DOCSLIB.ORG

Ciena 5410 Service Aggregation Switch Datasheet

Total Page：16

File Type：pdf, Size：1020Kb

Download full-text PDF Read full-text

Load more

5410 SERVICE AGGREGATION SWITCH Ciena’s 5410 Service Aggregation Switch is a next-generation Carrier Ethernet/MPLS service aggregation platform that provides terabit switching capacity, massive service scalability, carrier-class resiliency, and a cutting-edge Features and Benefits feature set that supports advanced QoS, > Provides 100 Gb/s per slot and 1 Tb/s per system full duplex switching capacity with rich features OAM, and L2VPN capabilities. and scalable services The 5410 enables service providers to keep pace with constantly increasing demand > Implements advanced Ethernet and for bandwidth capacity and service density in metro networks to support mobile comprehensive MPLS features that support demanding mobile backhaul, backhaul and Ethernet transport applications, business VPNs, and business and business, transport, and residential residential triple play services. service applications, including 3G/4G wireless backhaul, L2VPN service Optimized for deployment in large central offices and Points of Presence (PoPs), the delivery and aggregation, L2 backhaul of L3VPNs, and FTTx/IP DSLAM 5410 features ten 100 Gb/s-capable line module slots, 1 Tb/s full duplex switching Aggregation capacity, and fully redundant common equipment that supports six-9s reliability > Sophisticated virtual switching and ensures service continuity. The 5410 offers high-density Gigabit Ethernet (GbE) architecture provides enhanced scale and Ten Gigabit Ethernet (10GbE) connectivity to subscriber edge and first-tier and flexibility for tens of thousands of aggregation switches, and high-performance 10GbE uplinks to IP/MPLS core services on a single system, including up to 128,000 VLANs, 64,000 virtual switch/routers. With its 1 Tb/s full duplex capacity today, the system can easily switches, and up to ten million MAC accommodate future line modules that will offer 100GbE interfaces. The 5410 addresses architecture and system resources also provide robust scalability for MAC addresses: > Delivers high reliability, six-9s one million MAC addresses per line module and up to ten million MAC addresses per availability, and 50 ms protection system. A sophisticated virtual switching architecture enables each 5410 to support switching resiliency using state-of- over 128,000 individual services (VLANs and Ethernet Virtual Circuits) for enhanced the-art hardware and software design coupled with advanced control plane scale and flexibility, and advanced ingress and egress hierarchical QoS capabilities and Ethernet OAM capabilities ensure SLAs can be honored while fully utilizing system assets. Protocol-agnostic and > Provides broad service stratification standards-based, the 5410 supports a full range of MPLS and Ethernet transport and and robust bandwidth allocation control protocols and L2VPN capabilities, including: MPLS Pseudowires/Virtual for guaranteed SLAs via MEF- Circuits; VPLS and Hierarchical VPLS; MPLS Label Edge Router and Label Switch 14-compliant hierarchical QoS capabilities Router functions; MPLS-TP; IEEE 802.1Qay PBB-TE, IEEE 802.1D/802.1Q/802.1ad/ bridging; and ITU-T G.8032 Ethernet Ring Protection Switching. The 5410 implements Continued on page 2 resiliency options for these protocols that enable 50 ms recovery from link or port failures, and provides a robust suite of Carrier Ethernet connectivity fault management and performance monitoring capabilities. D Datasheet The 5410 system hardware includes all the elements required for Synchronous Ethernet Features and Benefits (continued) allowing the 5410 to easily support Synchronous Ethernet applications, such as next- generation mobile backhaul, when software support is added in a future release. 5410 > Supports interworking between Line Modules and Control Modules include synchronization and clocking circuitry; Q-in-Q VLANs, MPLS Virtual Circuits, VPLS/H-VPLS L2VPNs, dedicated timing traces on the backplane enable precise network timing and and PBB-TE connections for synchronization signals to be delivered to and derived from all Ethernet interface ports; complete service flexibility and the system I/O module includes T1 or E1 BITS interfaces that can receive or provide and optimization of network resources BITS synchronization timing signals. These design elements will enable the delivery and distribution of transport-class timing for network-synchronized applications. > Provides operational efficiency and consistent system and service attributes with a field-proven service-aware Access Metro/Aggregation Core operating system, and supports comprehensive management for high-velocity service rollout and faster time to revenue MTU nxG, 10G (Q, P, H) G.8032 Ring 10G > Complies with IEEE/IETF/ITU/ Q-in-Q (Q) PBB-TE (P) MEF standards and features MTU Remote Terminal nxG H-VPLS (H) MEF, MPLS/VPLS, VLAN, and (Q, P) MPLS-TP PBB-TE services on every port G (Q) G.8032 National MPLS to support interoperability in 10G multivendor deployments SMB G (Q) (Q, P, H) > Facilitates implementation PoP/RT of latest standards and new SMB service capabilities through G (Q, P) G.8032 programmable hardware, ensuring future-proof 1G investment protection MTU (Q, P, H) Figure 1: 5410 Ethernet aggregation The 5410 is based on Ciena’s field-proven technology, deployed by dozens of network operators in tens of thousands of PoPs, central offices, businesses, office parks, cell sites, remote terminals, and residences. It combines the low cost and high capacity of Ethernet with the reliability, management, and service quality usually associated with SONET/SDH networking systems. The 5410 software is based on an advanced service- aware operating system—used in all Ciena service delivery and aggregation switches— which provides consistent system and service attributes and a common operator interface, provisioning, and management capabilities. These features leverage customer investment in training and operational procedures, provide significant operational cost savings, enhance service creation velocity for faster time to revenue, and deliver synergy with Ciena’s best-in-class portfolio of right-sized service aggregation and service delivery Carrier Ethernet systems. Advanced MPLS Functionality The 5410 offers a comprehensive set of MPLS, VPLS, and H-VPLS features that support resilient MPLS L2VPNs and enable service providers to offer MPLS-based connectivity 5410 chassis, front view services on metro Ethernet networks, extending the functionality and scalability of MPLS core networks. These MPLS features include: > MPLS Pseudowire Emulation Edge-to-Edge (PWE3), which supports MPLS Virtual Private Wire Services (VPWS), sometimes called MPLS Virtual Leased Lines. VPWS enables the provisioning of MPLS-based Ethernet Virtual Circuits (EVCs) and supports MPLS-based Ethernet Private Line and Ethernet Virtual Private Line services 2 > Virtual Private LAN Services (VPLS) and Hierarchical-VPLS > MPLS FRR further enhances resiliency for MPLS/VPLS services (H-VPLS) based on MPLS Label Switched Paths (LSPs) and by enabling LSPs to be restored within 50ms in case of MPLS Virtual Circuits (VCs). 5410 VPLS capabilities support physical or logical link failure Layer 2 VPNs that are based on Ethernet transport with the > MPLS OAM capabilities include LSP Ping and LSP Traceroute, benefits of an MPLS/VPLS control plane. This control plane which can be used to verify the LSP status and topology includes Label Distribution Protocol (LDP) for VPLS VC signaling; OSPF-TE and IS-IS-TE for MPLS Tunnel Routes; The 5410 incorporates the latest innovations in Ethernet and RSVP-TE for the establishment of the LSPs that switching technology and control plane protocols and utilizes constitute the VPLS high-performance programmable processors to deliver new features and functionality through simplified software > MPLS label edge router functionality, which enables the 5410 upgrades, with minimal disruption to customer traffic. to function as a VPLS/H-VPLS Provider Edge switch (PE-rs). This programmable architecture keeps capital and operating The 5410 also supports H-VPLS Dual Homing, which provides costs to a minimum and simplifies the migration to emerging resiliency by enabling spoke connections from an MTU-s to standards and services—maximizing revenue potential and two different PE Hub switches. For additional flexibility, providing superior flexibility and investment protection. H-VPLS spoke connections can be implemented using MPLS pseudowires (VCs), Q-in-Q Ethernet VLAN Virtual Circuits, The 5410 provides interworking between Q-in-Q VLANs, Ethernet Private Lines, Ethernet Virtual Private Lines and MPLS/VPLS/H-VPLS, and PBB/PBB-TE encapsulation formats. PBB-TE service instances By enabling Ethernet and MPLS/VPN and connectivity services to seamlessly traverse multiple domains and transmission > MPLS tunnel groups can be configured on the 5410 to facilities—each with its own encapsulation and transport— support LSP redundancy Ciena’s 5410 creates interworking functionality that supports complete service flexibility and maximum optimization of network resources and traffic engineering architectures. Technical Information IEEE 802.3u Fast Ethernet MEF 9 Ethernet Virtual Private Line Interfaces IEEE 802.3z GbE MEF 9 Ethernet Virtual Private LAN 10 x 10GbE XFP NNI/UNI ports per 10GbE IEEE 802.1D MAC Bridges MEF 9 Ethernet Tree Line Module IEEE 802.1Q VLANs - Including .1p Priority MEF 14 Traffic Management - 0–10 10GbE line modules per system IEEE 802.1ad Provider Bridging (Q-in-Q) - 0-100 x 10GbE ports per system VLANs with full S-VLAN range
Recommended publications
  • Ciena 5305 Service Aggregation Switch Datasheet

    Ciena 5305 Service Aggregation Switch Datasheet

    5305 SERVICE AGGREGATION SWITCH Features and Benefits The 5305 Ethernet/MPLS service aggregation > Features advanced Ethernet and MPLS to support demanding business, mobile switch is purpose-built for Carrier Ethernet backhaul, transport, and residential applications including L2VPN service to deliver cost-effective capacity, scalability, delivery and aggregation, 3G/4G wireless backhaul, FTTx and IP DSLAM and resiliency. With this switch, service Aggregation and L2 backhaul of L3VPNs > Delivers optimal density and service providers can keep pace with the constantly flexibility with a compact modular chassis, supporting incremental increasing demand for bandwidth and next- expansion of service and bandwidth capacity with linear CAPEX outlay generation services that support business, > Supports tens of thousands of services mobile backhaul, transport, and residential on a single system with robust scalability of up to 30,000+ VLANs and applications in metro networks. two million MAC addresses per chassis > Delivers high reliability, five-9s The 5305 is a modular, chassis-based system optimized for metro-edge deployments availability, and 50 ms protection in a wide variety of network topologies, including fiber and microwave rings, point-to- switching resiliency using state-of- the-art hardware and software design point fiber, microwave mesh, and fiber or copper to the subscriber. The switch coupled with advanced control plane supports high-density Gigabit Ethernet (GbE) connectivity to the subscriber edge and and Ethernet OAM capabilities
  • Data Center Ethernet 2

    Data Center Ethernet 2

    DataData CenterCenter EthernetEthernet Raj Jain Washington University in Saint Louis Saint Louis, MO 63130 [email protected] These slides and audio/video recordings of this class lecture are at: http://www.cse.wustl.edu/~jain/cse570-15/ Washington University in St. Louis http://www.cse.wustl.edu/~jain/cse570-15/ ©2015 Raj Jain 4-1 OverviewOverview 1. Residential vs. Data Center Ethernet 2. Review of Ethernet Addresses, devices, speeds, algorithms 3. Enhancements to Spanning Tree Protocol 4. Virtual LANs 5. Data Center Bridging Extensions Washington University in St. Louis http://www.cse.wustl.edu/~jain/cse570-15/ ©2015 Raj Jain 4-2 Quiz:Quiz: TrueTrue oror False?False? Which of the following statements are generally true? T F p p Ethernet is a local area network (Local < 2km) p p Token ring, Token Bus, and CSMA/CD are the three most common LAN access methods. p p Ethernet uses CSMA/CD. p p Ethernet bridges use spanning tree for packet forwarding. p p Ethernet frames are 1518 bytes. p p Ethernet does not provide any delay guarantees. p p Ethernet has no congestion control. p p Ethernet has strict priorities. Washington University in St. Louis http://www.cse.wustl.edu/~jain/cse570-15/ ©2015 Raj Jain 4-3 ResidentialResidential vs.vs. DataData CenterCenter EthernetEthernet Residential Data Center Distance: up to 200m r No limit Scale: Few MAC addresses r Millions of MAC Addresses 4096 VLANs r Millions of VLANs Q-in-Q Protection: Spanning tree r Rapid spanning tree, … (Gives 1s, need 50ms) Path determined by r Traffic engineered path spanning tree Simple service r Service Level Agreement.
  • Converged Networking in the Data Center

    Converged Networking in the Data Center

    Converged Networking in the Data Center Peter P. Waskiewicz Jr. LAN Access Division, Intel Corp. [email protected] Abstract data center as a whole. In addition to the general power and cooling costs, other areas of focus are the physical The networking world in Linux has undergone some sig- amount of servers and their associated cabling that re- nificant changes in the past two years. With the expan- side in a typical data center. Servers very often have sion of multiqueue networking, coupled with the grow- multiple network connections to various network seg- ing abundance of multi-core computers with 10 Gigabit ments, plus they’re usually connected to a SAN: ei- Ethernet, the concept of efficiently converging different ther a Fiber Channel fabric or an iSCSI infrastructure. network flows becomes a real possibility. These multiple network and SAN connections mean large amounts of cabling being laid down to attach a This paper presents the concepts behind network con- server. Converged Networking takes a 10GbE device vergence. Using the IEEE 802.1Qaz Priority Group- that is capable of Data Center Bridging in hardware, ing and Data Center Bridging concepts to group mul- and consolidates all of those network connections and tiple traffic flows, this paper will demonstrate how dif- SAN connections into a single, physical device and ca- ferent types of traffic, such as storage and LAN traf- ble. The rest of this paper will illustrate the different fic, can efficiently coexist on the same physical connec- aspects of Data Center Bridging, which is the network- tion.
  • Interactions Between TCP and Ethernet Flow Control Over Netgear

    Interactions Between TCP and Ethernet Flow Control Over Netgear

    Interactions between TCP and Ethernet flow control over Netgear XAVB2001 HomePlug AV links Radika Veera Valli, Grenville Armitage, Jason But, Thuy Nguyen Centre for Advanced Internet Architectures, Technical Report 130121A Swinburne University of Technology Melbourne, Australia [email protected], [email protected], [email protected], [email protected] Abstract—HomePlug AV links are usually slower than buffers along network paths. A side-effect of such loss- the surrounding wired LAN links, creating a bottleneck based CC behaviour is that all traffic sharing the bottle- where traffic may be buffered and delayed. To investigate neck will experience additional latency due to increased the interactions between TCP congestion control and average queueing delays [5]. This is particularly prob- Ethernet flow control we trialled four TCP variants over a lematic for real-time application flows (such as VoIP and HomePlug AV link using two Netgear XAVB2001 devices. We observed that the XAVB2001’s use of Ethernet flow online games) given the growth of arguably-gratuitous control effectively concatenates the buffers in both the buffering (“buffer bloat”) in network devices, interface XAVB2001 and the directly attached sending host. This led cards and network software stacks [6]. to multi-second RTTs when using NewReno and CUBIC In this report we borrow from, and extend, previous (loss-based) TCPs, which is significantly troublesome for work by one of the current authors [7]. Our new work multimedia traffic on home LANs. In contrast, Vegas explores in greater detail the latency and flow control and CDG (delay-based) TCPs kept buffer utilisation low, issues that emerge when end-to-end TCP is allowed to and induced five to ten milliseconds RTT.
  • Chapter 2 Link Layer Verilog Hardware Implementation, and One Wireless Broadcast Link, I.E

    Chapter 2 Link Layer Verilog Hardware Implementation, and One Wireless Broadcast Link, I.E

    Modern Computer Networks: An open source approach Chapter 2 Modern Computer Networks: An open source approach Chapter 2 package, one wired broadcast link, i.e. Ethernet in Section 2.4 along with its Chapter 2 Link Layer Verilog hardware implementation, and one wireless broadcast link, i.e. wireless LAN (WLAN) in Section 2.5 plus a brief on Bluetooth and WiMAX. PPP is Problem Statement popularly used in the last-mile dial-up services or routers carrying various network protocols over point-to-point links. Ethernet has occupied more than 95 percent of To effectively and efficiently transmit data over physical links from one node wired LANs. It is also poised to be ubiquitous in MANs and WANs. In contrast to to one or more nodes, there is much more to do than simply modulating or desktop PCs, which usually use wired links, many devices such as laptop PCs encoding bit stream into signal. Transmission impairments, such as crosstalk and cellular phones are mobile and prefer wireless links such as WLAN, Bluetooth, between two adjacent pairs, can unexpectedly change transmission signal and and WiMAX. thus result in errors. The transmitter might transmit faster than the receiver can Table 2.1 Link protocols. handle. The transmitter has to somehow indicate the destination(s), if on a PAN/LAN MAN/WAN broadcast link, i.e. LAN, and usually needs to name itself to let the receiver know Token Bus (802.4) DQDB (802.6) Token Ring (802.5) HDLC where the source is. If multiple stations share a LAN, an arbitration mechanism is HIPPI X.25 required to determine who can transmit next.
  • Understanding Ethernet

    Understanding Ethernet

    Chapter 1 Understanding Ethernet In This Chapter ▶ Exploring carrier sensing and collision detection methods ▶ Understanding directional traffic and simultaneous transmissions ▶ Examining Ethernet speed limits and wire-rated standards ▶ Exploring Gigabit Ethernet ▶ Distinguishing between local and backbone Ethernet ▶ Recognizing Ethernet interfaces ▶ Understanding Ethernet’s simplicity n today’s connected business world, ubiquitous access to Idigitized information and data is critical. Networking has not only reshaped the information landscape, but it has also vastly increased speeds at which information moves and is consumed. In fact, networking is the key to accessing the com- plex services and applications within which information plays a starring role. Although access to the Internet is essential, all networking begins at the local level. For modern networks, Ethernet is the standard infrastructure upon which local net- works rest. Ethernet comprises a family of frame-based protocols and tech- nologies designed to connect local area networks (LANs) — computersCOPYRIGHTED and devices situated in MATERIAL close proximity and able to communicate with one another. Ethernet draws its name from the fanciful Latin terminology luminiferous aether, which trans- lates to a light-bearing medium once thought to fill space and propagate magnetic waves. As such, Ethernet is a metaphorical reference for the physical transmission medium in the work- place (and at home) that propagates digitized information in electronic form. The term “Ethernet” was originally coined by 4 Carrier Ethernet For Dummies Bob Metcalfe while jointly developing basic Ethernet network computing with David Boggs at Xerox Palo Alto Research Center (PARC). Sensing a Carrier and Handling Collisions Two people on opposite ends of a phone conversation can sense carrier presence (either a dial-tone or a connected call) and handle collisions (overlapping conversations).
  • Storage and Network Convergence Using Fcoe and Iscsi

    Storage and Network Convergence Using Fcoe and Iscsi

    Front cover Storage and Network Convergence Using FCoE and iSCSI Learn how to improve IT service performance and availability Simplify your storage and network infrastructure See how to reduce data center network costs Sangam Racherla Silvio Erdenberger Harish Rajagopal Kai Ruth ibm.com/redbooks International Technical Support Organization Storage and Network Convergence Using FCoE and iSCSI January 2014 SG24-7986-01 Note: Before using this information and the product it supports, read the information in “Notices” on page xi. Second Edition (January 2014) This edition applies to the latest supported Converged Network Adapters and Switches in the IBM System Networking Portfolio of products. © Copyright International Business Machines Corporation 2012, 2014. All rights reserved. Note to U.S. Government Users Restricted Rights -- Use, duplication or disclosure restricted by GSA ADP Schedule Contract with IBM Corp. Contents Notices . xi Trademarks . xii Preface . xiii Authors. xiii Now you can become a published author, too! . .xv Comments welcome. .xv Stay connected to IBM Redbooks . xvi Part 1. Overview of storage and network convergence . 1 Chapter 1. Introduction to convergence . 3 1.1 What convergence is. 4 1.1.1 Calling it what it is . 4 1.2 Vision of convergence in data centers . 4 1.3 The interest in convergence now . 5 1.4 Fibre Channel SANs today . 5 1.5 Ethernet-based storage today. 6 1.6 Benefits of convergence in storage and network . 7 1.7 Challenge of convergence . 8 1.8 Conclusion . 10 Chapter 2. Fibre Channel over Ethernet . 11 2.1 Background: Data Center Bridging . 12 2.1.1 Priority-based Flow Control: IEEE 802.1Qbb .
  • TROUBLESHOOTING FEATURES for Media Conversion Products Table of Contents

    TROUBLESHOOTING FEATURES for Media Conversion Products Table of Contents

    WHITE PAPER : TROUBLESHOOTING FEATURES for Media Conversion Products Table of Contents Introduction ......................................................................................................... 3 Link Integrity ....................................................................................................... 4 Ethernet Troubleshooting Features ................................................................... 4 1. Auto Negotiation (AN) ....................................................................................... 4 a. Full Duplex ................................................................................................ 4 b. 100 Mbps Speed ..................................................................................... 4 c. Selective Advertising ................................................................................... 5 d. Rate Converters ....................................................................................... 5 e. FX Auto Negotiation (FX AN) ..................................................................... 5 2. FX LinkLoss (FXLL) ........................................................................................... 6 3. TX LinkLoss (TXLL) ........................................................................................... 6 4. FiberAlert (FA) ................................................................................................... 7 5. LinkLoss & FiberAlert ........................................................................................ 7
  • Fairness in a Data Center

    Fairness in a Data Center

    University of New Hampshire University of New Hampshire Scholars' Repository Doctoral Dissertations Student Scholarship Winter 2012 Fairness in a data center MIkkel Hagen University of New Hampshire, Durham Follow this and additional works at: https://scholars.unh.edu/dissertation Recommended Citation Hagen, MIkkel, "Fairness in a data center" (2012). Doctoral Dissertations. 694. https://scholars.unh.edu/dissertation/694 This Dissertation is brought to you for free and open access by the Student Scholarship at University of New Hampshire Scholars' Repository. It has been accepted for inclusion in Doctoral Dissertations by an authorized administrator of University of New Hampshire Scholars' Repository. For more information, please contact [email protected]. Fairness in a data center BY Mikkel Hagen M.S. Computer Science, University of New Hampshire (2008) B.S. Computer Science/Biology/Psychology, St. Bonaventure University (2004) DISSERTATION Submitted to the University of New Hampshire in partial fulfillment of the requirements for the degree of Doctor of Philosophy in Computer Science December 2012 UMI Number: 3537816 All rights reserved INFORMATION TO ALL USERS The quality of this reproduction is dependent upon the quality of the copy submitted. In the unlikely event that the author did not send a complete manuscript and there are missing pages, these will be noted. Also, if material had to be removed, a note will indicate the deletion. Dissertation Publishing UMI 3537816 Published by ProQuest LLC 2013. Copyright in the Dissertation held by the Author. Microform Edition © ProQuest LLC. All rights reserved. This work is protected against unauthorized copying under Title 17, United States Code. uest ProQuest LLC 789 East Eisenhower Parkway P.O.
  • 19540 - Telematics 7Th Tutorial - Media Access, Ethernet & Wireshark

    19540 - Telematics 7Th Tutorial - Media Access, Ethernet & Wireshark

    19540 - Telematics 7th Tutorial - Media Access, Ethernet & Wireshark Bastian Blywis Department of Mathematics and Computer Science Institute of Computer Science 26. November, 2009 Institute of Computer Science – Telematics Tutorial – 26. November, 2009 1 Outline 1. Frame Size 2. Packets and Cells 3. FDDI Performance 4. ATM 5. Tunneling 6. Bridge Classification 7. Virtual LANs 8. IEEE 802.1Q and IEEE 802.2 9. Features of Layer 2 Protocols 10. Security Considerations 11. IEEE 802.1D Institute of Computer Science – Telematics Tutorial – 26. November, 2009 2 Frame Size Consider a 10 MBit/s CSMA/CD LAN with a bus of 50m length. The speed of the signal within the transmission medium is 2 ∗ 108m=s. 1. Calculate the upper bound of the collision detection time. 2. Specify the minimum frame length. Institute of Computer Science – Telematics Tutorial – 26. November, 2009 3 Frame Size Maximum time depends on maximum distance: AB 50m 5*10^-7s – Worst case: station A senses a free medium and starts to transmit – Station B senses a free medium and starts to transmit just at the time the signal from A arrives at B – Signal from B has to travel through the whole network until A detects the collision – Maximum time until collision detection, is twice the time of the signal propagation time for the whole network buslength 50m t = 2 ∗ = 2 ∗ = 5 ∗ 10−7s 8 m signalspeed 2 ∗ 10 s Institute of Computer Science – Telematics Tutorial – 26. November, 2009 4 Frame Size To ensure the stations are able to detect a collision, the frame has be take at least t to send: framelength min > t capacity 10MBit framelength > ∗ 5 ∗ 10−7s = 5Bit min s Institute of Computer Science – Telematics Tutorial – 26.
  • Network Transport for Data Centers

    Network Transport for Data Centers

    X_405082 Advanced Computer Networks Data Center Transport Lin Wang ([email protected]) Period 2, Fall 2020 Course outline Warm-up Video ■ Fundamentals ■ Video streaming ■ Forwarding and routing ■ Video stream analytics ■ Network transport Networking and ML Data centers ■ Networking for ML ■ Data center networking ■ ML for networking ■ Data center transport � Mobile computing Programmability ■ Wireless and mobile ■ Sofware defined networking ■ Programmable forwarding 2 Learning objectives What are the new challenges in data center transport? What design choices do we have for data centers transport design? 3 What is special about data center transport Network ■ Extremely high speed (100+ Gbps) ■ Extremely low latency (10-100s of us) Diverse applications and workloads ■ Large variety in performance requirements Traffic patterns ■ Large long-lived flows vs small short-lived flows ■ Scatter-gather, broadcast, multicast Built out of commodity components: no expensive/customized hardware 4 Congestion control recall Network Do you still remember the goal of congestion control? 5 Congestion control recall Network Congestion control aims to determine the rate to send data on a connection, such that (1) the sender does not overrun the network capability and (2) the network is efficiently utilized 6 TCP Sender Receiver Network ACK ACK ACK Application The transport layer in the network model: Reliable TCP ■ Reliable, in-order delivery using acknowledges ■ Make sure not to overrun the receiver (receiving window, IP rwnd) and the network (congestion window,
  • NC-SI Over MCTP Binding Specification 6

    NC-SI Over MCTP Binding Specification 6

    1 2 Document Number: DSP0261 3 Date: 2013-08-22 4 Version: 1.0.0 5 NC-SI over MCTP Binding Specification 6 7 Document Type: Specification 8 Document Status: DMTF Standard 9 Document Language: en-US 10 11 NC-SI over MCTP Binding Specification DSP0261 12 Copyright notice 13 Copyright © 2013 Distributed Management Task Force, Inc. (DMTF). All rights reserved. 14 DMTF is a not-for-profit association of industry members dedicated to promoting enterprise and systems 15 management and interoperability. Members and non-members may reproduce DMTF specifications and 16 documents for uses consistent with this purpose, provided that correct attribution is given. As DMTF 17 specifications may be revised from time to time, the particular version and release date should always be 18 noted. 19 Implementation of certain elements of this standard or proposed standard may be subject to third party 20 patent rights, including provisional patent rights (herein "patent rights"). DMTF makes no representations 21 to users of the standard as to the existence of such rights, and is not responsible to recognize, disclose, 22 or identify any or all such third party patent right, owners or claimants, nor for any incomplete or 23 inaccurate identification or disclosure of such rights, owners or claimants. DMTF shall have no liability to 24 any party, in any manner or circumstance, under any legal theory whatsoever, for failure to recognize, 25 disclose, or identify any such third party patent rights, or for such party’s reliance on the standard or 26 incorporation thereof in its product, protocols or testing procedures.