Understanding Latency in Software Defined Networks



Similar documents
Mazu: Taming Latency in Software Defined Networks

Measuring Control Plane Latency in SDN-enabled Switches

Securing Local Area Network with OpenFlow

Software Defined Networking What is it, how does it work, and what is it good for?

Networking Virtualization Using FPGAs

Software Defined Networking What is it, how does it work, and what is it good for?

基 於 SDN 與 可 程 式 化 硬 體 架 構 之 雲 端 網 路 系 統 交 換 器

Router Architectures

ON THE IMPLEMENTATION OF ADAPTIVE FLOW MEASUREMENT IN THE SDN-ENABLED NETWORK: A PROTOTYPE

Outline. Institute of Computer and Communication Network Engineering. Institute of Computer and Communication Network Engineering

Software-defined Latency Monitoring in Data Center Networks

DESIGN AND ANALYSIS OF TECHNIQUES FOR MAPPING VIRTUAL NETWORKS TO SOFTWARE- DEFINED NETWORK SUBSTRATES

WHITE PAPER. SDN Controller Testing: Part 1

OpenFlow with Intel Voravit Tanyingyong, Markus Hidell, Peter Sjödin

LTE - Can SDN paradigm be applied?

Data Center Load Balancing Kristian Hartikainen

Networking Topology For Your System

Software Defined Networks

Towards Software Defined Cellular Networks

Security Challenges & Opportunities in Software Defined Networks (SDN)

IxNetwork OpenFlow Solution

Empowering Software Defined Network Controller with Packet-Level Information

Deploying in a Distributed Environment

How To Test For Performance And Scalability On A Server With A Multi-Core Computer (For A Large Server)

High-Fidelity Switch Models for Software-Defined Network Emulation

On real-time delay monitoring in software-defined networks

ATCN 2014: SDN - Mobility and SDN: Mobility Management and Mobile Networks

SDN and Data Center Networks

Software Defined Networking and OpenFlow: a Concise Review

The Lagopus SDN Software Switch. 3.1 SDN and OpenFlow. 3. Cloud Computing Technology

Dynamic Controller Deployment in SDN

OpenFlow Overview. Daniel Turull

Transparent Optimization of Grid Server Selection with Real-Time Passive Network Measurements. Marcia Zangrilli and Bruce Lowekamp

An Intelligent Framework for Vehicular Ad-hoc Networks using SDN Architecture

Understanding OpenFlow

B4: Experience with a Globally-Deployed Software Defined WAN TO APPEAR IN SIGCOMM 13

D1.2 Network Load Balancing

Ethernet-based Software Defined Network (SDN) Cloud Computing Research Center for Mobile Applications (CCMA), ITRI 雲 端 運 算 行 動 應 用 研 究 中 心

Achieving Low-Latency Security

OpenFlow - the key standard of Software-Defined Networks. Dmitry Orekhov, Epam Systems

OpenDaylight Performance Stress Tests Report

Effective Switch Memory Management in OpenFlow Networks

The State of OpenFlow: Advice for Those Considering SDN. Steve Wallace Executive Director, InCNTRE SDN Lab Indiana University

...DYNAMiC INTERNET SOLUTiONS >> Reg.No. 1995/020215/23

What you need to know about SDN control and data planes

Flexible SDN Transport Networks With Optical Circuit Switching

Scaling Database Performance in Azure

SDN. WHITE PAPER Intel Ethernet Switch FM6000 Series - Software Defined Networking. Recep Ozdag Intel Corporation

Disaster-Resilient Backbone and Access Networks

TCP Offload Engines. As network interconnect speeds advance to Gigabit. Introduction to

Best Practices for Monitoring Databases on VMware. Dean Richards Senior DBA, Confio Software

How To Understand The Power Of A Network In A Microsoft Computer System (For A Micronetworking)

Products for the registry databases and preparation for the disaster recovery

OpenFlow and Onix. OpenFlow: Enabling Innovation in Campus Networks. The Problem. We also want. How to run experiments in campus networks?

Assessing the Performance of Virtualization Technologies for NFV: a Preliminary Benchmarking

SDN CENTRALIZED NETWORK COMMAND AND CONTROL

Autonomous Fast Rerouting for Software Defined Network

How To Understand The Power Of The Internet

Ethernet Fabric Requirements for FCoE in the Data Center

Scalable and Reliable control and Management for SDN-based Large-scale Networks. CJK CFI

Software Defined Networking (SDN)

CS 78 Computer Networks. Internet Protocol (IP) our focus. The Network Layer. Interplay between routing and forwarding

Definition of a White Box. Benefits of White Boxes

HP OpenFlow Protocol Overview

Data Center Networks and Basic Switching Technologies

Performance of Host Identity Protocol on Nokia Internet Tablet

Abstraction of a failure free Software Defined Network (SDN Application)

Poisoning Network Visibility in Software-Defined Networks: New Attacks and Countermeasures Sungmin Hong, Lei Xu, Haopei Wang, Guofei Gu

Towards Correct Network Virtualization. Soudeh Ghorbani Brighten Godfrey UIUC

Ahsay Online Backup Suite v5.0. Whitepaper Backup speed

Implementing OpenDaylight in a Carrier Network

Virtualization and SDN Applications

COLO: COarse-grain LOck-stepping Virtual Machine for Non-stop Service

How good can databases deal with Netflow data

Network Instruments white paper

Software Defined Networking Seminar

Why Software Defined Networking (SDN)? Boyan Sotirov

OpenFlow. Ihsan Ayyub Qazi. Slides use info from Nick Mckeown

50. DFN Betriebstagung

MailEnable Scalability White Paper Version 1.2

Advanced Study of SDN/OpenFlow controllers

How To Make A Vpc More Secure With A Cloud Network Overlay (Network) On A Vlan) On An Openstack Vlan On A Server On A Network On A 2D (Vlan) (Vpn) On Your Vlan

A Fuzzy Logic-Based Information Security Management for Software-Defined Networks

Cloud Cost Management for Customer Sensitive Data

Wire-speed Packet Capture and Transmission

Radhika Niranjan Mysore, Andreas Pamboris, Nathan Farrington, Nelson Huang, Pardis Miri, Sivasankar Radhakrishnan, Vikram Subramanya and Amin Vahdat

SOFTWARE-DEFINED NETWORKING AND OPENFLOW

How OpenFlow-based SDN can increase network security

Cisco Application Networking Manager Version 2.0

WHITE PAPER: BEST PRACTICES. Sizing and Scalability Recommendations for Symantec Endpoint Protection. Symantec Enterprise Security Solutions Group

Collecting Packet Traces at High Speed

SDN, OpenFlow and the ONF

Sockets vs. RDMA Interface over 10-Gigabit Networks: An In-depth Analysis of the Memory Traffic Bottleneck

Gigabit Ethernet Design

Open-source routing at 10Gb/s

Agenda. Enterprise Application Performance Factors. Current form of Enterprise Applications. Factors to Application Performance.

I/O virtualization. Jussi Hanhirova Aalto University, Helsinki, Finland Hanhirova CS/Aalto

Transcription:

Understanding Latency in Software Defined Networks Junaid Khalid Keqiang He, Sourav Das, Aditya Akella Li Erran Li, Marina Thottan 1

Latency in SDN Mobility Centralized Controller Openflow msgs 1 flows Time taken to install 1 rules? Can Some be as XXX large msecs?? as 1 secs!!! 2

Do applications care about latency? Intra-DC Traffic Engineering Latency is critical to many applications MicroTE routes predictable traffic on time scales of 1-2s Route Limits updates the applicability can take as long of SDN as applications.5s Latency can undermine MicroTE s effectiveness Fast Failover Applications assume latency is low and constant Reroute the affected flows quickly in face of failures Longer update time increases congestion and drops Latency can inflate failover time by nearly 2s! 3

Factors contributing to Latency? Robust Control Software Design and Distributed Controllers Speed of Control Programs and Network Latency Not received much attention Latency in network switches 4

Our Work Two contributions: Systematic experiments to explore latencies in production switches Latency in network switches Design a framework to overcome the impact of latencies 5

Outline Motivation Elements of Latency Measurement Methodology Inbound Latency Outbound Latency Insertion Modification Deletion 6

Outline Motivation Elements of Latency Measurement Methodology Inbound Latency Outbound Latency Insertion Modification Deletion 7

PCI Elements of Latency 1. Inbound Latency Controller Switch Switch Fabric CPU board I1 Memory DMA Forwarding Engine Lookup I2 ASIC SDK CPU I3 No Match Hardware Tables OF Agent Inbound Latency I1: Send to ASIC SDK I2: Send to OF Agent I3: Send to Controller PHY PHY Packet 8

PCI Elements of Latency 1. Inbound Latency 2. Outbound Latency Controller O1 Switch Switch Fabric CPU board Memory Forwarding Engine DMA Lookup ASIC SDK CPU OF Agent O4 Hardware Tables O2 O3 Outbound Latency O1: Parse OF Msg O2: Software schedules the rule O3: Reordering of rules in table O4: Rule is updated in table PHY PHY 9

Outline Motivation Elements of Latency Measurement Methodology Inbound Latency Outbound Latency Insertion Modification Deletion 1

Latency Measurements - Setup POX Controller Pktgen Libpcap eth eth1 eth2 Control Channel Flows IN Flows OUT Openflow Switch Switch CPU RAM Flow table size Data Plane Vendor A 1 Ghz 1 GB 896 14*1Gbps + 4*4Gbps Vendor B 2 Ghz 2 GB 496 4*1Gbps +4*4Gbps 11

Outline Motivation Elements of Latency Measurement Methodology Inbound Latency Outbound Latency Insertion Modification Deletion 12

Inbound Latency Increases with flow arrival rate CPU Usage is higher for higher flow arrival rates Flow Arrival Rate (packets/sec) Low Power CPU Mean Delay per packet_in (msec) CPU Usage ( % ) - 7.1 1 3.32 15.7 2 8.33 26.5 vendor A switch 13

inbound delay (ms) inbound delay (ms) Inbound Latency Increases with interference from outbound msgs 3 25 2 15 1 5 2 4 6 8 1 flow # (a) with flow_mod/pkt_out 3 25 Low Power 2CPU 15 1 Software Inefficiency 5 2 4 6 8 1 flow # (b) w/o flow_mod/pkt_out vendor A switch 2 flows/sec 14

Outline Motivation Elements of Latency Measurement Methodology Inbound Latency Outbound Latency Insertion Modification Deletion 15

Outbound Latency Latency for three different flow_mod operations Insertion Modification Deletion Impact of key factors on these latencies Table occupancy Rule priority structure 16

Outline Motivation Elements of Latency Measurement Methodology Inbound Latency Outbound Latency Insertion Modification Deletion 17

insertion delay (ms) insertion delay (ms) Insertion Latency Priority Effects 3 25 2 15 1 5 2 4 6 8 1 2 4 6 8 1 (a) Burst size 1, same priority (b) Burst size 1, increasing priority Vendor B switch 3 25 2 15 1 5 18

insertion delay (ms) insertion delay (ms) Insertion Latency Priority Effects 3 25 2 15 1 5 2 4 6 8 1 2 4 6 8 1 (a) Burst size 1, same priority (b) Burst size 1, increasing priority Vendor B switch 3 25 2 15 1 5 1 1 x TCAM 19

insertion delay (ms) insertion delay (ms) Insertion Latency Priority Effects 3 25 2 15 1 5 2 4 6 8 1 2 4 6 8 1 (a) Burst size 1, same priority (b) Burst size 1, increasing priority Vendor B switch 3 25 2 15 1 5 1 1 x 99 TCAM 2

insertion delay (ms) insertion delay (ms) Insertion Latency Priority Effects 3 25 2 15 1 5 2 4 6 8 1 2 4 6 8 1 (a) Burst size 1, same priority (b) Burst size 1, increasing priority Vendor B switch 3 25 2 15 1 5 1 1 99 x 11 TCAM 21

insertion delay (ms) insertion delay (ms) Insertion Latency Priority Effects 3 25 2 15 1 5 2 4 6 8 1 2 4 6 8 1 (a) Burst size 1, same priority (b) Burst size 1, increasing priority Vendor B switch 3 25 2 15 1 5 11 1 1 99 x TCAM 22

insertion delay (ms) insertion delay (ms) Insertion Latency Priority Effects 1 8 1 8 6 4 2 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 (a) Burst size 8, same priority (b) Burst size 8, decreasing priority Vendor A switch 6 4 2 23

insertion delay (ms) insertion delay (ms) Insertion Latency Priority Effects 1 8 1 8 6 4 2 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 (a) Burst size 8, same priority (b) Burst size 8, decreasing priority Vendor A switch 6 4 2 x TCAM 24

insertion delay (ms) insertion delay (ms) Insertion Latency Priority Effects 1 8 1 8 6 4 2 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 (a) Burst size 8, same priority (b) Burst size 8, decreasing priority Vendor A switch 6 4 2 x 1 11 12 13 14 TCAM 25

insertion delay (ms) insertion delay (ms) Insertion Latency Priority Effects 1 8 1 8 6 4 2 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 (a) Burst size 8, same priority (b) Burst size 8, decreasing priority Vendor A switch 6 4 2 99 x 1 11 12 13 14 TCAM 26

insertion delay (ms) insertion delay (ms) Insertion Latency Priority Effects 1 8 1 8 6 4 2 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 (a) Burst size 8, same priority (b) Burst size 8, decreasing priority Vendor A switch 6 4 2 99 x 1 11 12 13 14 TCAM 27

insertion delay (ms) insertion delay (ms) Insertion Latency Priority Effects 1 8 1 8 6 4 2 1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8 (a) Burst size 8, same priority (b) Burst size 8, decreasing priority Vendor A switch 6 4 2 99 x 1 11 12 13 14 TCAM 28

avg completion time (ms) avg completion time (ms) Insertion Latency Table occupancy Effects Rule Priority Structure and Table Occupancy Affected by the types of rules in the table 2 18 16 14 12 1 8 6 4 2 table 1 table 4 2 TCAM Organization and 18 # Of Slices 16 14 12 1 Switch Software Overhead 8 1 2 3 4 5 6 burst size (a) low priority rules into a table with high priority existing rules 6 4 2 table 1 table 4 1 2 3 4 5 6 burst size (b) high priority rules into a table with low priority existing rules Vendor B switch 29

Outline Motivation Elements of Latency Measurement Methodology Inbound Latency Outbound Latency Insertion Modification Deletion 3

modification dely (ms) modification delay (ms) Modification Latency Higher than Insertion latency for vendor B Not affected by rule priority but affected by table occupancy 16 14 12 1 8 6 4 2 16 14 12 1 Poorly optimized software in vendor B 8 6 4 2 2 4 6 8 1 5 1 15 2 (a) 1 rules in the table (b) 2 rules in table Vendor B switch 31

Outline Motivation Elements of Latency Measurement Methodology Inbound Latency Outbound Latency Insertion Modification Deletion 32

deletion delay (ms) deletion delay (ms) Deletion Latency Higher than Insertion latency for both vendor A and B Not affected by rule priority but affected by table occupancy 12 1 12 1 8 6 4 2 2 4 6 8 1 (a) 1 rules in table 8 6 4 2 5 1 15 2 (b) 2 rules in table Vendor B switch 33

deletion delay (ms) deletion delay (ms) Deletion Latency Higher than Insertion latency for both vendor A and B Not affected by rule priority but affected by table occupancy 14 12 1 8 6 4 2 8 Deletion is causing TCAM 6 Reorganization 4 2 4 6 8 1 (a) 1 rules in table 14 12 1 2 5 1 15 2 (b) 2 rules in table Vendor A switch 34

Summary Latency in SDN is critical to many applications Assumption: Latency is small or constant Latency is high and variable Varies with Platforms, Type of operations, Rule priorities, Table occupancy, Concurrent operations Key Factors: TCAM Organization, Switch CPU and inefficient Software Implementation 35

Summary Latency in SDN is critical to many applications Assumption: Latency is small or constant Latency is high and variable Varies with Platforms, Type of operations, Rule priorities, Table occupancy, Concurrent operations Key Factors: TCAM Organization, Switch CPU and inefficient Software Implementation 36

Backup Slides 37

Impact of Concurrent CPU jobs Impacts insertion delay. E.g. Polling Statistics Polling stats impacts more when table occupancy is higher table occupancy table occupancy 5 12 1 8 6 4 2 Low Power CPU delays concurrent jobs 1 2 3 4 5 flow stats polling count Vendor B 38