NRT: Collaborative Research: A Unified Experimental Environment for Diverse Networks

点击翻译按钮获取中文摘要

A critical issue for t e development of future network and distributed system technologies is the availabil-ity of testbed infrastructure for conducting research.Availability as three components.First,it requiresthat testbeds provide a set of diverse and modern networking resources in realistic scenarios:environmentsthat include PCs,(potentially mobile)wireless and sensor nodes,commercial-quality switches and routers,and modern network interconnects.Second,availability requires software to e .ectively manage the use ofthese resources:this requires providing "hardware level" access to experimenters,through a uni .ed and re-motely accessible user interface,in a manner that also ensures security both inside and outside experimentalenvironments.Third,availability requires infrastructure for experiment manipulation,monitoring,measure-ment,and validation:e.g.,the ability to inject network tra .c and faults,the ability to obtain precise datafrom experiments,and the assurance that the testbed is an accurate model of the real world.The work described in this proposal is to build on the University of Utah 's Netbed to address the threeaspects of availability described above.The 165,000 line Netbed software system is e .ectively an operatingsystem for experiments in distributed systems and networking.The bulk of its code is independent ofthe network 's actual mec anisms,so it presents a common interface to experimenters across a variety ofmec anisms,including real networks,emulated networks in a cluster,and simulated networks.Netbed 'ssupport for ybrid experimentation allows .exible combinations of real code,abstracted tra .c generation,synthetic network environments,and real networks.Emulab Netbed 's primary resource,is an extremelysuccessful 170-node,time-and space-shared automated cluster testbed that as been in 24/7 production usefor several years.Netbed is currently being used by 440 researchers,from 50 institutions,for over 80 projects.The proposed work will provide deliverables in four areas.First,the existing Emulab cluster and itsinterconnect will be enlarged and modernized.Emulab will grow to provide continuing service to the public,including user support.The complementary Utah and Wisconsin facilities (WAIL)will be linked,allowingresource allocation and co-scheduling across them.This will provide a seamless transition from simulation,through emulation on PC-based routers,to large-scale testing on real routers.Second,Netbed will beextended to control,as transparently to users as possible,many additional kinds of network devices anddeployments.Netbed will provide four di .erent platforms of wireless,sensor,and mobile devices,rangingfrom a .xed array of wireless nodes (a "tabletop" environment)to an actual set of mobile devices deployedon a campus scale.The Netbed facilities will also be generalized with the addition of "hard" devices -IP routers,layer 1 switches,optical devices,Intel IXPs,and nodes deployed in the wide-area - as wellas new "soft" devices suc as virtual nodes and links to enable experiments at 1 - 2 orders of magnitudegreater scale.Third,the Netbed system,speci .cally the Emulab cluster,will be "hardened" so that,whendeployed,it can safely quarantine malicious code while still being remotely accessible to researchers.Theenhanced Emulab will be able to serve as a testbed for most network security researc and evaluation,sucas worm containment,IP traceback,intrusion detection,and counter measures for denial-of-service (DoS),and distributed denial-of-service (DDoS)attacks.Support for researc in these areas will require Emulab tosupport osts running Microsoft Windows.Fourth,new software will provide a comprehensive system forrealistic network tra .c generation,network tra .c measurements,and a measurement arc ive.A wide areanetwork monitoring system and measurement archive will provide important data to Netbed researchers andthe community at large.It will also serve as the primary source of data from whic we can validate ourmodeling and test e .orts.The intellectual merit of the proposed work lies in the creation of new technologies for networking anddistributed system experimentation.This work will develop new techniques for integrating widely variedresources,both actual and simulated,both hardware and software,within a single and uni .ed experimentalframework.This work will be realized in an expanded Emulab cluster,whic will continue to be available toresearchers,and it will also be realized in Netbed 's software that can be used by other institutions that wishto run their own testbeds - separately,or in cooperation with Netbed.This ties directly to the broaderimpacts of the proposed work.The creation and delivery of an expanded Emulab,as well as the softwarecapable of managing similar testbeds,will result in new testbeds and new testbed capabilities being madeavailable to researchers worldwide.This will directly advance the development of future systems,as well asthe teac ing and training of students and computing professionals.