TCP/Generic Segmentation Offload and Its Application in Xen Herbert - PowerPoint PPT Presentation

Aug 13, 2023 •138 likes •279 views

TCP/Generic Segmentation Offload and Its Application in Xen Herbert Xu Principal Software Engineer Red Hat Asia Pacific What is TSO? Faster Ethernet (Gigabit) => higher CPU load: 1500-byte Ethernet MTU set in 70's. Amount of data

TCP/Generic Segmentation Offload and Its Application in Xen Herbert Xu Principal Software Engineer Red Hat Asia Pacific
What is TSO? ● Faster Ethernet (Gigabit) => higher CPU load: – 1500-byte Ethernet MTU set in 70's. – Amount of data per second 100 times higher. – CPU load per second 100 times higher. – Jumbo frames (9000-byte MTU or higher) help. – Hard to deploy due to PMTU issues.
What is TSO? ● Solution: Offload segmentation to NIC: – Effectively increases local MTU to ~64KB. – Stateless offload => easy OS support (cf TOE). – Supported by major NIC vendors and OSes. – Complements checksum offload. – Greatly reduces MTU-related CPU load.
TSO in Linux ● Added in August 2002 by Alexey Kuznetsov. ● Original version was incredibly fast. ● It ignored congestion window requirements :) ● Stable since 2.6.16.10/2.6.17. ● Enabled by default on TG3 and E1000.
Xen Paravirtual Networking ● Simulates a NIC in software. ● dom0/netback <=> domU/netfront. ● Operates on Ethernet packets. ● Uses ring buffer like a real NIC. ● Uses page flipping unlike a real NIC. ● Checksum offload support.
Xen Paravirtual Networking ● Performance problem with TCP: – loopback in domU: 5543.91Mb/s – domU => dom0: 1228.08Mb/s – domU => domU: 323.91Mb/s ● Loopback performance on par with baremetal. ● domU/domU performance less than wire speed.
Xen Paravirtual Networking ● MTU(lo) = 16436, MTU(eth0) = 1500. ● Change MTU(eth0) to 16436? ● 16436 bytes > 4KB (page size), requires SG. ● Change MTU(lo) to 1500. ● Throughput down to 2178.86Mb/s.
Xen Paravirtual Networking ● Solution: Implement SG for Xen. ● domU => dom0 throughput reaches 3097.49Mb/s with MTU of 16436. ● Comparable with 5543.91Mb/s on lo. ● Remaining difference due to longer code path. ● Higher MTU unrealistic due to PMTU.
Xen Paravirtual Networking ● Solution: TCP Segmentation Offload. ● No segmentation at all within Xen. ● Effective MTU of ~64KB within Xen. ● domU => dom0 throughput: 3208.41Mb/s. ● domU => domU throughput: 1678.52Mb/s. ● domU/domU has extra copy for memory protection. Can be removed with MMU help.
Generic Segmentation Offload ● Problem: Fails if packet hits non-TSO NIC. ● Solution: Simulate TSO in dom0. ● Add TCP knowledge to generic path? ● Add GSO infrastructure first. ● Packet => GSO => IPv4 => TCP.
Generic Segmentation Offload ● Reality check: Solves more than one problem? – TCP/ECN support with TSO. – TSO over IPv6. – Share code with UFO (UDP Fragment Offload). – Potential to enable TSO on all NICs. – Support more protocols, e.g., DCCP.
Gory Details of GSO/SG in Xen ● Extend ring buffers as real NICs do. ● Chaining descriptors for SG. ● Add extra descriptor for GSO. ● Same strategy can be used fo TX checksum. ● Maintain compatibility with feature negotiation.
Questions

Recommend

HIERARCHICAL QOS HARDWARE OFFLOAD Yossi Kuperman, Maxim Mikityanskiy, 2020 AGENDA Hierarchical

HIERARCHICAL QOS HARDWARE OFFLOAD Yossi Kuperman, Maxim Mikityanskiy, 2020 AGENDA Hierarchical Token Bucket Brief description of HTB and its issues HTB offload solution Modifications to HTB to solve the issues and offload the logic Current

1.06k views • 13 slides

Segmentation of tomographic images and caracterisation of porous media : Application to tuffeau,

Segmentation of X-ray tuffeau images September 4, 2008 1 de 1 Summary: Segmentation of tomographic images and caracterisation of porous media : Application to tuffeau, historical Loire-Valley building stone. E. Le Trong, O. Rozenbaum, S.

664 views • 45 slides

GCN INTRODUCTION AND ITS APPLICATION IN 3D POINT CLOUD SEMANTIC SEGMENTATION Yisong Li (NVIDIA),

GCN INTRODUCTION AND ITS APPLICATION IN 3D POINT CLOUD SEMANTIC SEGMENTATION Yisong Li (NVIDIA), Guohao Li (KAUST) Grid Data vs General Graphs CNN vs GCN ResGCN OUTLINE Experiments on 3D Cloud Point Segmentation Sequential

1.31k views • 102 slides

1 Definition of a simple generic class Why generic programming (cont.) class Pair <T> {

Topics background and goals of generic programming basics of generic classes = parameterized types generic methods for general algorithms inheritance rules for generic types Generic programming in Java bounded type parameters

261 views • 5 slides

VIDEO SIGNALS Segmentation WHAT IS SEGMENTATION WHAT IS SEGMENTATION Segmentation is a

VIDEO SIGNALS Segmentation WHAT IS SEGMENTATION WHAT IS SEGMENTATION Segmentation is a fundamental low level operation on Segmentation is a fundamental low-level operation on images. If an image is already partitioned into segments

639 views • 51 slides

Application of optimal sampling lattices on CT image reconstruction and segmentation or three

Application of optimal sampling lattices on CT image reconstruction and segmentation or three dimensional printing X I QI AN G ZH E N G Division of Health and Natural Sciences, Voorhees College, Denmark, SC 29042 Abstract First introduce

794 views • 31 slides

Segmentation Bottom-up Segmentation Semantic / instance segmentation Many Slides from L.

Segmentation Bottom-up Segmentation Semantic / instance segmentation Many Slides from L. Lazebnik. Outline Bottom-up segmentation Superpixel segmentation Semantic segmentation Metrics Architectures

1.28k views • 49 slides

Open tools and methods for large scale segmentation of Very High Resolution satellite images

Introduction Data representation and conversion Generic framework for large scale segmentation Pre and post processing Conclusion Open tools and methods for large scale segmentation of Very High Resolution satellite images Julien Michel

592 views • 29 slides

Multicriteria 3D PET image segmentation Application to oncology and cancer lesion extraction

Introduction Methodology Experimental study Future works Multicriteria 3D PET image segmentation Application to oncology and cancer lesion extraction Francisco Javier Alvarez Padilla, lose Grossiord, Barbara Romaniuk, Benot Naegel,

803 views • 34 slides

Topics to be covered Motivation: match application domains Generic match operator

A survey of approaches to automatic schema matching E. Rahm, and P. A. Bernstein. The VLDB Journal , 10(3): 334-350, 2001. Presented by Joel So (j2so@cs.uwaterloo.ca) Topics to be covered Motivation: match application domains Generic

397 views • 8 slides

Generic vs Alternative Specific Coefficients in Conditional Logits: An Application to Party

Generic vs Alternative Specific Coefficients in Conditional Logits: An Application to Party Choice Toni Rodon www.tonirodon.cat 20 October 2016 Stata Users Group meeting Universitat pompeu Fabra Motivation Car Bus Train Decision-maker

339 views • 21 slides

speaker change detection using fundamental frequency with application to multi-talker segmentation

speaker change detection using fundamental frequency with application to multi-talker segmentation May 16, 2019 Aidan Hogg, Christine Evers and Patrick Naylor Electrical and Electronic Engineering, Imperial College London, UK diarization

735 views • 29 slides

Topology-preserving discrete deformable model: Application to multi-segmentation of brain MRI

Introduction Background notions Method Discussion Topology-preserving discrete deformable model: Application to multi-segmentation of brain MRI Sanae Miri 1 , 2 , Nicolas Passat 1 , Jean-Paul Armspach 2 1 LSIIT, UMR 7005 CNRS/ULP - Universit

171 views • 16 slides

Sparsity Based Spectral Embedding: Application to Multi-Atlas Echocardiography Segmentation

Sparsity Based Spectral Embedding: Application to Multi-Atlas Echocardiography Segmentation Ozan Oktay, Wenzhe Shi, Jose Caballero, Kevin Keraudren, and Daniel Rueckert Department of Compu.ng Imperial

289 views • 24 slides

Blobworld Image segmentation using EM and its application to image querying [Carson et. al. ]

Blobworld Image segmentation using EM and its application to image querying [Carson et. al. ] Presented by: Nikhil V. Shirahatti Introduction: Why do we need an image retrieval system? So does text retrieval method work for images?

988 views • 34 slides

Hardware accelerating Linux network functions Roopa Prabhu, Wilson Kok Proceedings of netdev

Hardware accelerating Linux network functions Roopa Prabhu, Wilson Kok Proceedings of netdev 0.1, Feb 14-17, 2015, Ottawa, On, Canada Agenda Recap: offload models, offload drivers Introduction to switch asic hardware L2 offload

457 views • 26 slides

Segmentation Segmentation Segmentation Define the accurate boundaries of all objects in an image

Day 4 Lecture 2 Segmentation Segmentation Segmentation Define the accurate boundaries of all objects in an image Segmentation: Datasets Pascal Visual Object Classes Microsoft COCO 20 Classes 80 Classes ~ 5.000 images ~ 300.000 images

995 views • 26 slides

Pixel-Level Im Image Understanding wit ith Semantic Segmentation and Panoptic Segmentation

Pixel-Level Im Image Understanding wit ith Semantic Segmentation and Panoptic Segmentation Hengshuang Zhao The Chinese University of Hong Kong May 29, 2019 Part I: I: Semantic Segmentation Semantic Segmentation background car person

510 views • 39 slides

Generic Methods 36 What are Generic Methods? Generic methods = methods that introduce type

Generic Methods 36 What are Generic Methods? Generic methods = methods that introduce type parameters Similar to declaring a generic type but type parameter's scope is limited to method where it is declared The following are

584 views • 6 slides

Semantic Segmentation / Instance Segmentation Based on Deep learning Yiding Liu 2018.12.08

Semantic Segmentation / Instance Segmentation Based on Deep learning Yiding Liu 2018.12.08 Outline Overview of segmentation problem Semantic segmentation Instance Segmentation Our work Definition of segmentation problem Image

900 views • 54 slides

Segmentation using Segmentation using Bayesian Decision Theory Bayesian Decision Theory

Segmentation Segmentation Segmentation using Segmentation using Bayesian Decision Theory Bayesian Decision Theory Segmentation of images is the separation of pixels into different categories depending upon their intensities and/or other

109 views • 8 slides

InSilicoLab : from Generic Solutions to Application-Specific Science Gateway Joanna Kocot ,

InSilicoLab : from Generic Solutions to Application-Specific Science Gateway Joanna Kocot , Mariusz Sterzel, Tomasz Szepieniec, Maciej Golik, Tomasz Grabarczyk, Tomasz Twarg, Piotr Wjcik ACC CYFRONET AGH, ul. Nawojki 11, 30- 059 Krakw,

173 views • 16 slides

CPACE Generic Card & Mobile Payment Application Specification Collectively, the ECPC

CPACE Generic Card & Mobile Payment Application Specification Collectively, the ECPC Participating Schemes represent more than 300 million cards SRC Security Research Consulting, mandated Bancontact Company 9 Million 15 Million by the

1.76k views • 5 slides

Image Segmentation using Seg3D Segmentation From Clinical Scans RA RA LA RV LA LV RV LV

Image Segmentation using Seg3D Segmentation From Clinical Scans RA RA LA RV LA LV RV LV Tools Layers Viewers Final Segmenting Original Scan Segmentation using Seg3D 2 3 Tools Menu Interactive tools for operating on volume

448 views • 10 slides