|
|
The EMIX Concept Written by Dr. Roman Krzanowski & Bruno Giguère Internet MIX or IMIX is a well known and accepted concept of an average size composition of layer 3 (IP) packets that traverse the IP network [1]. The operational importance of IMIX lays in the fact that it provides a reference model of the aggregated flow of packets that can be used to test different aspects of the network performance and configuration of network elements. IMIX is by no way carved in stone and different IMIX models are used and proposed across the industry [2]. Yet, even if IMIX does not have the firmness of the gravitational constant of Newton’s law, IMIX has been embedded in test tools and test procedures. It does not define as we said, a unique testing model, but it does facilitate the comparison of test results and testing procedures. For the similar reason and with the similar hopes and objectives the EMIX concept has been created. What is EMIX? Ethernet MIX or EMIX is a reference composition of layer 2 (Ethernet) frames that traverse layer 2 Metro Ethernet Networks (MEN). The size of frames in EMIX would be defined by the service frame size, as defined by MEF [4]. To be precise the MEF concept of the service frame excludes the Preamble and Interface gaps, but includes FCS field [5].
MEN may be characterized by the more homogenous types of flows than can be seen in the IP backbone or IP network. Thus, the EMIX concept may have to be augmented with different variants such as EMIX for Ethernet Backhaul (EBH), EMIX for Business Services, EMIX for High-Speed Internet Access (HSIA), possibly the global interconnect and other Ethernet services. With these variants in mind and with the addition of new technologies to deliver applications over the MEN, we find that having a frame size distribution that is based on “actual” network measurements would be impractical as network engineering is done differently from one service providers to another. For these reasons will shall define an EMIX frame distribution that should reflect the high-level use of the different variants and not the detailed implementation. It is also suggested that EMIX may be a completely synthetic composition of packets that the industry would agree to use as a standard reference flow composition. Such an EMIX would facilitate standard testing and standard verification of MEN design. For example, current service acceptance testing methods for the Ethernet Mobile Backhaul service are using selected ranges of packet sizes similar to those used in testing of traffic polices in a generic switch. The test results would be much more relevant to service providers and customers, if such tests would use traffic composed of a mix of packet sized expected in the EBH traffic flow. While in principle it may not be possible to get an agreement on the Generic Reference EMIX across providers and services, it may be possible to agree to working EMIX compositions for specific services or tests. It would be possible then to use of the EMIX defined this way as a part of a standard test procedures for service activation, service verification and service and design testing. In addition to the flow composition based on the mix of different frame sizes, EMIX may be also characterized by the time sequence of frames. Such a definition, of an extension of the definition of EMIX would be of value in defining test sequences for testing traffic profiles with CIR, CBS, EIR, EBS components. Needless to say that currently there is no across-industry standard and accepted way for verification of layer 2 traffic profiles, and we find that such a standard would be welcomed. The EMIX concept is already getting some traction. It has been applied in the proposal for ITU-T Y.156sam document. As we are writing this technical note (Oct. 2010), Y.156sam is under development. Authors invite comments and suggestions on the further work on characterization and definition of the EMIX concept for Metro Ethernet Networks. Notes and References [1] “While IMIX, short for Internet Mix, is not officially defined by a standards organization, it has become an increasingly popular concern in the networking test arena. Its origins are derived in part out of a need to identify and simulate Internet network traffic according to frame size usage. Network performance in terms of throughput, latency, and packet loss can vary greatly depending on the traffic mix, thereby affecting the effectiveness of computer applications and ultimately the end user experience. Studies indicate that Internet traffic consists of fixed percentages of different frame sizes at a given point and time.” Test Methodology Journal, January 2006, accessed on Sept 28, 2010, on http://gospirent.com/whitepaper/IMIX_Test_Methodolgy_Journal.pdf [2] Performance Modeling of a Network Processor Data Path Using Queuing Systems. M. Kolendavelu, R. Thudt, accessed on Sept 23, 2010 at http://comsnets.org/2009/comsnets09-camera-12-20/comsnets09-final79.pdf [3] S. McCreary and K. C. Claffy, “Trends in Wide Area IP Traffic Patterns – A view from the Ames Internet Exchange”, Cooperative Association for Internet Data Analysis, CAIDA, USA. [4] Service frame: An Ethernet frame transmitted across the UNI toward the Ser-vice Provider or an Ethernet frame transmitted across the UNI toward the Subscriber (MEF 10.2.). [5] The MEF definition of the service frame is aligned with the definition of the ETH layer characteristic information as described in ITU-T G.8010/ Y.1306. p.7
|
