Presentation is loading. Please wait.

Presentation is loading. Please wait.

S. Dasilva, D. Florissi, Y. Yemini (YY) ++ Distributed Computing & Communications (DCC) Lab Columbia University; D CC.

Similar presentations


Presentation on theme: "S. Dasilva, D. Florissi, Y. Yemini (YY) ++ Distributed Computing & Communications (DCC) Lab Columbia University; D CC."— Presentation transcript:

1 S. Dasilva, D. Florissi, Y. Yemini (YY) ++ Distributed Computing & Communications (DCC) Lab Columbia University; http://www.cs.columbia.edu/dcc D CC

2 Overview * Background: communications SW architectures * Introduction to active networks * The NetScript approach

3 Background: Evolution of Com SW Architectures

4 The Differences * Closed vertically integrated * Slow committee standards * Standardize algorithms & data formats * Fragmented SW market * Open multivendor product * Fast market standards * Standardize interfaces * Mass market for SW Intermediate NodeEnd Node

5 Traditional Com SW * Architecture l Rich features built-in closed vertically-integrated elements l Signalling protocols coordinate & control l Built to last * Process l Standardization through slow committees l Changes through slow evolution l Slow time to market

6 The Emergence of Mgmt * Architecture lReduced-complexity elements: instrumentation + agent lCoordination & control through NMS applications lSingle protocol, multiple MIBs * Process lFeatures are designed into MIBs lShort time-to-market lThrow-away SW

7 What Did SNMP Simplify? * SNMP created a new communication SW architecture * Vendors SW Problem 1: Minimal time-to-market * Vendors SW Problem 2: Featurizing products * Vendors SW problem 3: Reduced investment * But… lComplexity is shifted to users OAM lIncreasing operations risks & costs dominate end-users

8 Bridging The Paradigms * Key new: how to build com SW that enables l Fast time to market l Competitive featurization l Automation of mgmt & control * Driving opportunity: mgmt SW market * Enabling technology: mobile agents Instrumentation Agnt Srvc

9 Introduction to Active Nets

10 Active Nets = Programmable Nets * Making intermediate nodes programmable * By using mobile agents lPacket capsule approach lDelegated agents approach Instrumentation Agnt Srvc Pkt Transport/Switching HW

11 What Are Active Nets Good For? * Active protocols: lMulticast/multimedia protocols, signalling protocols * Active network management & control lConfiguration mgmt; fault mgmt, traffic analysis & shaping lSelf-managing networks lEnd-end mgmt (e.g., QoS) * Enable new network architectures l E.g., STM over IP Instrumentation Agnt Srvc Pkt Transport/Switching HW

12 Technical Rationale (Continued) * Active network security: lActive firewall filters & proxies, active monitoring of intrusion attacks, active response handlers, * Active application layer: lCaching/routing srvrs, filtering/compression, active phone/video

13 Market Rationale * Reduce time-to-market increase featurization * Vendor-based standardization * Create mass com SW markets * Create SW pull for communication HW Instrumentation Agnt Srvc Pkt Transport/Switching HW

14 The NetScript Approach

15 Goals & Approach * Goals l Language & environment to build, deploy, manage & protect active services * Approach l Language to simplify & unify composition of active elements l Program networks --end-end services -- not just nodes l Manageability & protection derived from language abstractions

16 NetScript Abstracts Packet Processing * The postscript analogy l Layer of abstractions to define page display l Separate HW printing engines from app SW * Abstract packet- stream processing primitives l Create a layer of abstractions of packet-stream processing l Separate HW engines from app SW

17 * Dataflow model: l active element = packet stream processor engine * Active elements are composed from boxes (computational channels) lBoxes encapsulate computations & resource allocation lComposition through interconnection lSynchronization of streams motions lAllocation of underlying resources lSimple model of interoperability The NetScript Language

18 Run-Time Model * Multiple implementation models l Java-based: translates to Java, executes on JVM l C…: NetScript =>c => bin * Secure delegation protocol

19 Virtual Active Networks (VAN) * How to deploy, manage & protect large multi- domain active nets? * VAN is a composable unit of end-end service l Composition through interconnection, layering and bridging * VAN is a unit of coordinated resource mgmt l Mapping active computations to underlying resources * VAN is a unit of protection

20 The Challenge of Active Nets Mgmt * Active components change dynamically * How will MIBs be created? Deployed? * How will mgrs monitor, analyze, control ANet?

21 Towards Compiler-Generated Mgmt * Goal: systemic design-time manageability * Managed properties are integrally designed * Compiler-generated instrumentation MIBs * A universal MIB structure unifies semantics

22 Conclusions * Active networks enable new com SW paradigm lRich applications potential; revolutionary improvement in mgmt/control and in network applications and services lNew roles for network nodes; new relationships between end- nodes, boundary nodes and intermediate nodes * NetScript provides a language system to support lSystematic, end-end programming of networks lDevelopment, deployment and mgmt of active systems


Download ppt "S. Dasilva, D. Florissi, Y. Yemini (YY) ++ Distributed Computing & Communications (DCC) Lab Columbia University; D CC."

Similar presentations


Ads by Google