Work Load Design for NFV performance evaluation

Intel Mobile Communications

SG1 NIP#29

1

Background

• NFV enables use of high-volume servers to reduce TCO

• NFV is under standardization within ETSI technology cluster

• NFV evaluation is a crucial business need

• Multiple Stakeholders in the Telecom Ecosystem that are involved in the evaluation

2

Service Providers

EPC Vendors

IaaS Vendors

Compute Element Vendors

NFV as an Active Research Topic

Active Research to study multiple aspects of virtualization• Characterizing performance of NFV [2]• Understanding bottlenecks in NFV [3] • Evaluation of architecture and infrastructures using NFV [4]

References[2] Hirschman B., Mehta P., et.al., “High-performance evolved packet core signaling

and bearer processing on general-purpose processors”, IEEE Network, Pages 6-14, 2015, Vol. 29, Issue. 3.

[3] Rajan A.S., Gobriel S., et.al., “ Understanding the bottlenecks in virtualizing cellular core network functions”, In Proceedings of IEEE International Workshop on Local and Metropolitan Area Networks (LANMAN), Pages 1-6, April 2015.

[4] Technical Brief, “Supporting Evolved Packet Core for One Million Mobile Subscribers with Four Intel® Xeon® Processor-Based Servers” https://networkbuilders.intel.com/docs/MESH_Group_Intel_EPC_TB_FINAL.pdf

3

Importance

• Performance on virtualized cores vary for different workloads• Important to evaluate performance for

– Planning infrastructure deployment– Network roll-out– Business decisions

• There is no standardized framework for exchange of work-load descriptions between the various stakeholders to enable an effective performance evaluation

4

Work Load Design

5

• Information Exchange Standardization

Service Provider 1

Service Provider 2

Service Provider N

vEPC OEM 1

vEPC OEM 2

vEPC OEM M

Common Work Load Design Description

Impact

Enables• fair comparison of multiple

NFV platforms that comply with a similar scope

• network vendors to develop and compare solutions that satisfy trade-offs of features v/s performance v/s cost

• software vendors an integrators to develop solutions for different deployments

6

Evaluation

WorkLoad L

WorkLoad 2

WorkLoad 1

Summary

• vEPC work-load design for NFV performance evaluation• This NIP complements

– 3GPP TS 32.455 : Telecommunication Management; Key Performance Indicators (KPI) for the Evolved Packet Core (EPC); Definitions

7

8

Backup

ETSI NFV Work

• GS NFV-INF 001 NFV; Infrastructure Overview

• GS NFV-INF 004 NFV; Infrastructure; Hypervisor Domain

• GS NFV-REL 001 NFV; Resiliency Requirements

• GS NFV 002 NFV; Architectural Framework

• GS NFV 003 NFV; Terminology for Main Concepts in NFV

• GS NFV-INF 003 NFV; Infrastructure; Compute Domain

• GS NFV-INF 005 NFV; Infrastructure; Network Domain

• GS NFV-INF 010 NFV; Service Quality Metrics

• GS NFV-MAN 001 NFV; Management and Orchestration

• GS NFV-SWA 001 NFV; Virtual Network Functions Architecture

• GS NFV-SEC 003 NFV; NFV Security; Security and Trust Guidance

• GS NFV-PER 001 NFV; NFV Performance & Portability Best Practises

• GS NFV-PER 002 NFV; Proofs of Concept; Framework

• GS NFV-INF 007 NFV; Infrastructure; Methodology to describe Interfaces and Abstractions

• GS NFV-SEC 001 NFV; NFV Security; Problem Statement

• GS NFV 001 NFV; Use Cases

9

ETSI GS NFV-INF 010 (Service Quality Metrics)

• Purpose of NFV Service Quality Metrics

10

ETSI GS NFV-INF 010 (Service Quality Metrics)

• Service Quality Metrics in context of NFV Reference Architecture

11

ETSI GS NFV-INF 010 (Service Quality Metrics)

• Taxonomy of Service Quality Metrics

12