Advanced Ethernet Features

1
Local Area Networks

• Advanced Ethernet Features

2
IEEE 802.3 Family of LAN Protocols Advanced
Features

• Introduction
• Along with order-of-magnitude increases in speed
  and the application of switching to Ethernet,
  there are a number of advanced features that have
  been developed
• Full-duplex operation
• Jumbo Frames
• Flow Control
• Link Aggregation
• Virtual LANs (VLANs)
• Priority Transport
• Port Authentication
• There are other networking technologies that can
  take advantage of these features we are only
  discussing these in the context of Ethernet

3
Advanced Ethernet Features Full-Duplex Operation

• Introduction
• Traditional Ethernet is half-duplex a station
  cannot send and receive data simultaneously
• While upgrading from shared to switched LANs
  increases network capacity, it does not
  completely eliminate CSMA/CD from network
  operation
• The collision domain shrinks, but there are still
  two stations competing for the medium
• Full-duplex operation allows nodes to transmit
  and receive simultaneously
• Effectively no collision domain because there is
  only one transmitter and one receiver on a
  switched LAN leading to direct transmission
  without CSMA/CD.
• 100 link utilization with well-designed LAN
  switches
• 100BASE-TX, and 100BASE-FX support full duplex as
  transmit and receive signal paths can be
  simultaneously active.

4
Advanced Ethernet Features Full-Duplex Operation

• Collision Domains
• Lets take a moment to explore the impact of
  switching full-duplex operation on the
  collision domains found in a network
• Half-duplex shared
• Half-duplex switched
• Full-duplex switched
• A tangent switch backplane capacity
  full-duplex operation

5
Advanced Ethernet Features Full-Duplex Operation

• IEEE 802.3x
• The IEEE 802.x Working Group developed a standard
  that enhanced switched mode operation by defining
  full-duplex data transfer operation
• IEEE 802.3x finalized in 1997
• Allow independent (and simultaneous) transmission
  and reception of data by an Ethernet node
• Though mistakenly associated with Fast Ethernet,
  IEEE 802.3x is applicable to all flavors of
  Ethernet
• Also included specifications for flow control
  across full-duplex links
• Requires backwards compatibility (though manual
  configuration of mixed half and full-duplex
  equipment requires care!

6
Advanced Ethernet Features Full-Duplex Operation

• Requirements
• Requires the use of a star topology with a
  central wiring closet
• The Ethernet PHY must have independent transmit
  receive paths
• Baseband coax -gt NO!
• Fiber and UTP -gt YES
• There must be only two nodes on a dedicated
  point-to-point link
• Both nodes must be configured for full-duplex
  operation
• Requires switches beware of full-duplex
  Ethernet hubs!

7
Advanced Ethernet Features Full-Duplex Operation

• Operational Considerations
• While full-duplex ports are CSMA/CD capable, the
  MAC algorithm is disabled in full-duplex
  operation
• Frames are transmitted as soon as possible with
  only the IFG (Inter-frame Gap) between frames
• While CSMA/CD distance (delay) requirements no
  longer apply, the noise attenuation
  characteristics of the physical media still
  impose distance limits
• For UTP, half and full-duplex limits are the same
• For fiber, full-duplex links can span great
  distances (up to 40km with single-mode fiber and
  expensive optical components)

8
Advanced Ethernet Features Full-Duplex Operation

• Flow Control across full-duplex links
• While great for efficiency, care is needed to
  prevent full-duplex nodes from being overwhelmed
  with traffic
• The committee incorporated MAC-layer flow control
  into 802.3x to help prevent buffer overflow at
  nodes
• Without CSMA/CD the methods used for flow control
  in a half-duplex environment disappeared (e.g.-
  backpressure)
• The 802.3x MAC Control Protocol
• The standard defined new MAC control frames to
  use between nodes in a full-duplex connection
• These control frames are special Ethernet frames
  of type 0x8808 hex that contain opcodes for
  various functions

9
Advanced Ethernet Features Full-Duplex Operation

• The PAUSE MAC Control Frame
• The only defined MAC Control Command assigned
  opcode 0x0001 hex
• Sent to the reserved destination multicast
  address 01-80-c2-00-00-01
• Included in the frame is the Pause time
• Can be a value between 0 and 65,535
• This value is a multiplier receiver multiplies
  value by 512 bit times to determine the Pause
  time
• During that interval Receiver (of the PAUSE
  frame) should transmit no frames

10
Advanced Ethernet Features Jumbo Frames

• Introduction
• The effect of frame size on efficiency
• Current maximum ethernet frame size is 1518 bytes
  (1522 with VLAN tagging)
• Compare this with other technologies like Token
  Ring and FDDI
• Not efficient for certain applications
• File transfers
• Compute clusters
• Storage Area Networks
• Jumbo frames nominally refer to any ethernet
  equipment that can transport frames larger than
  the standard frame size

11
Advanced Ethernet Features Jumbo Frames

• Implementation
• While Jumbo frames are implemented by many
  manufacturers, there is no current standard
• This means you must be careful with network
  design and interoperability
• Even in a vendors product line there can be
  differences!
• A variety of sizes has been proposed a very
  common size is around 9000 bytes
• Large enough to allow a complete NFS message to
  ride in a single Ethernet frame
• Another common size is around 4470 bytes (to be
  compatible with other protocols like FDDI)
• See http//darkwing.uoregon.edu/joe/jumbo-clean-g
  ear.html for a list of equipment frame sizes

12
Example (TCP packet transfer)

• Due to overhead caused by TCP relaibility, the
  TCP throughput lt 0.7 MSS / (rtt
  sqrt(packet_loss)), where MSS is maximum segment
  size (MTU minus TCP/IP headers), rtt is round
  trip time, and packet_loss is probability of
  frame loss.
• Suppose rtt is about 40 msec, and let's say
  packet_loss is 0.1. With an frame size of 1500
  bytes (MSS of 1460). Evaluate the TCP throughput.
• Answer TCP throughput 0.7x1460x8/(0.04x0.033)6.
  5 Mbps based on TCP's ability to detect and
  recover from congestion (loss).
• Assume Jumbo Frame of 9000 byte. Evaluate the TCP
  throughput.
• AnswerTCP throughput 0.7x(9000-40)x8/(0.04x0.033
  ) 40 Mbps.
• Evaluate the packet loss rate to achieve a
  throughput of 500 Mbps with 1500 bytes and 9000
  byte frames.
• Answer packet_loss (0.7 x MSS/(rtt x
  TCP_throughput))2.
• We would need a packet loss rate of no more than
  1x10-5 when frame size is 9KBytes.
• However, with 1500 byte frames, the required
  packet loss rate is down to 2.8x10-7!
• While the jumbo frame is only 6 times larger, it
  allows us the same throughput in the face of 36
  times more packet loss.

13
Example (Continue)

• A 9000 byte GE packet takes the same amount of
  time to transmit as a 900 byte F-Ethernet packet
  or a 90 byte 10 Mbps Ethernet packet.
• Jumbo frames on GE at worse add less delay
  variation than 1500 byte frames do on slower
  Ethernets.
• No one is suggesting that slower Ethernets use
  9000 byte frames.
• As for queueing delay concerns, that could happen
  whether packets are large or small. If delivery
  QoS is required, then the routers need to
  implement some kind of priority or expedited
  forwarding, regardless of the packet sizes.
• Tiny frames (including 53 byte ATM cells) may be
  helpful when multiplexing lower bit rate streams,
  but they become increasingly inefficient on
  gigabit and beyond links.
• Conclusion leaving the local area network at
  high speed, the dynamics of TCP will require to
  use large frame sizes. Without them, the packet
  loss rate over a high bandwidth-delay product
  path would have to be extraordinarily low. Core
  internet infrastructure, from campus backbones to
  Network Access Points (NAPs), should be
  particularly careful not to limit the permitted
  MTU to 1500 bytes. In the long run there is no
  reason to stop at 9000 byte frames, but given the
  current ethernet CRC limitation it is a good
  evolutionary step for gigabit data rates.

14
Advanced Ethernet FeaturesLink Aggregation

• Introduction
• Allows an increase in network capacity
  availability without changing the underlying
  network technology
• Another alternative to use when upgrading to a
  higher speed is neither feasible or possible
• Sometimes called inverse multiplexing
• Benefits
• Increased bandwidth capacity
• Granular capacity increase
• Higher link availability
• Uses existing hardware

15
Advanced Ethernet FeaturesLink Aggregation

• Disadvantages
• Like any technology there is always disadvantages
  -- these have kept link aggregation a niche
  solution
• More space network interfaces necessary
• More complexity
• More maintenance overhead
• Harder to troubleshoot aggregated links
• Performance improvements depend on traffic flow

16
Advanced Ethernet FeaturesLink Aggregation

• Application Scenarios
• Switch-to-Switch
• Switch-to-Station (possibly a server or router)
• Station-to-Station (very rare)

17
Advanced Ethernet FeaturesLink Aggregation

• Issues Considerations
• Addressing Interfaces
• Need to assign logical MAC across all aggregated
  NICs
• Traffic Distribution Algorithm
• How to allocate traffic among aggregated links
• Want efficiency, but have to worry about other
  factors
• Transparency to upper layers
• Non-duplication
• Ordering
• Performance

18
Advanced Ethernet FeaturesLink Aggregation

• Other Considerations
• Mixing aggregate-able technologies
• Mixing speeds (versions) of aggregate-able
  technologies
• Using aggregation with shared LAN technologies
• Must worry about the operation of the MAC
  algorithm
• Really not recommended for use with shared LANs

19
Advanced Ethernet FeaturesLink Aggregation

• The IEEE 802.3ad Link Aggregation Standard
• Work on the standard began in 1998 with final
  approval in 2000
• Standard applies to Ethernet only
• All links in an aggregation must be the same
  speed
• Other restrictions
• Only full-duplex links
• Only one aggregation group per pair of devices
  (there can be other individual links between
  these devices)
• Many technical details taken from Etherchannel
  an earlier proprietary link aggregation protocol
  developed by Cisco

20
Advanced Ethernet FeaturesLink Aggregation

• The IEEE 802.3ad Goals
• Incremental bandwidth across logical channels
• Linearly incremental bandwidth
• Increased link availability
• Automatic configuration fault tolerance via
  rapid link reconfiguration
• Maintenance of link invariants
• Transparency to upper layers/applications
• Backwards compatibility with non-802.3ad
  equipment
• No change to Ethernet frame
• Network Management support

21
Advanced Ethernet FeaturesLink Aggregation
Operation
22
Advanced Ethernet FeaturesLink Aggregation

• The IEEE 802.3ad Control Protocols
• To maintain control of the aggregated links and
  traffic between devices, two control protocols
  are used
• Marker Protocol
• Link Aggregation Control Protocol (LACP)
• Marker Protocol
• Used to move data flows from one aggregated
  link to another
• Uses fixed and link-constrained 128 byte Ethernet
  frames
• Uses request/response operational commands

23
Advanced Ethernet FeaturesLink Aggregation

• Link Aggregation Control Protocol (LACP)
• Used to automatically configure maintain
  aggregated links between cooperating systems
• Protocol is generally passive, operating in the
  following manner
• Devices advertise their configuration capability
  via LACP messages
• There are no response messages devices are
  supposed to read incoming messages and configure
  themselves appropriately (usually the best common
  configuration)
• If device status or capability changes the
  change is to be advertised via LACP
• LACP messages can be overridden by manual
  configuration

24
Advanced Ethernet FeaturesLink Aggregation

• Link Aggregation Control Protocol (LACP)
• What is exchanged in LACP messages
• System ID used to ensure all aggregated links
  belong to the same (far-end) device
• Port Numbers Priority uniquely identifies
  links optionally assigns an aggregation
  priority to a link
• Aggregation Link Characteristics not all links
  between devices are aggregation candidates a key
  (unique value) shared by all links that can be
  aggregated
• Operational Mode LACP can operate in either
  Active (send LACP messages without prompt) or
  Passive mode
• LACP Message Transmission Frequency the
  configurable time interval between generation of
  LACP messages

25
Advanced Ethernet FeaturesVirtual LANs (VLANs)

• Introduction
• With LANs as discussed so far, there is no
  difference between the physical logical network
  layout
• With VLANs the physical topology can be different
  than the logical topology
• In other words, the set of stations that can
  communicate as if they are part of the same LAN
  (via direct MAC layer frames) can now be
  physically separated
• Requires the use of VLAN-aware switches
• VLAN applications
• Moves, Adds, Changes
• LAN Security (traffic isolation)
• User Mobility
• Bandwidth Efficiency

26
Advanced Ethernet FeaturesVirtual LANs (VLANs)

• The old way

27
Advanced Ethernet FeaturesVirtual LANs (VLANs)

• The new way

28
Advanced Ethernet FeaturesVirtual LANs (VLANs)

• Requirements for VLAN Operation
• Frame tags
• VLAN awareness
• VLAN association rules
• Frame distribution
• How do switches know the logical (virtual)
  groupings?
• First, frames belong to VLANs
• Two methods of associating frames with VLANs
• Implicit tagging
• Each switch examines the frame and based on its
  characteristics associates
• Explicit tagging
• Fields in the frame carry VLAN information

29
Advanced Ethernet FeaturesVirtual LANs (VLANs)

• Varieties of VLAN Associations
• Switch port-number based VLAN membership
• MAC-address (48 bit) based VLAN membership
• Protocol-based
• IP-based
• Application-based

30
Advanced Ethernet FeaturesVirtual LANs (VLANs)

• The IEEE 802.1Q Standard
• Due to demand, development of 802.1Q began in
  1995
• Final approval in December 1998
• Does not replace IEEE 802.1D (MAC-layer Bridges)
  but compliments it
• Extends it for VLAN-aware switches
• Maintains backward compatibility for non-VLAN
  operation and use in mixed environments
• Defined VLAN use with Ethernet, Token Ring, and
  FDDI
• What the standard covers
• Frame tagging
• Forwarding/Filtering Database
• Priority Operation
• Encapsulation of Token Ring/FDDI frames
• Automatic distribution of VLAN information
• Management of VLAN-aware switches

31
Advanced Ethernet FeaturesVirtual LANs (VLANs)

• IEEE 802.1Q VLAN Tags Frame Format
• Required modification of the basic Ethernet frame
• Designed to fit in allow operation with
  non-tagged frames
• Required extensive testing to ensure expanded
  frame (1522 bytes) caused no problems in existing
  equipment
• Maximum Frame size changed in IEEE 802.3ac (1998)

32
Advanced Ethernet FeaturesVirtual LANs (VLANs)

• IEEE 802.1Q VLAN Tag Fields
• VLAN Protocol ID (16 bits) set to 0x8100 hex in
  all tagged frames
• Priority (3 bits) used for the Priority
  function described later
• Canonical Format Indicator (1 bit) only
  important when using 802.1Q to bridge between
  technologies
• VLAN Identifier (12 bits) allows a maximum of
  4094 VLANs in a network
• The value 0xFFF hex is reserved
• The value 0x000 hex denotes a priority tag only
  frame has no VLAN association
• E-RIF optional field used with Token Ring
  FDDI LANs employing native source routing

33
Advanced Ethernet FeaturesVirtual LANs (VLANs)

• IEEE 802.1Q VLAN Switch Operation
• Normal switches examine the destination MAC
  address in frames to determine what ports to
  transmit it on
• The filtering/forwarding database matches
  destination MAC address to a port (unicast)
• Unknown MAC addresses are flooded until learned
• Broadcast MAC addresses transmitted out all port
  except the one it was received on
• VLAN-aware switches add a second variable into
  the filtering/forwarding database corresponding
  decision
• VLAN association for the frame
• The switch needs a way to determine which
  physical ports to associate with a VLAN
• Static (manual) mapping
• Dynamic mapping

34
Advanced Ethernet FeaturesVirtual LANs (VLANs)

• Automatic VLAN configuration
• To allow easier administration of VLAN-based
  networks, a control protocol was developed for
  communicating VLAN information between switches
• The protocol, call GVRP (GARP VLAN Registration
  Protocol), is part of a family of switch-related
  control protocols
• GVRP allows the switch at one end of a link to
  advertise the VLANs associations for the physical
  port to the far-end device
• Uses a reserved destination MAC address
  01-80-c2-00-00-21
• Similar definitions and operation to LACP

35
Advanced Ethernet FeaturesTraffic Classes
Quality of Service

• Introduction
• While priorities are included in several of the
  802 protocols and the IEEE 802.1D specification
  outlines how bridges should map priorities
  between different protocols, the original
  specifications do not define a true set of
  traffic classes
• Ethernet as defined has no traffic classes
• Bridges are not told how to prioritize frames for
  transmission)
• To provide better MAC layer support for
  time-critical data an update to the 802.1D
  specification defines and outlines the use of
  traffic classes
• This work is sometimes called 802.1p because it
  was the committee that developed the priority
  mechanisms

36
Advanced Ethernet Features Traffic Classes
Quality of Service

• The Use of Traffic Classes in IEEE 802.1D
• The updated standard relies on three concepts
• User Priority the priority found in the
  priority field of the MAC frame usually carried
  end-to-end unless it needs to be modified to fit
  the rules of a transit MAC protocol
• Access Priority the priority with which a
  bridge accesses the outbound LAN to transmit a
  frame it must relay
• Traffic Class if bridge has multiple queues for
  outbound traffic, the traffic class is used to
  determine the relative priority of the queues
• The Traffic Class is assigned by the bridge on
  the basis of incoming user priority
• Traffic classes help reduce the queuing delay
  seen by high priority frames, though there is
  still an access delay for transmission (based on
  the outbound MAC protocol)

37
Advanced Ethernet Features Traffic Classes
Quality of Service

• The Use of Traffic Classes in 802.1D (continued)
• Eight traffic classes are defined usable for
  each outbound port, corresponding to eight
  distinct outbound queues
• Within a queue a FIFO discipline is typically
  used strict ordering in the traffic class
• Output scheduling
• There are two basic choices for determining how
  to pull data out of the queues for transmission
• Obviously, strict round robin is not one of the
  choices!
• Strict Priority transmit any waiting frames out
  of the highest priority queues exclusively
• Weighted Fair Queuing (WFQ)
• Uses round robin but is modified by a weight
• Higher weights assigned to higher priority queues

38
Advanced Ethernet Features Traffic Classes
Quality of Service

• The Use of Traffic Classes in 802.1D (continued)
• Bridge Queue Diagram

39
Advanced Ethernet Features Traffic Classes
Quality of Service

• Weighted Fair Queuing

40
Advanced Ethernet Features Traffic Classes
Quality of Service

• Mapping of User Priority to Traffic Class
• The recommended 802.1D mapping of incoming user
  priority to traffic class is shown in Table 12.4
• If the 802.1Q (virtual LAN) specification is in
  use the priority field in the tag header can be
  used to determine the traffic class if necessary
• How this is done was already discussed
• Even protocols like 802.3 and 802.11 that support
  only one access priority can support multiple
  traffic classes (queues) but how the queues are
  filled are determined by other parameters (e.g.
  protocol type)

41
Advanced Ethernet Features Traffic Classes
Quality of Service

• Internet Traffic Quality of Service
• Many higher layer protocols have QoS mechanisms
  there should be some way to map higher level QoS
  levels to 802.1D traffic classes
• There are a couple of difficulties with this
• Internet traffic usually traverses a set of
  diverse networks making mapping between layers
  difficult
• Lower layers typically cannot see the QoS
  mechanisms used in higher layers
• In IP and ATM networks this would allow higher
  layer QoS (like the IP TOS bits or ATM service
  class) to be used to set the priorities of frames
  wherever possible
• Though important for the construction of
  internets with end-to-end QoS support, this is an
  area of ongoing research that currently lacks
  approved standards

42
Advanced Ethernet Features Traffic Classes
Quality of Service

• The Final Word on Traffic Classes
• Remember there is a difference between Quality
  of Service and Class(es) of Service
• Quality of Service provides or tries to guarantee
  the system/network will provide certain agreed
  upon service levels
• Minimum bandwidth
• Maximum Delay
• With Classes of Services there are not
  guarantees per say

43
Advanced Ethernet Features Port Authentication

• Introduction
• Once upon a time the Internet was a friendly
  place
• There was not much worry about security nowadays
  that is obviously not the case
• The IEEE saw the need to secure the edge of the
  network as a key security weakness
• Main targets
• Wireless networks (IEEE 802.11)
• Edge Ethernet ports

44
Advanced Ethernet Features Port Authentication

• IEEE 802.1x
• A working group was established to address edge
  security and develop a standard solution (Sum
  2001)
• Solution is not Ethernet-specific, but it has
  been adopted mainly for securing 802.3 802.11
  (WLANs)
• General Operation
• IEEE 802.1x allows a network device (switch) to
  query and authenticate a node before allowing
  network access
• Supported on Win XP and CE nodes third-party
  add-ins for other operating systems
• If the 802.1x-capable device does not hold
  authentication information, it must have some
  back-end access to an authentication server
  (via another authentication protocol such as
  RADIUS or TACACs)

45
Advanced Ethernet Features Port Authentication

• IEEE 802.1x Operation
• Allows authentication via username/password or
  some other user-based credentials
• Uses EAPOL (Extensible Authentication Protocol
  over LAN) for the node (802.1x Supplicant) to
  exchange authentication data with the switch
  (802.1x Authenticator)
• EAPOL adapted from earlier authentication
  protocols developed for PPP (Point-to-Point
  Protocol)
• Port remains locked until successful
  authentication occurs
• Some vendors support more advanced features based
  on 802.1x (VLAN selection), but this is not part
  of the standard

46
Advanced Ethernet Features Port Authentication

• IEEE 802.1x Operation

47
Advanced Ethernet Features Port Authentication

• IEEE 802.1x Operation (2)

48
Advanced Ethernet Features Port Authentication

• IEEE 802.1x Operation (3)

49
IEEE 802.3 Family of LAN ProtocolsHomework
Reading

• Homework 3 - Due at Class 6 in two weeks
• Chapter 7 7.2, 7.7
• Additional Question (5 points) download
  Ethereal (www.ethereal.com) or an equivalent
  sniffer program and install on a PC you have
  access to and permission to use. Capture a web
  session in a file (at least getting a couple of
  different pages) and explain what traffic you
  see. If any frames appear that are not part of
  the web session explain why they are there.
• Complete Lab1 in the OPNet Lab Manual, submit
  your answers to the questions at the end of the
  Lab
• Reading
• This weeks material Stallings chapters 7 and
  12.5
• Next week Chapter 8, 9, and 10