an o s perspective on networks active messages and u net
play

An O/S perspective on networks: Active Messages and U-Net Theo - PowerPoint PPT Presentation

Feb 11, 2023 •271 likes •582 views

An O/S perspective on networks: Active Messages and U-Net Theo Jepsen Cornell University 17 October 2013 Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 1 / 30 Brief History Overview of Parallel Computing

  1. An O/S perspective on networks: Active Messages and U-Net Theo Jepsen Cornell University 17 October 2013 Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 1 / 30

  2. Brief History Overview of Parallel Computing ◮ 1962: Burroughs Corporation introduces D825 (4 CPUs connected to 16 memory modules) ◮ 1967: Amdahl’s Law: predicting the maximum speedup when new CPUs are added ◮ 1969: Multics introduced (8 CPUs) ◮ 1970: C.mmp multiprocessor (16 CPUs) ◮ 1976: ILLIAC IV (up to 256 CPUs): ”perhaps the most infamous of Supercomputers” Source: en.wikipedia.org Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 2 / 30

  3. Networked Parallel Computers ◮ Tightly coupled multiprocessor machines ◮ Task parallelism ◮ Communication should be responsive ◮ Communication should not have high overhead Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 3 / 30

  4. Connection Machine 5 (CM-5) ◮ Hypercubic routing network ◮ Intended for AI, but used for scientific computing ◮ communication channel over a hypercubic routing network Source: Photographer: Tom Trower, 1993 Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 4 / 30

  5. Networking Problems Faced by Parallel Computers ◮ High-latency ◮ Throughput optimization based on big messages ◮ Disconnect between software and hardware design ◮ Synchronous messaging Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 5 / 30

  6. Active Messages: a Mechanism for Integrated Communication and Computation In Proceedings of the 19th Annual International Symposium on Computer Architecture , 1992 Thorsten von Eicken David E. Culler Seth Copen Goldstein Klaus Erik Schauser Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 6 / 30

  7. Outline ◮ Motivation for Active Messages ◮ Solution – how do they achieve the goals ◮ Implementation ◮ Evaluation of Active Messages ◮ Perspective Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 7 / 30

  8. Active Messages: Motivation ◮ CPU design based on raw performance, not so much on networking ◮ Synchronous messaging is slow due to high latency ◮ Communication and computation should be combined Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 8 / 30

  9. Active Messages: Solution ◮ Take advantage of fast DMA and specific network interface capabilities ◮ Use asynchronous messaging paradigm; latency becomes less of a problem ◮ Implement queues to buffer data so that it is ready for computation Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 9 / 30

  10. Asynchronous Messaging Source: same paper

  11. Asynchronous Messaging Source: same paper

  12. Utilizes resources (almost) as predicted Source: same paper Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 12 / 30

  13. Active Messages: Conclusion ◮ Case for hardware to better support networking ◮ Good basis for parallel computing (Split-C) ◮ Design is specific to parallel computing Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 13 / 30

  14. Active Messages: Perspective ◮ What happened to Active Messages? ◮ Where is parallel computing? Niche computing? ◮ Change in hardware? We stopped increasing CPU frequency Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 14 / 30

  15. More Than Low Latency ◮ Low latency communication is essential in parallel computing ◮ What happens with multiple processes? ◮ What about isolation? ◮ Do you let the kernel handle isolation? Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 15 / 30

  16. Networking Through Kernel ◮ Kernel abstracts networking device ◮ Processes must use the underlying protocols ◮ Kernel responsible for management and isolation (policy) Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 16 / 30

  17. User Space Networking ◮ Remove kernel from critical path ◮ Reduce the latency between network interface and process ◮ Program can decide protocols Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 17 / 30

  18. U-Net: A User-Level Network Interface for Parallel and Distributed Computing 15th SOSP, December 1995. Thorsten von Eicken ? Anindya Basu Vineet Buch Werner Vogels Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 18 / 30

  19. Outline ◮ Motivation/Goals for U-Net ◮ Design ◮ Implementation ◮ Evaluation of U-Net ◮ Perspective Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 19 / 30

  20. U-Net Goals ◮ Low latency and high bandwidth with small messages ◮ Application access to underlying protocols ◮ Practical integration with existing systems Solution: virtual network interfaces (endpoints) Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 20 / 30

  21. Source: same paper Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 21 / 30

  22. Source: same paper Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 22 / 30

  23. Better utilization of available bandwidth Source: same paper Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 23 / 30

  24. Supports other protocols Source: same paper Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 24 / 30

  25. Performs well compared to other systems (*run on different h/w) Source: same paper Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 25 / 30

  26. U-Net: Conclusion ◮ Reasonable performance ◮ Is there process isolation? ◮ TCP/IP implementation should facilitate integration with other networks ◮ Adoption? Applicability in existing systems? ◮ Approach reminiscent of a microkernel’s: move networking to userspace Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 26 / 30

  27. U-Net: Perspective ◮ Is there sufficient granularity of network resources? ◮ Do we have user-level interfaces today? Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 27 / 30

  28. What happened to U-Net? ◮ Virtual Interface Network (1997) – userspace zero-copy networking ◮ RDMA – Remote DMA: zero-copy over network done by network card ◮ Myrinet (1995): two fibre optic cables, packet switching, lower protocol overhead than Ethernet, low latency ◮ Infiniband (1999): high speed I/O (storage) ◮ Virtualization: hypervisor kernel-bypass Source: en.wikipedia.org Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 28 / 30

  29. Active Messages and U-Net: Comparison ◮ Both are in parallel computing environment ◮ However, U-Net also focused on TCP and UDP ◮ Do we use this today? ◮ What happened to parallel computing Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 29 / 30

  30. Perspective ◮ Less tightly-coupled processing today ◮ Active Networks in mid 90s ◮ Disliked for disobeying end-to-end argument ◮ 15 years later: SDN ◮ Is parallel computing still important today? Theo Jepsen (Cornell University) CS 6410: Advanced Systems 17 October 2013 30 / 30

Recommend

AN O/S PERSPECTIVE ON NETWORKS Adem Efe Gencer 1 October 4 th , 2012 1 Department of Computer

AN O/S PERSPECTIVE ON NETWORKS Adem Efe Gencer 1 October 4 th , 2012 1 Department of Computer

AN O/S PERSPECTIVE ON NETWORKS Adem Efe Gencer 1 October 4 th , 2012 1 Department of Computer Science, Cornell University Papers 2 Active Messages: A Mechanism for Integrated Communication and Control, Thorsten von Eicken, David E. Culler,

506 views • 33 slides
Computer Networks I Computer Networks I Networks A networks connection structure is known as

Computer Networks I Computer Networks I Networks A networks connection structure is known as

Computer Networks I Computer Networks I Networks A networks connection structure is known as its network topology . Host nodes source and destination of messages Communication nodes where messages pass through; can be

304 views • 26 slides
Active Adversary Lecture 7 CCA Security MAC Active Adversary Active Adversary An active

Active Adversary Lecture 7 CCA Security MAC Active Adversary Active Adversary An active

Active Adversary Lecture 7 CCA Security MAC Active Adversary Active Adversary An active adversary can inject messages into the channel Active Adversary An active adversary can inject messages into the channel Eve can send ciphertexts to Bob

2.08k views • 160 slides
MAC. SKE in Practice. Lecture 5 Active Adversary Active Adversary An active adversary can

MAC. SKE in Practice. Lecture 5 Active Adversary Active Adversary An active adversary can

MAC. SKE in Practice. Lecture 5 Active Adversary Active Adversary An active adversary can inject messages into the channel Active Adversary An active adversary can inject messages into the channel Eve can send ciphertexts to Bob and get them

2.08k views • 155 slides
Active Adversary Lecture 7 CCA Security MAC Active Adversary An active adversary can inject

Active Adversary Lecture 7 CCA Security MAC Active Adversary An active adversary can inject

Active Adversary Lecture 7 CCA Security MAC Active Adversary An active adversary can inject messages into the channel Eve can send ciphertexts to Bob and get them decrypted Chosen Ciphertext Attack (CCA) If Bob decrypts all ciphertexts for Eve,

352 views • 14 slides
Deep networks CS 446 The ERM perspective These lectures will follow an ERM perspective on deep

Deep networks CS 446 The ERM perspective These lectures will follow an ERM perspective on deep

Deep networks CS 446 The ERM perspective These lectures will follow an ERM perspective on deep networks: Pick a model/predictor class (network architecture). (We will spend most of our time on this!) Pick a loss/risk. (We will almost

608 views • 40 slides
Active Networks Towards an Active Network Architecture, D. Tennenhouse, D. Wetherall, CCR

Active Networks Towards an Active Network Architecture, D. Tennenhouse, D. Wetherall, CCR

Active Networks Towards an Active Network Architecture, D. Tennenhouse, D. Wetherall, CCR 1996 Winner of ACM SIGCOMM 2007 Test-of-Time Award Internet in 1996: Routers are passive -- just move bits around Bits are either

1.09k views • 80 slides
Resources for Controlled Languages for Alert Messages and Protocols in the European Perspective

Resources for Controlled Languages for Alert Messages and Protocols in the European Perspective

Resources for Controlled Languages for Alert Messages and Protocols in the European Perspective Sylviane Cardey 1 , Krzysztof Bogacki 2 , Xavier Blanco 3 , Ruslan Mitkov 4 1 Centre Tesnire, Universit de Franche-Comt (FR) 2 Uniwersytet

491 views • 22 slides
Graphs and social networks Social networks Active area of research motivated in part by

Graphs and social networks Social networks Active area of research motivated in part by

csci 210: Data Structures Graphs and social networks Social networks Active area of research motivated in part by growing public fascination withe the complex connectedness of modern society Social networks have grown in

590 views • 11 slides
Learning Algorithms for Active Learning Plan Background Matching Networks

Learning Algorithms for Active Learning Plan Background Matching Networks

Learning Algorithms for Active Learning Plan Background Matching Networks Active Learning Model Applications: Omniglot and MovieLens Critique and discussion Background: Matching Networks (Vinyals et al. 2016)

241 views • 20 slides
Distribution Networks Laura Ramrez Elizondo Learning Objectives What is an active

Distribution Networks Laura Ramrez Elizondo Learning Objectives What is an active

Distribution Networks Laura Ramrez Elizondo Learning Objectives What is an active distribution network? What are the main differences between the passive and active distribution networks? What are some challenges in the

85 views • 8 slides
Fast classification using sparsely active spiking networks Hesham Mostafa Institute of neural

Fast classification using sparsely active spiking networks Hesham Mostafa Institute of neural

Fast classification using sparsely active spiking networks Hesham Mostafa Institute of neural computation, UCSD Artificial networks vs. spiking networks backpropagation output layer ...... ...... Multi-layer networks are extremely hidden

431 views • 18 slides
Capsule Networks and Active Learning Chris Aasted, PhD Lockheed Martin Autonomous Systems

Capsule Networks and Active Learning Chris Aasted, PhD Lockheed Martin Autonomous Systems

Higher Performance with Less Data via Capsule Networks and Active Learning Chris Aasted, PhD Lockheed Martin Autonomous Systems Outline Problem Statement Capsule Networks Transfer Learning Active Learning Datasets

571 views • 30 slides
Active Implants in MRI: The MRI Industry's Perspective Michael Steckner, PhD, MBA Toshiba

Active Implants in MRI: The MRI Industry's Perspective Michael Steckner, PhD, MBA Toshiba

Active Implants in MRI: The MRI Industry's Perspective Michael Steckner, PhD, MBA Toshiba Medical Research Institute USA Disclosure Employee of Toshiba Chair of NEMA/MITA (industry organization) MR Technical committee Co-convenor of

408 views • 26 slides
Midterm review CSCI 466: Networks Keith Vertanen

Midterm review CSCI 466: Networks Keith Vertanen

Midterm review CSCI 466: Networks Keith Vertanen Network architecture 2 Encapsula8on High-level messages encapsulated in low-level messages

1.25k views • 111 slides
An approach to modeling short messages in spatio- temporal networks Amosse EDOUARD, PhD student

An approach to modeling short messages in spatio- temporal networks Amosse EDOUARD, PhD student

An approach to modeling short messages in spatio- temporal networks Amosse EDOUARD, PhD student Nhan LE-THAN, Supervisor Outline 1. Concept & Objective 2. The Radio Social Platform 3. Realisation 4. Contribution 5. Perspectives 6. Current

426 views • 24 slides
Proposal for a Register of Distributed Energy Resources A Networks Perspective Dr Stuart

Proposal for a Register of Distributed Energy Resources A Networks Perspective Dr Stuart

Proposal for a Register of Distributed Energy Resources A Networks Perspective Dr Stuart Johnston General Manager, Network Transformation Energy Networks Australia 27 March 2018 Energy Networks Australia Energy Networks Australia Members

457 views • 11 slides
The Active Card An Active Mind in an Active Body More people, More Active, More often! The

The Active Card An Active Mind in an Active Body More people, More Active, More often! The

The Active Card An Active Mind in an Active Body More people, More Active, More often! The Active vision Active The Olympic Legacy 100,000 card holders by 2012 SMART card technology Improved communications for

421 views • 21 slides
The Role of Enpr-EMA - A Neonatal Networks Perspective Wolfgang Gpel German Neonatal

The Role of Enpr-EMA - A Neonatal Networks Perspective Wolfgang Gpel German Neonatal

The Role of Enpr-EMA - A Neonatal Networks Perspective Wolfgang Gpel German Neonatal Network University of Lbeck Expectations for Enpr-EMA A Neonatal Networks perspective Who we are 40 Neonatal Intensive Care Units from

344 views • 10 slides
Sharing Multiple Messages over Mobile Networks Yuxin Chen, Sanjay Shakkottai, Jeffrey G.

Sharing Multiple Messages over Mobile Networks Yuxin Chen, Sanjay Shakkottai, Jeffrey G.

2011 Infocom, Shanghai April 12, 2011 Sharing Multiple Messages over Mobile Networks Yuxin Chen, Sanjay Shakkottai, Jeffrey G. Andrews Information Spreading over MANET users over a unit area Each user wishes to spread

419 views • 23 slides
Sending Messages on Communication Networks on Time Ronald Koch, Britta Peis, Martin Skutella,

Sending Messages on Communication Networks on Time Ronald Koch, Britta Peis, Martin Skutella,

Sending Messages on Communication Networks on Time Ronald Koch, Britta Peis, Martin Skutella, Andreas Wiese TU Berlin The Message Routing Problem In a distributed system, processes residing at different nodes of the network communicate by

1.48k views • 82 slides
Multimedia Data Distribution and Processing in IP Networks Active Network Legacy Eva Hladk

Multimedia Data Distribution and Processing in IP Networks Active Network Legacy Eva Hladk

Multimedia Data Distribution and Processing in IP Networks Active Network Legacy Eva Hladk a, CESNET & Masaryk University, Czech Republic The Ninth International Conference on Networks, ICN 2010 Menuires, The Three Valley, 12. 4. 2010

643 views • 39 slides
Flow Networks A new perspective of complex systems Contents 1 Flow Networks 2 Common Patterns

Flow Networks A new perspective of complex systems Contents 1 Flow Networks 2 Common Patterns

LOGO School of Systems Beijing Normal University Jiang Zhang @ Ali forum of complexity science 2014 Flow Networks A new perspective of complex systems Contents 1 Flow Networks 2 Common Patterns 3 Constructal Law? School of Systems

420 views • 25 slides
(Active) Flow and Noise Control A (Limited) Experimental Perspective F. S. Alvi Florida State

(Active) Flow and Noise Control A (Limited) Experimental Perspective F. S. Alvi Florida State

Some Challenging Problems in (Active) Flow and Noise Control A (Limited) Experimental Perspective F. S. Alvi Florida State University Florida Center for Advanced Aero-Propulsion Future Directions in CFD Research Aug 6-8, 2012, Hampton Roads,

931 views • 32 slides

More recommend