H Mcast: A High-Throughput Middleware for a Universal Future - - PowerPoint PPT Presentation

H∀Mcast: A High-Throughput Middleware for a Universal Future Internet Multicast Service

Sebastian Meiling

Dominik Charousset, Thomas C. Schmidt, Matthias Wählisch

iNET RG, Department of Computer Science Hamburg University of Applied Sciences

August 2nd, 2011

Introduction Architecture Prototype Evaluation Conclusion

Outline

1 Introduction 2 Architecture 3 Prototype 4 Evaluation 5 Conclusion H A Mcast – Meiling 2

Introduction Architecture Prototype Evaluation Conclusion

Motivation

Network nodes are equipped with enhanced resources

advanced endsystem intelligence support for complex operations

This offers new service deployment options For example hybrid group communication

combine native IP and overlay multicast endsystems participate in routing and forwarding

H A Mcast – Meiling 3

Introduction Architecture Prototype Evaluation Conclusion

Objectives

Decouple application development from service deployment Universal service access through standardized APIs Adaptive service instantiation at runtime, depending on local network environment and node capabilities Provide incremental deployment and service evolution H A Mcast exemplary implements a universal multicast service

H A Mcast – Meiling 4

Introduction Architecture Prototype Evaluation Conclusion

Issues for Future Internet Services

Globally available network services rely on:

uniform deployment within networks and endsystems widely available (standardized) service APIs

Requires support by vendors of hardware and operating systems Multicast specific issues:

Divergent deployment states of multicast technologies Incompatible APIs for various multicast flavors Conflicting incentives for usage and deployment

H A Mcast – Meiling 5

Introduction Architecture Prototype Evaluation Conclusion

H A Mcast Architecture

System oriented multi service architecture Building blocks for new services:

technology transparent, universal service API extended middleware functionality on endsystems evolutionary, incremental service deployment

H A Mcast multicast service consists of:

an abstract naming scheme based on URIs (LocID split) the common multicast API, conforms to IRTF draft [1] a middleware component for endsystems Interdomain Multicast Gateways (IMGs)

H A Mcast – Meiling 6

Introduction Architecture Prototype Evaluation Conclusion

Incremental Deployment Scenario

HAMcast Network Stack

IPv4 IPv4 IPv6

Standard Network Stack

Scribe

<

IMG IMG Member of „G“ Member of „F“ Member of „F“ Member of „G“ Member of „F“ Member of „G“ IMG

IMGs inter-connect heterogeneous multicast domains Group members (F, G) independent of domain or technology Coexistence of standard and H A Mcast network stack

H A Mcast – Meiling 7

Introduction Architecture Prototype Evaluation Conclusion

Overview

Prototype implementation to demonstrate concepts of the H A Mcast architecture Utilizes hybrid group communication to provide a universal multicast service Late binding of multicast technology at runtime Implemented in C/C++ including boost library Multi OS support, currently runs on Linux and Mac OS

H A Mcast – Meiling 8

Introduction Architecture Prototype Evaluation Conclusion

Components

Common multicast API

Transparent multicast calls Implemented as client library

Middleware Component

User space daemon Instantiated once per host

Service Modules

Implement specific technology e.g. IP multicast, Scribe

HAMcast API-Library Underlay Socket Stub join/leave send/receive Group Application Middleware

Pluggable Technology Modules

IPC

...

Service Selection HAMcast Socket

Sockets

IPv4 IPv4 IPv4 IPv4 IPv4 IPv6 IPv4 IPv4 Scribe ALM Service- Discovery Group- Mapping

H A Mcast – Meiling 9

Introduction Architecture Prototype Evaluation Conclusion

Evaluation

Analyzing system performance of H A Mcast prototype Single sender-receiver scenario Hardware:

Hosts with QuadCore CPU, 8 GB RAM Network link with bandwidth of 1 Gbit/s

Comparison of H A Mcast-IP, H A Mcast-OM and IP multicast Metrics: throughput, loss and CPU usage Packet payload size from 100 to 1400 Bytes

H A Mcast – Meiling 10

Introduction Architecture Prototype Evaluation Conclusion

Packet Throughput

200000 400000 600000 800000 200 400 600 800 1000 1200 1400 Packet Throughput [Pakete/s] Payload [Bytes] IP−Stack HAMcast−IP HAMcast−OM MAX

H A Mcast – Meiling 11

Introduction Architecture Prototype Evaluation Conclusion

Data Throughput

200 400 600 800 1000 200 400 600 800 1000 1200 1400 Data Throughput [MBit/s] Payload [Bytes] IP−Stack HAMcast−IP HAMcast−OM MAX

H A Mcast – Meiling 12

Introduction Architecture Prototype Evaluation Conclusion

Packet Loss

1 2 3 4 5 200 400 600 800 1000 1200 1400 Packet Loss [%] Payload [Bytes] IP−Stack HAMcast−IP HAMcast−OM

H A Mcast – Meiling 13

Introduction Architecture Prototype Evaluation Conclusion

CPU Utility

50 100 150 200 250 300 350 400 200 400 600 800 1000 1200 1400 CPU Utility [%] Payload [Bytes] IP−Stack HAMcast−IP HAMcast−OM

H A Mcast – Meiling 14

Introduction Architecture Prototype Evaluation Conclusion

Conclusion

Prototype demonstrates feasibility of H A Mcast architecture Design enables extension and integration of new features Promising evaluation results verify prototype performance First deployment of prototype enables a hybrid group communication service in G-Lab testbed environment Active participation in IRTF SAM RG:

Standardization of common multicast API Cooperation within research community

Further information and download of prototype at: http://hamcast.realmv6.org Visit demo presentation here at EuroView

H A Mcast – Meiling 15

Introduction Architecture Prototype Evaluation Conclusion

End of Talk Thank you for your attention. Questions?

H A Mcast – Meiling 16

Introduction Architecture Prototype Evaluation Conclusion

References I

• M. Waehlisch, T. Schmidt, and S. Venaas, “A Common API

for Transparent Hybrid Multicast,” IETF, Internet-Draft – work in progress 02, July 2011.

• S. Meiling, D. Charousset, T. C. Schmidt, and M. Wählisch,

“System-assisted Service Evolution for a Future Internet – The HAMcast Approach to Pervasive Multicast,” in Proc. of IEEE GLOBECOM 2010, Workshop MCS 2010. Piscataway, NJ, USA: IEEE Press, Dec. 2010, pp. 913–917.

• M. Wählisch, T. C. Schmidt, and G. Wittenburg, “On

Predictable Large-Scale Data Delivery in Prefix-based Virtualized Content Networks,” Computer Networks, 2011, accepted for Publication.

H A Mcast – Meiling 17

Introduction Architecture Prototype Evaluation Conclusion

References II

• T. Schmidt, M. Waehlisch, and G. Fairhurst, “Multicast

Mobility in Mobile IP Version 6 (MIPv6): Problem Statement and Brief Survey,” IETF, RFC 5757, February 2010.

• T. Schmidt, M. Waehlisch, and S. Krishnan, “Base Deployment

for Multicast Listener Support in Proxy Mobile IPv6 (PMIPv6) Domains,” IETF, RFC 6224, April 2011.

H A Mcast – Meiling 18