Performance Evaluation in LTE Networks (SGW-u Tunnels) Sina Shafaei - PowerPoint PPT Presentation

Chair for Network Architectures and Services Chair for Network Architectures and Services Technische Universität München Technische Universität München “IDP Final Talk” Performance Evaluation in LTE Networks (SGW-u Tunnels) Sina Shafaei Advisors: Daniel G. Raumer, Sebastian Gallenmüller Jan 18, 2016 1 Technische Universität München – Chair for Network Architectures and Services

Chair for Network Architectures and Services Technische Universität München Outline Definitions - LTE Networks - Evolved Packet Core - Serving Gateway (SGW) - GTP Packets Set Up and Installation - Software - MoonGen - Testbed Structure Performance Evaluation - Throughput - Packet Loss Conclusion and Summary - Final Remark - Future Work 2 Technische Universität München – Chair for Network Architectures and Services

Chair for Network Architectures and Services Technische Universität München Long Term Evolution (LTE) Networks 3 Technische Universität München – Chair for Network Architectures and Services

Chair for Network Architectures and Services Technische Universität München LTE Networks / The E-UTRAN (The access network) 4 Technische Universität München – Chair for Network Architectures and Services

Chair for Network Architectures and Services Technische Universität München LTE Networks / The Evolved Packet Core (EPC) 5 Technische Universität München – Chair for Network Architectures and Services

Chair for Network Architectures and Services Technische Universität München The Functions of the SGW The local Mobility Anchor point for inter-eNodeB handover • Mobility anchoring for inter-3GPP mobility • Lawful Interception • Packet routing and forwarding • Transport level packet marking in the up-link and the down-link • Accounting for inter-operator charging. For GTP-based links, the Serving GW generates • accounting data per UE and bearer 6 Technische Universität München – Chair for Network Architectures and Services

Chair for Network Architectures and Services Technische Universität München GPRS Tunnelling Protocol (GTP) IP-based protocol that allows tunnels between pairs of end point used in GSM, UMTS and LTE • core networks It provides mobility. (When UE is moving, the IP address remains same and packets are still • forwarded since tunneling is provided between PGW and eNodeB via SGW) GTP-U is used in user plane to carry user data traffic in GPRS, UMTS and LTE networks. UDP is used as transport protocol. 7 Technische Universität München – Chair for Network Architectures and Services

Chair for Network Architectures and Services Technische Universität München Set up and Installation Sgwuctl (SGW-u controller) • d eveloped originally by BISDN Revised Open Flow Libraries • adding OpenFlow support to the software to build control applications, controller frameworks and/or data path elements xDPd • a multi-platform, multi OpenFlow version data plane, open- source switch focusing on performance and extensibility DPDK • a set of libraries and drivers for fast packet processing ● MoonGen load generator developed by Paul 8 Technische Universität München – Chair for Network Architectures and Services

Chair for Network Architectures and Services Technische Universität München Why xDPd and not Open vSwitch? ● Virtual switching element is used to support the local traffic steering ● Both support the OpenFlow protocol as well as various tunneling protocols and dpdk ● Both have support for multiple logical switch instances But ... ● xDPd is designed to run multiple hardware platforms ( x86-gnu-linux, x86-dpdk, Cavium Octeon, Broadcom, EazyChip, and NetFPGA10G ) ● xDPd shows a somewhat cleaner architecture than OVS ● The implementation of the pipeline for different hardware platforms is eased in xDPd by the availability of ROFL 9 Technische Universität München – Chair for Network Architectures and Services

Chair for Network Architectures and Services Technische Universität München Set up and Installation Shell script to deploy the whole set up • GTP extension for MoonGen • Format of a GTP-u Header Our Set up in Baltikum Testbed GTP-u Packet Structure 10 Technische Universität München – Chair for Network Architectures and Services

Chair for Network Architectures and Services Technische Universität München Performance Evaluation Performance indicators: Throughput & Packet Loss • Number of registered bearers: 1, 100, 200, 300, 400 • Packet Size: 1400 Byte / 1024 Byte • Traffic rates: (100Mbit/s, 300Mbit/s, 500Mbit/s, 700Mbit/s) 1Gbit/s, 2Gbit/s, 3Gbit/s, • 4Gbit/s, 5Gbit/s, 6Gbit/s, 7Gbit/s, 8Gbit/s, 9Gbit/s, 10Gbit/s Duration of each test: 5 Minutes • Number of Tests: ~600 • 11 Technische Universität München – Chair for Network Architectures and Services

Chair for Network Architectures and Services Technische Universität München Performance Evaluation – Throughput / Packet Size: 1400Byte 12 Technische Universität München – Chair for Network Architectures and Services

Chair for Network Architectures and Services Technische Universität München Performance Evaluation – Throughput / Packet Size: 1024Byte 13 Technische Universität München – Chair for Network Architectures and Services

Chair for Network Architectures and Services Technische Universität München Performance Evaluation – Packet Loss / Packet Size: 1400Byte 14 Technische Universität München – Chair for Network Architectures and Services

Chair for Network Architectures and Services Technische Universität München Performance Evaluation – Packet Loss / Packet Size: 1024Byte 15 Technische Universität München – Chair for Network Architectures and Services

Chair for Network Architectures and Services Technische Universität München Summary and Conclusion Set up script which is prepared for this project can be used easily for further • investigation on both SGW and PGW and respectively their sub-components Behavior of SGW in higher traffic rates, can be said is, unpredictable • Processing time on GTP packets at SGW has an important impact on both of • throughput and packet loss َ Traffics with smaller packet sizes will result in higher throughput in high rates, in general • There are still limitations on SGW controller developed by BISDN (e.g, number of • registered bearers) xDPd and sgwuctl need a proper updated documentation(!) • 16 Technische Universität München – Chair for Network Architectures and Services

Chair for Network Architectures and Services Technische Universität München Future Work What happens if packets for different bearers are not sent in a round robin fashion but • rather at random? With regard to higher processing times when there are multiple bearers installed, does • the lookup speed depend on bearer ID and number of present bearers? In addition to GTP-u traffic, the SGW receives control requests, e.g., for adding • bearers. Investigations regarding the relationship between processing times of these requests and the data plane load can be performed. Evaluating the functions of xDPd as data plane with different set up / scenario • Is it possible to deploy the whole set up on one host? What would be the outcome? • 17 Technische Universität München – Chair for Network Architectures and Services

Chair for Network Architectures and Services Technische Universität München Thanks for your attention! Questions? 18 Technische Universität München – Chair for Network Architectures and Services