What SIP CLF is and what it is not
5.  Challenges in establishing a SIP CLF
6.  SIP CLF fields
7.  Relationship to other protocols
8.  Security Considerations
9.  Operational guidance
10.  IANA Considerations
11.  Acknowledgments
12.  References
    12.1.  Normative References
    12.2.  Informative References
§  Authors' Addresses

1.  Terminology

The key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "MAY", and "OPTIONAL" in this document are to be interpreted as described in RFC 2119 (Bradner, S., “Key words for use in RFCs to Indicate Requirement Levels,” March 1997.) [1].

RFC 3261 (Rosenberg, J., Schulzrinne, H., Camarillo, G., Johnston, A., Peterson, J., Sparks, R., Handley, M., and E. Schooler, “SIP: Session Initiation Protocol,” June 2002.) [2] defines additional terms used in this document that are specific to the SIP domain such as "proxy"; "registrar"; "redirect server"; "user agent server" or "UAS"; "user agent client" or "UAC"; "back-to-back user agent" or "B2BUA"; "dialog"; "transaction"; "server transaction".

This document uses the term "SIP Server" that is defined to include the following SIP entities: user agent server, registrar, redirect server, a SIP proxy in the role of user agent server, and a B2BUA in the role of a user agent server.

2.  Introduction

Servers executing on Internet hosts produce log records as part of their normal operations. A log record is, in essence, a summary of an application layer protocol data unit (PDU), that captures in precise terms an event that was processed by the server. These log records serve many purposes, including analysis and troubleshooting.

Well-known web servers such as Apache and Squid support event logging using a Common Log Format (CLF), the common structure for logging requests and responses serviced by the web server. It can be argued that a good part of the success of Apache has been its CLF because it allowed third parties to produce tools that analyzed the data and generated traffic reports and trends. The Apache CLF has been so successful that not only did it become the de-facto standard in producing logging data for web servers, but also many commercial web servers can be configured to produce logs in this format. An example of Apache CLF is depicted next:

          %h      %l     %u       %t   \"%r\"   %s    %b
     remotehost rfc931 authuser [date] request status bytes

remotehost:

Remote hostname (or IP number if DNS hostname is not available, or if DNSLookup is Off.

rfc931:

The remote logname of the user.

authuser:

The username by which the user has authenticated himself.

[date]:

Date and time of the request.

request:

The request line exactly as it came from the client.

status:

The HTTP status code returned to the client.

bytes:

The content-length of the document transferred.

The Session Initiation Protocol (Rosenberg, J., Schulzrinne, H., Camarillo, G., Johnston, A., Peterson, J., Sparks, R., Handley, M., and E. Schooler, “SIP: Session Initiation Protocol,” June 2002.) [2](SIP) is an Internet multimedia session signaling protocol that is increasingly used for other services besides session establishment. SIP does not currently have a CLF format and this memorandum serves to provide the rationale to establish a SIP CLF and identifies the required minimal information that must appear in any record.

3.  Motivation and use cases

As SIP becomes pervasive in multiple business domains and ubiquitous in academic and research environments, it is beneficial to establish a CLF for the following reasons:

Common reference for interpreting events:

In a laboratory environment or an enterprise service offering there will typically be SIP servers from multiple vendors participating in routing requests. Absent a CLF format, each server will produce output records in a native format making it hard to establish commonality for tools that operate on the log file.

Writing common tools:

A CLF format allows independent tool providers to craft tools and applications that interpret the CLF data to produce insightful trend analysis and detailed traffic reports. The format should be such that it retains the ability to be read by humans and processed using traditional Unix text processing tools.

Session correlation across diverse processing elements:

In operational SIP networks, a request will typically be processed by more than one SIP server. A SIP CLF will allow the network operator to trace the progression of the request (or a set of requests) as they traverse through the different servers to establish a concise diagnostic trail of a SIP session.

Note that tracing the request through a set of servers is considerably less challenging if all the servers belong to the same administrative domain.

Message correlation across transactions:

A SIP CLF can enable a quick lookup of all messages that comprise a transaction (e.g., "Find all messages corresponding to server transaction X, including all forked branches.")

Message correlation across dialogs:

A SIP CLF can correlate transactions that comprise a dialog (e.g., "Find all messages for dialog created by Call-ID C, From tag F and To tag T.")

Trend analysis:

A SIP CLF allows an administrator to collect data and spot patterns or trends in the information (e.g., "What is the domain where the most sessions are routed to between 9:00 AM and 12:00 PM?")

Train anomaly detection systems:

A SIP CLF will allow for the training of anomaly detection systems that once trained can monitor the CLF file to trigger an alarm on the subsequent deviations from accepted patterns in the data set. Currently, anomaly detection systems monitor the network and parse raw packets that comprise a SIP message -- a process that is unsuitable for anomaly detection systems [3] (Rieck, K., Wahl, S., Laskov, P., Domschitz, P., and K-R. Muller, “A Self-learning System for Detection of Anomalous SIP Messages,” 2008.). With all the necessary event data at their disposal, network operations managers and information technology operation managers are in a much better position to correlate, aggregate, and prioritize log data to maintain situational awareness.

Testing:

A SIP CLF allows for automatic testing of SIP equipment by writing tools that can parse a SIP CLF file to ensure behavior of a device under test.

Troubleshooting:

A SIP CLF can enable cursory trouble shooting of a SIP server (e.g., "How long did it take to generate a final response for the INVITE associated with Call-ID X?")

Offline analysis:

A SIP CLF allows for offline analysis of the data gathered. Once a SIP CLF file has been generated, it can be transported (subject to the security considerations in Section 8 (Security Considerations)) to a host with appropriate computing resources to perform subsequent analysis.

Real-time monitoring:

A SIP CLF allows administrators to visually notice the events occurring at a SIP server in real-time providing accurate situational awareness.

4.  What SIP CLF is and what it is not

With the success of SIP in traditional telephony domains, it is tempting to view the SIP CLF as a replacement for call logs and Call Detail Records (CDRs). However, this is expressly not our intent. The charging system of a telephone exchange produces a CDR. Insofar as a SIP entity is acting as a telephone exchange, it can continue producing CDR irrespective of whether it also produces a SIP CLF.

The SIP CLF is not a billing tool. It is not expected that enterprises will bill customers based on SIP CLF. The SIP CLF records events at the signaling layer only and does not attempt to correlate the veracity of these events with the media layer. Thus, SIP CLF must not be used to trigger customer billing.

The SIP CLF is not a quality of service (QoS) measurement tool. First, if QoS is defined as measuring the mean opinion score (MOS) of the received media, then SIP CLF does not aid in this task since it does not summarize events at the media layer. Second, insofar as QoS is defined as the time it takes a SIP server to issue a response to a request, some processing delay metric may be inferred from the SIP CLF. However, note that such inference is appropriate mostly from a proxy or a B2BUA; a UAS is limited by the amount of time it takes the physical user associated with the UAS to generate a (final) response.

The SIP CLF is a standardized manner of producing a text file. This format is used by SIP Servers, proxies, and B2BUAs. The SIP CLF is simply an easily digestible log of currently occurring events and past transactions. It contains enough information to allow humans and automata to derive relationships between discrete transactions handled at a SIP entity. For example, a SIP administrator should be able to issue a concise command to discover relationships between transactions or to search a certain dialog or transaction.

Note: The exact form of the "concise command" is left unspecified until the working group agrees to one or more formats for encoding the fields.

The SIP CLF is amenable to quick parsing (i.e., well-delimited fields) and it is platform and operating system neutral.

The SIP CLF is amenable to easy parsing and lends itself well to creating other innovative tools.

5.  Challenges in establishing a SIP CLF

Establishing a CLF for SIP is a challenging task. The behavior of a SIP entity is more complex when compared to the equivalent HTTP entity.

Base protocol services such as parallel or serial forking elicit multiple final responses. Ensuing delays between sending a request and receiving a final response all add complexity when considering what fields should comprise a CLF and in what manner. Furthermore, unlike HTTP, SIP groups multiple discrete transactions into a dialog, and these transactions may arrive at a varying inter-arrival rate at a proxy. For example, the BYE transaction usually arrives much after the corresponding INVITE transaction was received, serviced and expunged from the transaction list. Nonetheless, it is advantageous to relate these transactions such that automata or a human monitoring the log file can construct a set consisting of related transactions.

ACK requests in SIP need careful consideration as well. In SIP, an ACK is a special method that is associated with an INVITE only. It does not require a response, and furthermore, if it is acknowledging a non-2xx response, then the ACK is considered part of the original INVITE transaction. If it is acknowledging a 2xx-class response, then the ACK is a separate transaction consisting of a request only (i.e., there is not a response for an ACK request.) CANCEL is another method that is tied to an INVITE transaction, but unlike ACK, the CANCEL request elicits a final response.

While most requests elicit a response immediately, the INVITE request in SIP can pend at a proxy as it forks branches downstream or at a user agent server while it alerts the user. RFC 3261 (Rosenberg, J., Schulzrinne, H., Camarillo, G., Johnston, A., Peterson, J., Sparks, R., Handley, M., and E. Schooler, “SIP: Session Initiation Protocol,” June 2002.) [2] instructs the server transaction to send a 1xx-class provisional response if a final response is delayed for more than 200 ms. A SIP CLF log file needs to include such provisional responses because they help train automata associated with anomaly detection systems and provide some positive feedback for a human observer monitoring the log file.

Finally, beyond supporting native SIP actors such as proxies, registrars, redirect servers, and user agent servers (UAS), it is beneficial to derive a CLF format that supports back-to-back user agent (B2BUA) behavior, which may vary considerably depending on the specific nature of the B2BUA.

6.  SIP CLF fields

The inspiration for the SIP CLF is the Apache CLF. However, the state machinery for a HTTP transaction is much simpler than that of the SIP transaction (as evidenced in Section 5 (Challenges in establishing a SIP CLF)). The SIP CLF needs to do considerably more.

Accordingly, the following SIP CLF fields are defined as mininal information that must appear in any SIP CLF record:

date:

Date and time of the request or response represented as the number of seconds and milliseconds since the Unix epoch.

remotehost:

The DNS name or IP address of the upstream client.

authuser:

The user name by which the user has been authenticated. If the user name is unknown or when a request is challenged, the value in this field must be "-"

method:

The upper-case name of the SIP method.

request-uri:

The Request-URI, including any URI parameters.

from:

The From URI, including the tag. Whilst one may question the value of the From URI in light of RFC4744 (Peterson, J. and C. Jennings, “Enhancements for Authenticated Identity Management in the Session Initiation Protocol (SIP),” August 2006.) [4], the From URI, nonetheless, imparts some information. For one, the From tag is important and, in the case of a REGISTER request, the From URI can provide information on whether this was a third-party registration or a first-party one.

to:

The To URI, including tag.

callid:

The Call-ID.

status:

The SIP response status code returned upstream.

contactlist:

Contact URIs in the response, if any. A "-" field value may be used if there aren't any Contact URIs.

server-txn:

Server transaction identification code - the transaction identifier associated with the server transaction. Implementations can reuse the server transaction identifier (the topmost branch-id of the incoming request, with or without the magic cookie), or they could generate a unique identification string for a server transaction (this identifier needs to be locally unique to the server only.) This identifier is used to correlate ACKs and CANCELs to an INVITE transaction; it is also used to aid in forking as explained later in this section.

client-txn:

Client transaction identification code - this field is used to associate client transactions with a server transaction for forking proxies or B2BUAs. Upon forking, implementations can reuse the value they inserted into the topmost Via header's branch parameter, or they can generate a unique identification string for the client transaction. A more detailed explanation of why it is needed is provided next.

SIP Proxies may fork, creating several client transactions that correlate to a single server transaction. Responses arriving on these client transactions, or new requests (CANCEL, ACK) sent on the client transaction need log file entries that correlate with a server transaction. Similarly, a B2BUA may create one or more client transactions in response to an incoming request. These transactions will require correlation as well.

To best demonstrate the correlation directives "server-txn" and "client-txn", some examples are necessary. In order to do so, it helps to use a canonical representation for the SIP CLF. The most expedient way to do so is to use an ASCII representation for illustration purposes only, but to be safe, the working group should okay this since a specific SIP CLF format has not been defined yet. To get a gist of how these correlation directives help, please see Section 6 of a predecessor (Gurbani, V., Burger, E., Anjali, T., Abdelnur, H., and O. Festor, “The Common Log File (CLF) format for the Session Initiation Protocol (SIP),” March 2009.) [5] to this draft.

Finally, the SIP CLF should be extensible such that future SIP methods, headers and bodies can be represented as well.

Note: Not sure what else to say here since the specific format used to encode the CLF has some bearing on extensibility. A type-length-value (TLV) format or a PCAP format has advantages over an ASCII format. All three formats -- the ASCII, binary TLV format, and the PCAP format -- are available in Gurbani et al. (Gurbani, V., Burger, E., Anjali, T., Abdelnur, H., and O. Festor, “The Common Log File (CLF) format for the Session Initiation Protocol (SIP),” March 2009.) [5], Roach (Roach, A., “Binary Syntax for SIP Common Log Format,” May 2009.) [6], and Kaplan (Kaplan, H., “PCAP-compatible Binary Syntax for SIP Common Log File Format,” June 2009.) [7], respectively.

7.  Relationship to other protocols

During the 75th IETF, there was a substantial amount of discussion on the interplay of SIP CLF and other existing protocols. The IPFIX (Claise, B., “Specification of the IP Flow Information Export (IPFIX) Protocol for the Exchange of IP Traffic Flow Information,” January 2008.) [8] protocol is an encompassing solution to exporting information elements to a central manager over a multitude of transports. However, specifying the mechanics of exchanging and transporting SIP CLF records is out of scope of the current SIPCLF charter.

The syslog (Gerhards, R., “The Syslog Protocol,” March 2009.) [9] protocol allows the conveyance of event notification messages. Because of the frequency and number of messages that will need to be logged, it is not desirable to send every SIP CLF event to the syslog daemon. Instead, a judicious use of syslog could be that only certain events -- those that are pertinent from a network situational awareness perspective -- are sent to the syslog daemon preferably over a confidentiality and integrity protected channel such as TLS or DTLS.

Another protocol, IDMEF (Debar, H., Curry, D., and B. Feinstein, “The Intrusion Detection Message Exchange Format (IDMEF),” March 2007.) [10], defines data formats and exchange procedures for sharing information of interest to intrusion detection and response systems and management systems that may need to react with them. As was the case with syslog, instead of transporting every message to a detection system, it seems appropriate to digest certain high-value records from standard SIP CLF and turn only these into an IDMEF-expressible syntax.

8.  Security Considerations

A log file by its nature reveals the both the state of the entity producing it and the nature of the information being logged. To the extent that this state should not be publicly accessible and that the information is to be considered private, appropriate file and directory permissions attached to the log file should be used. In the worst case, public access to the SIP log file provides the same information that an adversary can gain using network sniffing tools (assuming that the SIP traffic is in clear text.) If all SIP traffic on a network segment is encrypted, then special attention must be directed to the file and directory permissions associated with the log file to preserve privacy such that only a privileged user can access the contents of the log file.

Transporting SIP CLF files across the network pose special challenges as well. While transporting SIP CLF files is out of scope in the current SIPCLF charter, it seems worth drawing attention to the fact that if the file is transported using unencrypted FTP or email, intermediaries and adversaries may have access to the raw SIP CLF records. Accordingly, if the SIP CLF file is to be moved from the generating host, secure FTP or secure email must be used instead.

The SIP CLF represents the minimum fields that lend themselves to trend analysis and serve as information that may be deemed useful. Other formats can be defined that include more headers (and the body) from Section 6 (SIP CLF fields). However, where to draw a judicial line regarding the inclusion of non-mandatory headers can be challenging. Clearly, the more information a SIP server logs, the longer time the logging process will take, the more disk space the log entry will consume, and the more potentially sensitive information could be breached. Therefore, adequate tradeoffs should be taken in account when logging more fields than the ones recommended in Section 6 (SIP CLF fields).

Implementers need to pay particular attention to buffer handling when reading or writing log files. SIP CLF entries can be unbounded in length. It would be reasonable for a full dump of a SIP message to be thousands of octets long. This is of particular importance to CLF log parsers, as a SIP CLF log writers may add one or more extension fields to the message to be logged.

9.  Operational guidance

Operational guidance for log file management with respect to size and rollover policies still needs to be documented.

10.  IANA Considerations

This document does not require any considerations from IANA.

11.  Acknowledgments

Members of the sipping, dispatch, ipfix and syslog working groups provided invaluable input to the formulation of the draft. These include Benoit Claise, Spencer Dawkins, David Harrington, Christer Holmberg, Hadriel Kaplan, Atsushi Kobayashi, Jiri Kuthan, Scott Lawrence, Simon Perreault, Adam Roach, Dan Romascanu, Robert Sparks, Brian Trammell, Dale Worley, Theo Zourzouvillys and others that we have undoubtedly, but inadvertently, missed.

Adam Roach proposed and documented the binary TLV-format and Hadriel Kaplan proposed and documented the PCAP-based format as alternative SIP CLF representations.

12.  References

12.1. Normative References

12.2. Informative References

Authors' Addresses