www.archivc.org
`41 5.561.6767
`41 5.840-0391 c-fax
`
`I nrcm cr /\ rchivc
`300 Funston Avenue
`San Francisco, CA 941 18
`
`AFFIDAVIT OF CHRISTOPHER BUTLER
`
`I. I am the Office Manager at the Internet Archive, located in San Francisco,
`California. 1 make this declaration of my own personal knowledge.
`2. The Internet Archive is a website that provides access to a digital library of
`Internet sites and other cultural artifacts in digital form. Like a paper library, we provide
`free access to researchers, historians, scholars, and the general public. The Internet
`Archive has partnered with and receives support from various institutions, including the
`Library of Congress.
`3. The Internet Archive has created a service known as the Way back Machine. The
`Wayback Machine makes it possible to surf more than 450 bill ion pages stored in the
`Internet Archive's web archive. Visitors to the Wayback Mach ine can search archives
`by URL (i.e., a website address). If archived records for a URL are available, the visitor
`will be presented with a list of available dates. The visitor may select one of those
`dates, and then begin surfing on an archived version of the Web. The links on the
`archived files, when served by the Wayback Machine, point to other archived files
`(whether HTML pages or images). If a visitor clicks on a link on an archived page, the
`Wayback Machine will serve the archived file with the closest available date to the page
`upon which the link appeared and was clicked.
`4. The archived data made viewable and browseable by the Wayback Machine is
`compi led using software programs known as crawlers, which surf the Web and
`automatically store copies of web tiles, preserving these files as they exist at the point of
`time of capture.
`5. The Internet Archive assigns a URL on its site to the archived files in the format
`http://web.archive.org/web/[Year in yyyy][Month in mm][Day in dd][Time code in
`hh:mm:ss]/[Archived URL]. Thus, the Internet Archive URL
`http://web.archive.org/web/19970 I 26045828/http://www.archive.org/ would be the
`URL for the record of the Internet Archive home page HTML file
`(http://www.archive.org/) archived on January 26, 1997 at 4:58 a.m. and 28 seconds
`( 1997/01126 at 04:58:28). A web browser may be set such that a printout from it will
`display the URL of a web page in the printout's footer. The date assigned by the Internet
`Archive applies to the HTML tile but not to image files linked therein. Thus images that
`appear on a page may not have been archived on the same date as the HTML file.
`Likewise, if a website is designed with "frames," the date assigned by the Internet
`Archive applies to the frameset as a whole, and not the individual pages within each
`frame.
`6. Attached hereto as Exhibit A are true and accurate copies of printouts of the
`Internet Archive's records of the .html files, .pdffiles, or .ps files for the URLs and the
`dates specified in the footer of the printout (HTML) or attached coversheet (PDF or PS).
`7. I declare under penalty of perjury that the foregoing is true and correct.
`
`DATE:
`
`INTEL Ex.1087.001
`
`

`

`CALIFORNIA JURA T
`
`See Attached Document.
`
`State of California
`County of San Francisco
`
`A notary public or other officer completing this
`certificate verifies only the identity of the
`individual who signed the document to which this
`certificate is attached, and not the truthfulness,
`accuracy, or validity of that document.
`
`Subscribed and sworn to (or affirmed) before me on
`this
`ZC"" day of 1,.,,.,14']
`Christopher Butler,
`
`' lt? 11- ' by
`
`proved to me on the basis of satisfactory evidence to be
`the person who appeared before me.
`
`SignaturecZ:j /L ft, _ _
`
`INTEL Ex.1087.002
`
`

`

`
`
`
`
`
`
`Exhibit A
`
`Exhibit A
`
`INTEL Ex.1087.003
`
`CAVIUM-1087
`Cavium, Inc. v. Alacritech, Inc.
`Page 003
`
`

`

`https://web.archive.org/web/19970113131433/http://www.alteon.com/techbrief
`.ps
`
`INTEL Ex.1087.004
`
`

`

`Gigabit Ethernet
`Technical Brief
`Achieving End-to-End Performance
`
`Alteon Networks, Inc.
`6351 San Ignacio Avenue
`San Jose, CA 95119
`1-408-574-5500
`
`First Edition
`September 1996
`
`INTEL Ex.1087.005
`
`

`

`© 1996 by Alteon Networks, Inc. All rights reserved.
`
`ii
`
`INTEL Ex.1087.006
`
`

`

`Contents
`
` Gigabit Ethernet Technical Brief
`Achieving End-to-End Performance
`Goals of Gigabit Ethernet 1
`Uses of Gigabit Ethernet 2
`Gigabit Ethernet History 3
`Gigabit Ethernet Momentum 4
`Workstation Speed 4
`Desktop Multimedia 5
`Multivendor Interoperability 6
`Migrating to Gigabit Ethernet 6
`Gigabit Ethernet Components 6
`Upgrade Scenarios 6
`Upgrading Connections to Centralized File and Compute Servers 7
`Upgrading Connections between Switches 8
`Upgrading a Switched Fast Ethernet Backbone 8
`Upgrading a Shared FDDI Backbone 9
`Protocol Architecture 10
`Physical Interface Characteristics 11
`Serializer/Deserializer 11
`8B/10B Encoding 11
`MAC Layer 11
`Upper Layers 11
`Cabling types and distances 12
`Flow Control 12
`Full-Duplex Transmission 12
`Half-duplex Transmission 13
`Carrier Extension 14
`Technology Advances 15
`Ethernet Adapters 15
`First Generation Ethernet Adapters 16
`Second Generation Ethernet Adapters 17
`Third Generation Ethernet Adapters—The Alteon Difference 18
`Ethernet Switches 20
`Specialized ASICs to meet the Performance Demands of Gigabit Ethernet 21
`Seamless Integration with Ethernet and Fast Ethernet 21
`Standard Ethernet Management 21
`Conclusion 22
`
`iii
`
`INTEL Ex.1087.007
`
`

`

`iv
`
`®Alteon Networks, Inc. 1996
`
`INTEL Ex.1087.008
`
`

`

` Gigabit Ethernet Technical Brief
`Achieving End-to-End Performance
`
`Intranet Web servers, centralized file and compute servers, data warehousing, groupware, medical
`imaging, CAD/CAM applications, 3-D modeling, animation, video, pre-press applications, server
`farms, seismic processing...
`The list goes on and on. The demand for high-speed network connections is proliferating at a pace
`almost as rapid as the speed requirements of the applications themselves. Evidence is everywhere:
`the rapid acceptance of 10/100 Mbps connections on today’s desktop computers, Ethernet
`switching at the department level, and the deployment of Fast Ethernet switches in corporate
`backbones are a few examples of the need for faster and faster networks.
`And still, bottlenecks remain. Server network connections have been limited to 100 Mbps since
`FDDI was shipping in volume in the late 1980’s. Fast Ethernet made it easier to build
`internetworking products, but did not provide a faster server interface. Today, centralized servers
`are often configured with multiple 100 Mbps network connections to meet bandwidth
`requirements.
`Enter Gigabit Ethernet.
`Gigabit Ethernet is a new technology that will provide seamless interoperability with Ethernet and
`Fast Ethernet. Gigabit Ethernet transfers data at a blazingly fast speed: one gigabit per second, or
`100 times the rate of standard Ethernet. Gigabit Ethernet is designed to deliver the same benefits
`as Fast Ethernet: seamless integration with installed Ethernets, dramatically higher performance
`than the previous standard, and a familiar management environment.
`It couldn’t happen at a better time. Multimedia over IP is just starting to take off with the proposed
`IETF standards, real-time transfer protocol (RTP) and resource reservation protocol (RSVP).
`These protocols, combined with gigabit networks, high-performance desktops, and Internet
`technology, will quickly change the way corporations access information.
`Ethernet. Fast Ethernet. Gigabit Ethernet. Networking made simple!
`Goals of Gigabit Ethernet
`Under development in the IEEE by the 802.3z Task Force, Gigabit Ethernet has the following
`primary goals:
`• Complete interoperability with Ethernet and Fast Ethernet
`– Retain the installed base of NICs
`– Leverage the investment in hubs, switches, and routers
`– Leverage the network management environment
`
`1
`
`INTEL Ex.1087.009
`
`

`

`Gigabit Ethernet Technical Brief
`• Conform to the Ethernet standard
`– Frame format
`– Minimum and maximum frame sizes
`– CSMA/CD access method
`– 802.2 LLC specifications
`• Provide a simple forwarding mechanism between 10, 100, and 1000 Mbps
`– No fragmentation of frames
`– No encapsulation of frames
`– No translation of frames
`• Provide 10 times the performance of Fast Ethernet
`Uses of Gigabit Ethernet
`Initially, Gigabit Ethernet will be deployed in three key areas:
`• Aggregating traffic between Ethernet clients and centralized file or compute servers
`• Connecting multiple 100Base-T Fast Ethernet switches through 100/1000 Mbps switches
`• Connecting both workstations and servers with Gigabit Ethernet to run high-bandwidth
`applications, such as CAD/CAM, medical imaging, and pre-press
`Figure 1 illustrates the projected initial uses of Gigabit Ethernet.
`
`High-Performance
`1000Base-X Server for
`Existing Ethernet Network
`PCI
`
`1000
`
`Backbone
`for 100Base-X
`Devices
`
`1000
`
`100
`
`100
`
`100
`
`100
`
`PCI
`
`Traditional
`High-Bandwidth
`Applications
`
`10 Mbps
`10 Mbps
`100 Mbps
`10 Mbps
`Figure 1. Initial deployment of Gigabit Ethernet
`
`100 Mbps
`
`In the future, the uses of Gigabit Ethernet will be extended to incorporate switches based entirely
`on Gigabit Ethernet on the campus backbone, centralized server farms connected to these
`all-gigabit switches, and specialized high-performance client workstations. Figure 2 illustrates
`some future Gigabit Ethernet possibilities.
`
`2
`
`®Alteon Networks, Inc. 1996
`
`INTEL Ex.1087.010
`
`

`

`Gigabit Ethernet Technical Brief
`
`Ethernet
`Campus
`Backbone
`
`1000
`
`Backbone
`for 100Base-X
`Devices
`
`High-Performance
`1000Base-X Server for
`Existing Ethernet Clients
`PCI
`
`1000
`
`100
`
`100
`
`100
`
`1000
`
`PCI
`PCI
`
`Server Farm
`with Compute and
`File Servers
`
`1000
`
`PCI
`Specialized
`High-Performance
`Clients
`
`100
`
`10 Mbps
`10 Mbps
`100 Mbps
`10 Mbps
`Figure 2. Future deployment of Gigabit Ethernet
`
`100 Mbps
`
`Gigabit Ethernet History
`In November of 1995, the IEEE 802.3 committee formed a High-Speed Study Group to examine
`the future direction of high-speed networking. The IEEE approved the 802.3z Task Force, focused
`exclusively on Gigabit Ethernet, in March of 1996. The IEEE 802.3z Task Force will develop the
`standard for Gigabit Ethernet.
`The Gigabit Ethernet Alliance (GEA) was formed in May of 1996. Modeling itself after the Fast
`Ethernet Alliance, the GEA is an open forum whose goal is to support the development,
`standardization, and deployment of Gigabit Ethernet.
`The primary goals of the Gigabit Ethernet Alliance are:
`• Fully support the Gigabit Ethernet standards activities of the IEEE 802.3z working group
`• Ensure product interoperability
`• Provide a mechanism for Gigabit Ethernet suppliers and consumers to communicate
`•
`Increase its membership
`More than 60 companies, including some of the biggest networking companies in the industry,
`participate in regular meetings of the GEA. The GEA home page is located on the World Wide
`Web at http://www.gigabit-ethernet.org.
`
`3
`
`INTEL Ex.1087.011
`
`

`

`Gigabit Ethernet Technical Brief
`Gigabit Ethernet Momentum
`There are a number of factors driving the development of Gigabit Ethernet. These include the
`increased speed of workstations using the PCI bus, the need for multimedia on the desktop, and
`corporate demands for multivendor interoperability.
`
`Workstation Speed
`With the advent of the PCI bus and high-performance CPUs, workstations are getting faster and
`faster. In fact, PCI-bus equipped workstations have more than enough compute power to handle
`data transfers at gigabit speeds. Table 1 shows the theoretical upper limit of throughput for four
`bus types that have been used in standard workstations over the years.
`
`Table 1. Workstation Throughput
`Bus Type
`Throughput
`ISA
`64 Mbps
`EISA
`264 Mbps
`MCA
`320 Mbps
`PCI (32 bit)
`1,056 Mbps
`PCI (64 bit)
`2,112 Mbps
`
`Faster bus speed is one important factor in achieving gigabit-speed readiness for standard
`workstations. In addition to having a faster bus, the interaction between the host workstation and
`the NIC must be optimized. The graph shown in Figure 3 illustrates the impact of greater
`throughput on workstation host CPU utilization.
`
`1000
`
`800
`
`600
`
`400
`
`200
`
`Throughput (Mbps)
`
`0
`
`0
`
`10Base-T
`NICs
`
`20
`
`40
`60
`% CPU Utilization
`Figure 3. Host throughput versus CPU utilization
`
`pti m iz e d A d a pters
`
`a n c e O
`
`P erfor m
`
`Existing Adapters
`100Base-T NICs
`100
`
`80
`
`4
`
`®Alteon Networks, Inc. 1996
`
`INTEL Ex.1087.012
`
`

`

`Gigabit Ethernet Technical Brief
`Using current NIC technology, as throughput approaches 100 Mbps, CPU utilization quickly
`approaches 100%. In order to support data transfers at gigabit speeds, the design of NICs cannot
`simply be scaled based on existing Ethernet or Fast Ethernet cards. Instead, a new, optimized
`method of transferring data between the network and the host must be implemented. Later in this
`paper, we will discuss how Alteon Networks is creating performance optimized adapters to answer
`the demands of Gigabit Ethernet.
`
`Desktop Multimedia
`Up until now, desktop multimedia applications have not experienced the explosion of use that has
`been predicted for the past few years. The reason for the sluggish implementation of desktop
`multimedia is simple: there has not been a combination of technologies available within most
`network managers’ budgets to support it.
`Multimedia applications need a mixture of ultra fast networking technology and an accompanying
`suite of protocols designed to simultaneously handle bursty data traffic and streaming voice/video
`traffic.
`The combination of Gigabit Ethernet technology and a new set of standard IP-based protocols is
`paving the way for desktop multimedia to become an affordable reality.
`The Internet Engineering Task Force (IETF) has standardized three new protocols that work
`together to provide desktop multimedia:
`• Real-time transfer protocol (RTP)
`• Real-time transfer control protocol (RTCP)
`• Resource reservation protocol (RSVP)
`RTP is a streaming-oriented protocol that works with the TCP/IP protocol suite to provide
`end-to-end delivery of multimedia traffic. RTP adds a 10-byte header to each packet. The header
`contains a time-stamp and sequencing information that allows time-sensitive data to be
`synchronized and reassembled in the correct order.
`RTCP is a control protocol that is used in conjunction with RTP. RTCP is responsible for
`providing feedback on the quality of network conditions, and allows applications to adapt to those
`conditions.
`RSVP is a protocol used for reserving network resources, such as buffers or bandwidth, for a
`session. It works by allowing applications to submit a bandwidth reservation request to the host
`PC. The host PC then requests a specific quality of service from RSVP-equipped routers along the
`data path, providing end-to-end allocation of network resources.
`These three protocols are leading the way for true desktop multimedia to become a reality in the
`very near future. Gigabit Ethernet is poised to handle the very large bandwidth requirements that
`desktop multimedia will require.
`
`5
`
`INTEL Ex.1087.013
`
`

`

`Gigabit Ethernet Technical Brief
`
`Multivendor Interoperability
`One of the key goals of the Gigabit Ethernet Alliance is to provide end-to-end multivendor
`interoperability for Ethernet, Fast Ethernet, and Gigabit Ethernet equipment.
`Initially, multivendor interoperability will be achieved using fiber optic cabling and Fibre Channel
`standards at the lowest levels of the Gigabit Ethernet standard. Standardizing on the use of fiber
`will help spur the early development of interoperable products, without impacting the deployment
`of Gigabit Ethernet into campus backbones and high-performance server environments. Vendors
`are committed to shipping interoperable products by the end of 1996 or early 1997.
`The IEEE 802.3z standard will include specifications for running Gigabit Ethernet on Category 5
`unshielded twisted pair. Due to greater technological challenges, twisted-pair implementations
`will follow the initial deployment of Gigabit Ethernet over fiber. The Task Force is also
`considering a 25 meter connection using twinax or similar cable. The standard will, once again,
`leverage Fibre Channel (FC-0) for this low-cost cabling solution.
`
`Migrating to Gigabit Ethernet
`Gigabit Ethernet is a campus technology, designed initially for use in the campus backbone.
`Gigabit Ethernet will be used between routers, switches, and hubs. It will also be used to connect
`individual servers, server farms, and power workstations requiring more throughput to the
`high-speed backbone. Eventually, Gigabit Ethernet will be used to connect both shared and
`point-to-point workgroups.
`Gigabit Ethernet Components
`To achieve the objectives of gigabit migration, four types of hardware are required:
`• Gigabit Ethernet Network Interface Cards (NICs)
`• Aggregating switches that connect 100Mbps Fast Ethernet and 1000Mbps Gigabit Ethernet
`segments
`• All-Gigabit Ethernet switches
`• All-Gigabit Ethernet Repeaters
`In the next section, we will discuss how to use Gigabit Ethernet components to migrate a variety of
`network designs to Gigabit Ethernet.
`Upgrade Scenarios
`In general, there are five scenarios for upgrading to Gigabit Ethernet:
`• Upgrading connections to centralized file and compute servers
`• Upgrading high performance workstations with Gigabit Ethernet NICs, and connecting them to
`Gigabit Ethernet switches or repeaters
`• Upgrading connections between switches to use 1000 Mbps fat pipes between 100/1000
`switches
`• Upgrading a switched Fast Ethernet backbone by connecting Fast Ethernet switches with a
`Gigabit Ethernet switch or repeater
`
`6
`
`®Alteon Networks, Inc. 1996
`
`INTEL Ex.1087.014
`
`

`

`Gigabit Ethernet Technical Brief
`• Upgrading a shared FDDI backbone by connecting FDDI concentrators or
`Ethernet-to-FDDI routers with Gigabit Ethernet switches or repeaters
`In all of the upgrade scenarios listed above, the network operating system, applications, and NIC
`drivers remain unchanged at the desktop. This allows companies to retain their investment in
`network management applications, tools, network hardware, and software. Existing multi-mode
`fiber runs, originally used for the existing backbone, are leveraged as well.
`
`Upgrading Connections to Centralized File and Compute Servers
`The simplest way to take advantage of the increased speed afforded by Gigabit Ethernet is to
`upgrade one or more Fast Ethernet switches to an aggregating Gigabit Ethernet switch. 10/100
`switches can be upgraded to 100/1000 switches, and high-performance super servers can be
`connected to the high-speed switches by installing Gigabit Ethernet NICs.
`In Figure 4, the illustration on the left shows the original network. Some servers have multiple 100
`Mbps connections to meet the bandwidth requirement of the applications. The illustration on the
`right shows the network using Gigabit Ethernet technology. The servers each have a single Gigabit
`Ethernet connection, simplifying network administration and providing much faster response time.
`
`End User Connections
`
`End User Connections
`
`100 M
`Repeater
`100 Mbps
`
`10 Mbps
`
`10/100 Switches
`10 Mbps
`10 Mbps
`
`100 M
`Repeater
`100 Mbps
`
`10 Mbps
`
`10/100 Switches
`10 Mbps
`10 Mbps
`
`100
`Mbps
`
`100
`Mbps
`
`100
`Mbps
`
`100
`Mbps
`
`100
`Mbps
`
`100
`Mbps
`
`100
`Mbps
`
`100
`Mbps
`
`Fast Ethernet Switch
`
`100 Mbps
`
`Gigabit Ethernet
`Switch or Repeater
`
`1000 Mbps
`
`Server Farm
`
`Server Farm
`New
`Figure 4. Upgrading Connections to Servers and High-Performance Workstations
`
`7
`
`INTEL Ex.1087.015
`
`

`

`Gigabit Ethernet Technical Brief
`Upgrading Connections between Switches
`Another simple way to take advantage of gigabit speeds is to upgrade the 100Mbps links between
`Fast Ethernet switches to 1,000Mbps. Fast Ethernet switches are replaced by 100/1000
`aggregating switches. This allows the 100/1000 aggregating switches to support more switched
`and shared Fast Ethernet segments, without replacing any desktop hardware or software.
`In Figure 5, the illustration on the left shows the original network, the one on the right shows the
`upgraded network.
`
`End User Connections
`100 M
`100 M
`Repeater
`Repeater
`100 Mbps
`100 Mbps
`
`10/100 Switches
`10 Mbps
`
`10/100 Switches
`10 Mbps
`
`10 Mbps
`
`End User Connections
`100 M
`100 M
`Repeater
`Repeater
`100 Mbps
`100 Mbps
`
`10 Mbps
`
`100
`Mbps
`
`100
`Mbps
`
`100
`Mbps
`
`100
`Mbps
`
`Fast Ethernet Switch
`
`100 Mbps
`
`100
`Mbps
`
`Fast Ethernet Switch
`
`100 Mbps
`
`100/1000 Switch
`100 Mbps
`
`1000
`Mbps
`
`100/1000 Switch
`100 Mbps
`
`Server Farm
`
`New
`
`Server Farm
`
`Figure 5. Upgrading Connections between Switches
`
`Upgrading a Switched Fast Ethernet Backbone
`In this scenario, a Fast Ethernet backbone is upgraded to a Gigabit Ethernet switch or repeater. The
`10/100 feeder switches are upgraded to 100/1000 aggregating switches. Once the backbone switch
`is upgraded to 1,000 Mbps, high-performance server farms can be connected directly to it by
`installing Gigabit Ethernet NICs. Additionally, the network can now support a greater number of
`segments, more bandwidth on each segment, and a greater number of nodes per segment.
`In Figure 6, the illustration on the left shows the original network, the one on the right shows the
`network migrated to take advantage of Gigabit Ethernet.
`
`8
`
`®Alteon Networks, Inc. 1996
`
`INTEL Ex.1087.016
`
`

`

`Gigabit Ethernet Technical Brief
`
`End User Connections
`
`10/100 Switches or Routers
`10 Mbps
`
`End User Connections
`
`10/100 Switches or Routers
`10 Mbps
`
`100
`Mbps
`
`Fast Ethernet Switch
`
`100 Mbps Backbone
`
`1000
`Mbps
`
`Gigabit Ethernet
`Switch or Repeater
`1000 Mbps Backbone
`
`New
`Figure 6. Upgrading a Switched Fast Ethernet Backbone
`
`Upgrading a Shared FDDI Backbone
`FDDI backbone networks are usually designed in one of two ways: a centralized backbone
`with FDDI concentrator and routers feeding into it, or a backbone that is distributed among a ring
`of FDDI concentrators or routers.
`To upgrade either of these FDDI backbone designs to Gigabit Ethernet, the FDDI concentrators
`and routers are replaced with a Gigabit Ethernet switch. In the centralized backbone approach, the
`FDDI concentrator is simply replaced with a Gigabit Ethernet switch or repeater, and the routers
`are upgraded with a Gigabit Ethernet connection. Figure 7 shows the before and after illustrations
`of a centralized backbone.
`
`Routers
`
`FDDI
`
`FDDI
`
`FDDI
`
`FDDI
`
`WAN
`
`FDDI
`Concentrator
`FDDI Backbone
`Figure 7. Upgrading a centralized FDDI backbone
`
`Router
`
`WAN
`
`Gigabit
`Ethernet
`Switch or
`Repeater
`
`1000 Mbps
`
`New
`
`To upgrade a distributed backbone, each FDDI concentrator and/or router is replaced with a
`Gigabit Switch. Upgrades can be done on a case by case basis, or all at the same time. If the
`upgrade is accomplished one piece at a time, existing FDDI routers and concentrators will need to
`have a Gigabit Ethernet interface installed in order to connect to the Gigabit Switch. Figure 8
`shows a distributed FDDI backbone upgrade. The one on the left is the original network. The
`network on the right shows the addition of a single Gigabit Ethernet switch.
`
`9
`
`INTEL Ex.1087.017
`
`

`

`Gigabit Ethernet Technical Brief
`
`Shared FDDI
`Backbone
`
`100 Mbps
`
`Routers
`
`FDDI
`Backbone
`
`FDDI
`Backbone
`
`or
`
`WAN
`
`Router
`
`Concentrator
`
`WAN
`
`FDDI
`Concentrator
`
`New
`
`FDDI
`Switch
`
`Gigabit
`Ethernet
`Switch
`1000 Mbps
`
`Figure 8. Upgrading a distributed FDDI backbone
`
`Protocol Architecture
`Gigabit Ethernet uses a mixture of proven protocol technologies adopted from both the original
`IEEE 802.3 Ethernet specification and the ANSI X3T11 Fibre Channel specification. Figure 9
`illustrates the way in which Gigabit Ethernet joins the two standards.
`
`IEEE 802.3
`Ethernet
`
`IEEE 802.2 LLC
`
`IEEE 802.3
`CSMA/CD
`IEEE 802.3
`Physical Layer
`
`FC-4 Upper Layer
`Mapping
`FC-3 Common
`Services
`
`FC-2 Signaling
`
`Upper Layers
`
`IEEE 802.2 LLC
`
`CSMA/CD or
`Full-Duplex MAC
`8B/10B
`Encode/Decode
`Serializer/
`Deserializer
`
`FC-1
`Encode/Decode
`FC-0 Interface
`and Media
`ANSI X3T11
`Fibre Channel
`Figure 9. Gigabit Ethernet Protocol Stack
`
`Connector
`
`IEEE 802.3z
`Gigabit Ethernet
`
`10
`
`®Alteon Networks, Inc. 1996
`
`INTEL Ex.1087.018
`
`

`

`Gigabit Ethernet Technical Brief
`Physical Interface Characteristics
`The lowest layer of the initial fiber version of the Gigabit Ethernet protocol stack uses layer FC-0
`from the Fibre Channel specification. FC-0 defines the physical characteristics of the interface and
`media, including the cables, connectors, drivers, transmitters and receivers. It also describes the
`speed and distance combinations of each media type.
`Serializer/Deserializer
`On top of the Physical (PHY) layer is the serializer/deserializer. The serializer/deserializer is a
`layer that supports multiple encoding schemes–one of which is 8B/10B, as specified in the Fibre
`Channel specification. Encoding is media dependent; 8B/10B encoding was designed for fiber
`optic media. When support of twisted pair cabling is added to the Gigabit Ethernet specification,
`the serializer/deserializer will provide a mechanism for the encoding scheme used for twisted pair
`to work with the Gigabit Ethernet PHY.
`8B/10B Encoding
`8B/10B encoding is another layer adapted from the FC-1 Fibre Channel specification. It describes
`the byte synchronization and the encode/decode scheme. The 8B/10B encoding scheme transmits
`eight bits as a 10-bit code group. 8B/10B encoding features low-cost component design and
`provides good transition density for easy clock recovery.
`MAC Layer
`Gigabit Ethernet supports both full-duplex and half-duplex transmission systems. In full-duplex
`mode, the Gigabit Ethernet Media Access Control (MAC) uses the IEEE 802.3x Full-Duplex
`specification, including IEEE 802.3x frame-based flow control.
`In half-duplex mode, the Gigabit Ethernet MAC supports the Carrier Sense Multiple Access with
`Collision Detection (CSMA/CD) access method, standardized in the original IEEE 802.3 Ethernet
`specification. It is expected that initial uses of Gigabit Ethernet will focus primarily on backbone
`connections using full-duplex transmission. Therefore, the first Gigabit Ethernet products to ship
`will support only full-duplex transmission. Support for half-duplex will be added later.
`Upper Layers
`Above the MAC layer, Gigabit Ethernet is unchanged from the original IEEE 802.3 standard,
`supporting both Ethernet and IEEE 802.2 Logical Link Control (LLC). Support for LLC allows
`multiple upper layer protocols (including TCP/IP, SPX/IPX, and so on) to run over Gigabit
`Ethernet.
`
`11
`
`INTEL Ex.1087.019
`
`

`

`Gigabit Ethernet Technical Brief
`Cabling types and distances
`The IEEE 802.3z Task Force investigated a variety of cabling types and distances to be supported
`by Gigabit Ethernet. Table 2 summarizes the proposed cabling types and distances.
`
`Table 2. Cabling Types and Distances
`Cable Type
`Distance
`Single-mode fiber
`2 kilometers
`Multimode fiber
`500 meters
`Category 5 Unshielded
`100 meters
`Twisted Pair (UTP)
`
`Initial Gigabit Ethernet installations will use single-mode and multimode fiber optic cable.
`Support for 4-pairs of Category 5 UTP will be added later, with a goal of 100 meter cable runs.
`
`Flow Control
`The flow control mechanism for Ethernet networks is different for full-duplex and half-duplex
`transmission types. The next section discusses these two transmission schemes, and the flow
`control mechanisms for each.
`Full-Duplex Transmission
`Using full-duplex transmission, signals travel in both directions on the same connection, at the
`same time. Simultaneous bi-directional transmission allows the aggregate data rate of an Ethernet
`network to be doubled. For example, a 10Mbps Ethernet network can achieve an aggregate
`20Mbps data rate, a 100Mbps Fast Ethernet network can achieve an aggregate 200Mbps, and a
`1000Mbps Gigabit Ethernet network can achieve and aggregate 2Gbps data rate.
`Full-duplex transmission is available for point-to-point connections only. Using full-duplex
`transmission, the issue of collisions on the network is completely eliminated and the CSMA/CD
`access control mechanism does not need to be invoked. Full-duplex can be used between a single
`workstation and a switch port, between two switch ports, or between two workstations.
`Full-duplex cannot be used for shared port connections, such as a repeater or hub port connecting
`to multiple workstations. Figure 10 shows a full-duplex, point-to-point network.
`
`12
`
`®Alteon Networks, Inc. 1996
`
`INTEL Ex.1087.020
`
`

`

`Gigabit Ethernet Technical Brief
`
`8 Independent,
`Point-to-Point Segments
`
`Segment 8
`
`Segment 7
`
`Segment 1
`
`Segment 2
`
`Switch
`
`Segment 3
`
`Segment 6
`Segment 5
`
`Segment 4
`
`Full-Duplex Transmission
`Figure 10. Full-duplex, point-to-point Ethernet segments
`
`An optional flow control mechanism being defined by IEEE 802.3x, is available for full-duplex
`transmission and works in a way similar to XON/XOFF flow control. A receiving station at one
`end of the point-to-point connection can send a packet to the sending station at the opposite end of
`the connection instructing the sending station to stop sending packets for a specified period of
`time. The sending station ceases to transmit packets until the period of time has passed, or until it
`receives a new packet from the receiving station with a time of zero, indicating that it is okay to
`resume transmission.
`Full-duplex Ethernet is being standardized by the IEEE 802.3x committee. The full-duplex
`standard is not specific to any particular speed for Ethernet. It can be used for Ethernet, Fast
`Ethernet, and Gigabit Ethernet.
`Half-duplex Transmission
`Using half-duplex transmission, signals travel in both directions on the wire, but not
`simultaneously. The original 802.3 standard specifies half-duplex transmission.
`To gain access to the network in a half-duplex environment, Ethernet has traditionally employed
`Carrier Sense Multiple Access with Collision Detection (CSMA/CD) as the standard access
`method. Using CSMA/CD, a station waits for a clear channel. When it detects clear channel, it
`begins to send frames onto the wire. If two stations start sending data at the same time, a collision
`occurs. Each station must detect the collision, abort the transmission, and wait for a random
`interval of time before attempting to transmit data on the network again.
`Half-duplex transmission is most commonly used on shared Ethernet segments. Unlike
`point-to-point connections, shared segments have two or more stations sharing a single port.
`Figure 11 shows a half-duplex, shared segment.
`
`13
`
`INTEL Ex.1087.021
`
`

`

`Gigabit Ethernet Technical Brief
`
`Half-Duplex
`Transmission
`
`Repeater
`
`Shared
`Segment
`
`Figure 11. Half-duplex, shared Ethernet segment
`
`Most switches manufactured today allow the user to select half-duplex or full-duplex on a port by
`port basis. This allows you to migrate your network from shared segments to point-to-point
`full-duplex segments over time. A switch port can be shared by front-ending the port with a
`repeater or hub. Figure 12 shows a network with both shared and dedicated switch ports.
`
`Half-Duplex
`Shared
`
`Full-Duplex
`Point-to-Point
`
`Switch
`
`Repeater
`
`Figure 12. Ethernet network with full-duplex and half-duplex ports
`
`Carrier Extension
`In order to abide by the rules of the CSMA/CD access method, all stations sharing an Ethernet
`segment must be able to hear and detect a collision that has occurred on the network. More
`specifically, a station must hear the collision for the frame it is sending before it has completed
`transmitting the entire frame. The amount of time it takes for a station to send the frame the full
`length of the wire and have the jam signal resulting from a collision travel back to the station is
`known as the slot-time.
`With very small (64 byte) frames travelling at speeds of 1,000 Mbps, the standard slot time used in
`the original IEEE 802.3 Ethernet specification is not long enough to accommodate a 100 meter
`cable run. The small frames are transmitted too quickly, and the sending station is finished
`transmitting the frame before it is aware of any collision that might have occurred at the other end
`of the segment.
`
`®Alteon Networks, Inc. 1996
`
`14
`
`INTEL Ex.1087.022
`
`

`

`Gigabit Ethernet Technical Brief
`In order to account for the problems of simply scaling CSMA/CD, and allow for minimum 64-byte
`frame sizes for compatibility across Ethernet devices, half-duplex Gigabit Ethernet implements a
`longer slot time using a technique called carrier extension. The frame size is not changed, but the
`time consumed on the wire is extended. Figure 13 shows a standard Ethernet frame and how
`extension is used, when necessary, to guarantee at least a 512-byte slot time.
`
`Original
`Ethernet
`Frame
`Start
`of
`Frame
`
`Idle
`
`Original Ethernet Slot-Time
`
`Preamble
`
`Start
`Delimiter
`
`Destination
`Address
`
`Source
`Address
`
`Length/
`Type
`
`Data
`
`Frame
`Check
`Sequence
`
`End
`of
`Frame
`
`Idle
`
`Idle
`
`Preamble