Internet Engineering Task Force (IETF) C. Davids Request for Comments: 7502 Illinois Institute of Technology Category: Informational V. Gurbani ISSN: 2070-1721 Bell Laboratories, Alcatel-Lucent S. Poretsky Allot Communications April 2015
Methodology for Benchmarking Session Initiation Protocol (SIP) Devices: Basic Session Setup and Registration
Abstract
This document provides a methodology for benchmarking the Session Initiation Protocol (SIP) performance of devices. Terminology related to benchmarking SIP devices is described in the companion terminology document (RFC 7501). Using these two documents, benchmarks can be obtained and compared for different types of devices such as SIP Proxy Servers, Registrars, and Session Border Controllers. The term "performance" in this context means the capacity of the Device Under Test (DUT) to process SIP messages. Media streams are used only to study how they impact the signaling behavior. The intent of the two documents is to provide a normalized set of tests that will enable an objective comparison of the capacity of SIP devices. Test setup parameters and a methodology are necessary because SIP allows a wide range of configurations and operational conditions that can influence performance benchmark measurements.
Status of This Memo
This document is not an Internet Standards Track specification; it is published for informational purposes.
This document is a product of the Internet Engineering Task Force (IETF). It represents the consensus of the IETF community. It has received public review and has been approved for publication by the Internet Engineering Steering Group (IESG). Not all documents approved by the IESG are a candidate for any level of Internet Standard; see Section 2 of RFC 5741.
Information about the current status of this document, any errata, and how to provide feedback on it may be obtained at http://www.rfc-editor.org/info/rfc7502.
Davids, et al. Informational [Page 1]
RFC 7502 SIP Benchmarking Methodology April 2015
Copyright Notice
Copyright (c) 2015 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 (http://trustee.ietf.org/license-info) in effect on the date of publication of this document. Please review these documents carefully, as they describe your rights and restrictions with respect to this document. Code Components extracted from this document must include Simplified BSD License text as described in Section 4.e of the Trust Legal Provisions and are provided without warranty as described in the Simplified BSD License.
This document describes the methodology for benchmarking Session Initiation Protocol (SIP) performance as described in the Terminology document [RFC7501]. The methodology and terminology are to be used for benchmarking signaling plane performance with varying signaling and media load. Media streams, when used, are used only to study how they impact the signaling behavior. This document concentrates on benchmarking SIP session setup and SIP registrations only.
The Device Under Test (DUT) is a network intermediary that is RFC 3261 [RFC3261] capable and that plays the role of a registrar, redirect server, stateful proxy, a Session Border Controller (SBC) or a B2BUA. This document does not require the intermediary to assume the role of a stateless proxy. Benchmarks can be obtained and compared for different types of devices such as a SIP proxy server, Session Border Controllers (SBC), SIP registrars and a SIP proxy server paired with a media relay.
The test cases provide metrics for benchmarking the maximum 'SIP Registration Rate' and maximum 'SIP Session Establishment Rate' that the DUT can sustain over an extended period of time without failures (extended period of time is defined in the algorithm in Section 4.10). Some cases are included to cover encrypted SIP. The test topologies that can be used are described in the Test Setup section. Topologies in which the DUT handles media as well as those in which the DUT does not handle media are both considered. The measurement of the performance characteristics of the media itself is outside the scope of these documents.
Benchmark metrics could possibly be impacted by Associated Media. The selected values for Session Duration and Media Streams per Session enable benchmark metrics to be benchmarked without Associated Media. Session Setup Rate could possibly be impacted by the selected value for Maximum Sessions Attempted. The benchmark for Session Establishment Rate is measured with a fixed value for maximum Session Attempts.
Finally, the overall value of these tests is to serve as a comparison function between multiple SIP implementations. One way to use these tests is to derive benchmarks with SIP devices from Vendor-A, derive a new set of benchmarks with similar SIP devices from Vendor-B and perform a comparison on the results of Vendor-A and Vendor-B. This document does not make any claims on the interpretation of such results.
In this document, the key words "MUST", "MUST NOT", "REQUIRED", "SHALL", "SHALL NOT", "SHOULD", "SHOULD NOT", "RECOMMENDED", "NOT RECOMMENDED", "MAY", and "OPTIONAL" are to be interpreted as described in BCP 14, conforming to [RFC2119] and indicate requirement levels for compliant implementations.
RFC 2119 defines the use of these key words to help make the intent of Standards Track documents as clear as possible. While this document uses these keywords, this document is not a Standards Track document.
Terms specific to SIP [RFC3261] performance benchmarking are defined in [RFC7501].
Test organizations need to be aware that these tests generate large volumes of data and consequently ensure that networking devices like hubs, switches, or routers are able to handle the generated volume.
The test cases enumerated in Sections 6.1 to 6.6 operate on two test topologies: one in which the DUT does not process the media (Figure 1) and the other in which it does process media (Figure 2). In both cases, the tester or Emulated Agent (EA) sends traffic into the DUT and absorbs traffic from the DUT. The diagrams in Figures 1 and 2 represent the logical flow of information and do not dictate a particular physical arrangement of the entities.
Figure 1 depicts a layout in which the DUT is an intermediary between the two interfaces of the EA. If the test case requires the exchange of media, the media does not flow through the DUT but rather passes directly between the two endpoints. Figure 2 shows the DUT as an intermediary between the two interfaces of the EA. If the test case requires the exchange of media, the media flows through the DUT between the endpoints.
During registration or re-registration, the DUT may involve backend network elements and data stores. These network elements and data stores are not shown in Figure 3, but it is understood that they will impact the time required for the DUT to generate a response.
Davids, et al. Informational [Page 6]
RFC 7502 SIP Benchmarking Methodology April 2015
This document explicitly separates a registration test (Section 6.7) from a re-registration test (Section 6.8) because in certain networks, the time to re-register may vary from the time to perform an initial registration due to the backend processing involved. It is expected that the registration tests and the re-registration test will be performed with the same set of backend network elements in order to derive a stable metric.
Test cases may be performed with any transport protocol supported by SIP. This includes, but is not limited to, TCP, UDP, TLS, and websockets. The protocol used for the SIP transport protocol must be reported with benchmarking results.
SIP allows a DUT to use different transports for signaling on either side of the connection to the EAs. Therefore, this document assumes that the same transport is used on both sides of the connection; if this is not the case in any of the tests, the transport on each side of the connection MUST be reported in the test-reporting template.
SIP allows a device to open one connection and send multiple requests over the same connection (responses are normally received over the same connection that the request was sent out on). The protocol also allows a device to open a new connection for each individual request. A connection management strategy will have an impact on the results obtained from the test cases, especially for connection-oriented transports such as TLS. For such transports, the cryptographic handshake must occur every time a connection is opened.
The connection management strategy, i.e., use of one connection to send all requests or closing an existing connection and opening a new connection to send each request, MUST be reported with the benchmarking result.
Some tests require Associated Media to be present for each SIP session. The test topologies to be used when benchmarking DUT performance for Associated Media are shown in Figure 1 and Figure 2.
The test cases specified in this document provide SIP performance independent of the protocol used for the media stream. Any media protocol supported by SIP may be used. This includes, but is not limited to, RTP and SRTP. The protocol used for Associated Media MUST be reported with benchmarking results.
4.6. Number of Associated Media Streams per SIP Session
Benchmarking results may vary with the number of media streams per SIP session. When benchmarking a DUT for voice, a single media stream is used. When benchmarking a DUT for voice and video, two media streams are used. The number of Associated Media Streams MUST be reported with benchmarking results.
The test cases specified in this document provide SIP performance independent of the media stream codec. Any codec supported by the EAs may be used. The codec used for Associated Media MUST be reported with the benchmarking results.
The value of the DUT's performance benchmarks may vary with the duration of SIP sessions. Session Duration MUST be reported with benchmarking results. A Session Duration of zero seconds indicates transmission of a BYE immediately following a successful SIP establishment. Setting this parameter to the value '0' indicates that a BYE will be sent by the EA immediately after the EA receives a 200 OK to the INVITE. Setting this parameter to a time value greater than the duration of the test indicates that a BYE will never be sent. Setting this parameter to a time value greater than the duration of the test indicates that a BYE is never sent.
The value of the DUT's performance benchmarks may vary with the Session Attempt Rate offered by the tester. Session Attempt Rate MUST be reported with the benchmarking results.
Davids, et al. Informational [Page 8]
RFC 7502 SIP Benchmarking Methodology April 2015
The test cases enumerated in Sections 6.1 to 6.6 require that the EA is configured to send the final 2xx-class response as quickly as it can. This document does not require the tester to add any delay between receiving a request and generating a final response.
In order to benchmark the test cases uniformly in Section 6, the algorithm described in this section should be used. A prosaic description of the algorithm and a pseudocode description are provided below, and a simulation written in the R statistical language [Rtool] is provided in Appendix A.
The goal is to find the largest value, R, a SIP Session Attempt Rate, measured in sessions per second (sps), which the DUT can process with zero errors over a defined, extended period. This period is defined as the amount of time needed to attempt N SIP sessions, where N is a parameter of test, at the attempt rate, R. An iterative process is used to find this rate. The algorithm corresponding to this process converges to R.
If the DUT vendor provides a value for R, the tester can use this value. In cases where the DUT vendor does not provide a value for R, or where the tester wants to establish the R of a system using local media characteristics, the algorithm should be run by setting "r", the session attempt rate, equal to a value of the tester's choice. For example, the tester may initialize "r = 100" to start the algorithm and observe the value at convergence. The algorithm dynamically increases and decreases "r" as it converges to the maximum sps value for R. The dynamic increase and decrease rate is controlled by the weights "w" and "d", respectively.
The pseudocode corresponding to the description above follows, and a simulation written in the R statistical language is provided in Appendix A.
Davids, et al. Informational [Page 9]
RFC 7502 SIP Benchmarking Methodology April 2015
; ---- Parameters of test; adjust as needed N := 50000 ; Global maximum; once largest session rate has ; been established, send this many requests before ; calling the test a success m := {...} ; Other attributes that affect testing, such ; as media streams, etc. r := 100 ; Initial session attempt rate (in sessions/sec). ; Adjust as needed (for example, if DUT can handle ; thousands of calls in steady state, set to ; appropriate value in the thousands). w := 0.10 ; Traffic increase weight (0 < w <= 1.0) d := max(0.10, w / 2) ; Traffic decrease weight
; ---- End of parameters of test
proc find_R
R = max_sps(r, m, N) ; Setup r sps, each with m media ; characteristics until N sessions have been attempted. ; Note that if a DUT vendor provides this number, the tester ; can use the number as a Session Attempt Rate, R, instead ; of invoking max_sps()
end proc
; Iterative process to figure out the largest number of ; sps that we can achieve in order to setup n sessions. ; This function converges to R, the Session Attempt Rate. proc max_sps(r, m, n) s := 0 ; session setup rate old_r := 0 ; old session setup rate h := 0 ; Return value, R count := 0
; Note that if w is small (say, 0.10) and r is small ; (say, <= 9), the algorithm will not converge since it ; uses floor() to increment r dynamically. It is best ; to start with the defaults (w = 0.10 and r >= 100).
while (TRUE) { s := send_traffic(r, m, n) ; Send r sps, with m media ; characteristics until n sessions have been attempted. if (s == n) { if (r > old_r) { old_r = r } else { count = count + 1
Davids, et al. Informational [Page 10]
RFC 7502 SIP Benchmarking Methodology April 2015
if (count >= 10) { # We've converged. h := max(r, old_r) break } }
r := floor(r + (w * r)) } else { r := floor(r - (d * r)) d := max(0.10, d / 2) w := max(0.10, w / 2) }
SIP Transport Protocol = ___________________________ (valid values: TCP|UDP|TLS|SCTP|websockets|specify-other) (Specify if same transport used for connections to the DUT and connections from the DUT. If different transports used on each connection, enumerate the transports used.)
Connection management strategy for connection oriented transports DUT receives requests on one connection = _______ (Yes or no. If no, DUT accepts a new connection for every incoming request, sends a response on that connection, and closes the connection.) DUT sends requests on one connection = __________ (Yes or no. If no, DUT initiates a new connection to send out each request, gets a response on that connection, and closes the connection.)
Session Attempt Rate _______________________________ (Session attempts/sec) (The initial value for "r" in benchmarking algorithm in Section 4.10.)
Session Duration = _________________________________ (In seconds)
Davids, et al. Informational [Page 11]
RFC 7502 SIP Benchmarking Methodology April 2015
Total Sessions Attempted = _________________________ (Total sessions to be created over duration of test)
Media Streams per Session = _______________________ (number of streams per session)
Associated Media Protocol = _______________________ (RTP|SRTP|specify-other)
Codec = ____________________________________________ (Codec type as identified by the organization that specifies the codec)
Media Packet Size (audio only) = __________________ (Number of bytes in an audio packet)
Establishment Threshold time = ____________________ (Seconds)
6.1. Baseline Session Establishment Rate of the Testbed
Objective: To benchmark the Session Establishment Rate of the Emulated Agent (EA) with zero failures.
Procedure: 1. Configure the DUT in the test topology shown in Figure 1. 2. Set Media Streams per Session to 0. 3. Execute benchmarking algorithm as defined in Section 4.10 to get the baseline Session Establishment Rate. This rate MUST be recorded using any pertinent parameters as shown in the reporting format of Section 5.1.
Expected Results: This is the scenario to obtain the maximum Session Establishment Rate of the EA and the testbed when no DUT is present. The results of this test might be used to normalize test results performed on different testbeds or simply to better understand the impact of the DUT on the testbed in question.
Objective: To benchmark the Session Establishment Rate of the DUT with no Associated Media and zero failures.
Procedure: 1. Configure a DUT according to the test topology shown in Figure 1 or Figure 2. 2. Set Media Streams per Session to 0. 3. Execute benchmarking algorithm as defined in Section 4.10 to get the Session Establishment Rate. This rate MUST be recorded using any pertinent parameters as shown in the reporting format of Section 5.1.
Expected Results: Find the Session Establishment Rate of the DUT when the EA is not sending media streams.
6.3. Session Establishment Rate with Media Not on DUT
Objective: To benchmark the Session Establishment Rate of the DUT with zero failures when Associated Media is included in the benchmark test but the media is not running through the DUT.
Davids, et al. Informational [Page 13]
RFC 7502 SIP Benchmarking Methodology April 2015
Procedure: 1. Configure a DUT according to the test topology shown in Figure 1. 2. Set Media Streams per Session to 1. 3. Execute benchmarking algorithm as defined in Section 4.10 to get the session establishment rate with media. This rate MUST be recorded using any pertinent parameters as shown in the reporting format of Section 5.1.
Expected Results: Session Establishment Rate results obtained with Associated Media with any number of media streams per SIP session are expected to be identical to the Session Establishment Rate results obtained without media in the case where the DUT is running on a platform separate from the Media Relay.
Objective: To benchmark the Session Establishment Rate of the DUT with zero failures when Associated Media is included in the benchmark test and the media is running through the DUT.
Procedure: 1. Configure a DUT according to the test topology shown in Figure 2. 2. Set Media Streams per Session to 1. 3. Execute benchmarking algorithm as defined in Section 4.10 to get the Session Establishment Rate with media. This rate MUST be recorded using any pertinent parameters as shown in the reporting format of Section 5.1.
Expected Results: Session Establishment Rate results obtained with Associated Media may be lower than those obtained without media in the case where the DUT and the Media Relay are running on the same platform. It may be helpful for the tester to be aware of the reasons for this degradation, although these reasons are not parameters of the test. For example, the degree of performance degradation may be due to what the DUT does with the media (e.g., relaying vs. transcoding), the type of media (audio vs. video vs. data), and the codec used for the media. There may also be cases where there is no performance impact, if the DUT has dedicated media-path hardware.
6.5. Session Establishment Rate with TLS-Encrypted SIP
Objective: To benchmark the Session Establishment Rate of the DUT with zero failures when using TLS-encrypted SIP signaling.
Davids, et al. Informational [Page 14]
RFC 7502 SIP Benchmarking Methodology April 2015
Procedure: 1. If the DUT is being benchmarked as a proxy or B2BUA, then configure the DUT in the test topology shown in Figure 1 or Figure 2. 2. Configure the tester to enable TLS over the transport being used during benchmarking. Note the ciphersuite being used for TLS and record it in Section 5.1. 3. Set Media Streams per Session to 0 (media is not used in this test). 4. Execute benchmarking algorithm as defined in Section 4.10 to get the Session Establishment Rate with TLS encryption.
Expected Results: Session Establishment Rate results obtained with TLS-encrypted SIP may be lower than those obtained with plaintext SIP.
6.6. Session Establishment Rate with IPsec-Encrypted SIP
Objective: To benchmark the Session Establishment Rate of the DUT with zero failures when using IPsec-encrypted SIP signaling.
Procedure: 1. Configure a DUT according to the test topology shown in Figure 1 or Figure 2. 2. Set Media Streams per Session to 0 (media is not used in this test). 3. Configure tester for IPsec. Note the IPsec profile being used for IPsec and record it in Section 5.1. 4. Execute benchmarking algorithm as defined in Section 4.10 to get the Session Establishment Rate with encryption.
Expected Results: Session Establishment Rate results obtained with IPsec-encrypted SIP may be lower than those obtained with plaintext SIP.
Objective: To benchmark the maximum registration rate the DUT can handle over an extended time period with zero failures.
Procedure: 1. Configure a DUT according to the test topology shown in Figure 3. 2. Set the registration timeout value to at least 3600 seconds. 3. Each register request MUST be made to a distinct Address of Record (AoR). Execute benchmarking algorithm as defined in
Davids, et al. Informational [Page 15]
RFC 7502 SIP Benchmarking Methodology April 2015
Section 4.10 to get the maximum registration rate. This rate MUST be recorded using any pertinent parameters as shown in the reporting format of Section 5.1. For example, the use of TLS or IPsec during registration must be noted in the reporting format. In the same vein, any specific backend processing (use of databases, authentication servers, etc.) SHOULD be recorded as well.
Expected Results: Provides a maximum registration rate.
Objective: To benchmark the re-registration rate of the DUT with zero failures using the same backend processing and parameters used during Section 6.7.
Procedure: 1. Configure a DUT according to the test topology shown in Figure 3. 2. Execute the test detailed in Section 6.7 to register the endpoints with the registrar and obtain the registration rate. 3. After at least 5 minutes of performing Step 2, but no more than 10 minutes after Step 2 has been performed, re-register the same AoRs used in Step 3 of Section 6.7. This will count as a re-registration because the SIP AoRs have not yet expired.
Expected Results: Note the rate obtained through this test for comparison with the rate obtained in Section 6.7.
Documents of this type do not directly affect the security of the Internet or corporate networks as long as benchmarking is not performed on devices or systems connected to production networks. Security threats and how to counter these in SIP and the media layer is discussed in RFC 3261, RFC 3550, and RFC 3711, and various other documents. This document attempts to formalize a set of common methodology for benchmarking performance of SIP devices in a lab environment.
[RFC3261] Rosenberg, J., Schulzrinne, H., Camarillo, G., Johnston, A., Peterson, J., Sparks, R., Handley, M., and E. Schooler, "SIP: Session Initiation Protocol", RFC 3261, June 2002, <http://www.rfc-editor.org/info/rfc3261>.
[Rtool] R Development Core Team, "R: A Language and Environment for Statistical Computing", R Foundation for Statistical Computing Vienna, Austria, ISBN 3-900051-07-0, 2011, <http://www.R-project.org>.
Davids, et al. Informational [Page 17]
RFC 7502 SIP Benchmarking Methodology April 2015
Appendix A. R Code Component to Simulate Benchmarking Algorithm
# Copyright (c) 2015 IETF Trust and the persons identified as # authors of the code. All rights reserved. # # Redistribution and use in source and binary forms, with or # without modification, are permitted provided that the following # conditions are met: # # The author of this code is Vijay K. Gurbani. # # - Redistributions of source code must retain the above copyright # notice, this list of conditions and # the following disclaimer. # # - Redistributions in binary form must reproduce the above # copyright notice, this list of conditions and the following # disclaimer in the documentation and/or other materials # provided with the distribution. # # - Neither the name of Internet Society, IETF or IETF Trust, # nor the names of specific contributors, may be used to # endorse or promote products derived from this software # without specific prior written permission. # # THIS SOFTWARE IS PROVIDED BY THE COPYRIGHT HOLDERS AND # CONTRIBUTORS "AS IS" AND ANY EXPRESS OR IMPLIED WARRANTIES, # INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF # MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE # DISCLAIMED. IN NO EVENT SHALL THE COPYRIGHT OWNER OR # CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, # INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES # (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE # GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS # INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, # WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING # NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE # OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH # DAMAGE.
w = 0.10 d = max(0.10, w / 2) DUT_max_sps = 460 # Change as needed to set the max sps value # for a DUT
Davids, et al. Informational [Page 18]
RFC 7502 SIP Benchmarking Methodology April 2015
# Returns R, given r (initial session attempt rate). # E.g., assume that a DUT handles 460 sps in steady state # and you have saved this code in a file simulate.r. Then, # start an R session and do the following: # # > source("simulate.r") # > find_R(100) # ... debug output omitted ... # [1] 458 # # Thus, the max sps that the DUT can handle is 458 sps, which is # close to the absolute maximum of 460 sps the DUT is specified to # do. find_R <- function(r) { s = 0 old_r = 0 h = 0 count = 0
# Note that if w is small (say, 0.10) and r is small # (say, <= 9), the algorithm will not converge since it # uses floor() to increment r dynamically. It is best # to start with the defaults (w = 0.10 and r >= 100).
cat("r old_r w d \n") while (TRUE) { cat(r, ' ', old_r, ' ', w, ' ', d, '\n') s = send_traffic(r) if (s == TRUE) { # All sessions succeeded
if (r > old_r) { old_r = r } else { count = count + 1
if (count >= 10) { # We've converged. h = max(r, old_r) break } }
r = floor(r + (w * r)) }
Davids, et al. Informational [Page 19]
RFC 7502 SIP Benchmarking Methodology April 2015
else { r = floor(r - (d * r)) d = max(0.10, d / 2) w = max(0.10, w / 2) } }
h }
send_traffic <- function(r) { n = TRUE
if (r > DUT_max_sps) { n = FALSE }
n }
Acknowledgments
The authors would like to thank Keith Drage and Daryl Malas for their contributions to this document. Dale Worley provided an extensive review that led to improvements in the documents. We are grateful to Barry Constantine, William Cerveny, and Robert Sparks for providing valuable comments during the documents' last calls and expert reviews. Al Morton and Sarah Banks have been exemplary working group chairs; we thank them for tracking this work to completion. Tom Taylor provided an in-depth review and subsequent comments on the benchmarking convergence algorithm in Section 4.10.
Davids, et al. Informational [Page 20]
RFC 7502 SIP Benchmarking Methodology April 2015
Authors' Addresses
Carol Davids Illinois Institute of Technology 201 East Loop Road Wheaton, IL 60187 United States
Phone: +1 630 682 6024 EMail: davids@iit.edu
Vijay K. Gurbani Bell Laboratories, Alcatel-Lucent 1960 Lucent Lane Rm 9C-533 Naperville, IL 60566 United States
Phone: +1 630 224 0216 EMail: vkg@bell-labs.com
Scott Poretsky Allot Communications 300 TradeCenter, Suite 4680 Woburn, MA 08101 United States