GigE & 10GigE Ethernet Technology Overview Testing Approaching€¦ · GigE & 10GigE Ethernet...
Transcript of GigE & 10GigE Ethernet Technology Overview Testing Approaching€¦ · GigE & 10GigE Ethernet...
GigE & 10GigE Ethernet Technology Overview&
Testing Approaching
Ray Lanier
Concord Communications Associates, Inc.
214 South Main Street
Concord, NH 03301
Tel: 603-228-8627 Fax: 603-224-3719
www.ccairep.com
AgendaEthernet Fundamentals
OSI Model
Ethernet and SONET comparison
Physical Layer (Layer 1)
Ethernet Frames (Layer 2)
Virtual LAN (VLAN) Concept
Auto-Negotiation Process
Flow Control – Pause Frames
The OSI (Open Systems Interconnect) model defines layers in a network. Understanding the function of each layer is key in understanding data
communication within Local, Metropolitan or Wide networks.
ApplicationPresentation
SessionTransportNetworkData LinkPhysicalLayer 1
Layer 2
Layer 3
Layer 4
Layer 5
Layer 6
Layer 7 Concerned with the end-users particular application
Provides for the representation of dataPerforms administrative tasks and security
Ensures end to end error free deliveryHandles addressing and routing between elementsHandles transfer of data over the channelHandles physical signaling, including connectors, timing, voltage T1, SONET
ATM, Frame Relay,Ethernet
IP, IPX
TCP, UDP
HTTP,FTPetc
The OSI Layer Model
Ethernet is generally placed in Layer 2 – Data Link Layer of the OSI model
Ethernet can be transported as follows:
• Ethernet over DWDM – this is a Layer 1 service
• Ethernet over SONET – could be a Layer 1 or Layer 2 service, depends on the network element deployed
• Ethernet Routed Network – this is a Layer 3 service
How does Ethernet fit into OSI Model?
SONETLayer 1
10/100/GigE and10GigE LAN
Layer 1 and 2
ApplicationPresentation
SessionTransportNetworkDatalinkPhysical
OSI Layer Model
IPATM
SONET Ethernet
IPEthernet
ApplicationPresentation
SessionTransportNetwork
PhysicalDatalink
SONET / Ethernet Comparison
Synchronous Timing Asynchronous TimingSONET Ethernet
OC-12RING
SONET / Ethernet Comparison
SONET / Ethernet Comparison
SONETEthernet
Bit Rate
Time
Designed for voice traffic• Fixed frame size• Constant bit rate
Designed for data traffic• Variable frame size• Bursty Traffic
SONET Ethernet
SONET
10/100/GigE and 10GigE LAN
Layer 1 – Physical Layer
Fiber Principles and Copper Characteristics
Ethernet Physical Layer Media – 10/100 and GigE
Most Common Physical Media for 10/100 and GigE• 10/100 Ethernet:
• 10BaseT/100BaseTX• Gigabit Ethernet:
• 1000BASE-SX (850 nm, Multi-Mode Fiber) • 1000BASE-LX (1310 nm, Single-Mode/Multi-Mode Fiber)
Other Interfaces• Gigabit Ethernet
• 1000BASE-CX, 1000BASE-T copper• 1000BASE-ZX (80km, 1550nm, single mode)
• 100 Mbps• 100BaseFX (multimode fiber)
Datalink
Network
Transport
Application
Presentation
Session
Physical
10GigE Physical Layer Media
Presentation
Application
Transport
Session
Datalink
Network
PhysicalSource: IEEE 802.3ae
10GigE LAN
10GigE WAN
Layer 2 – Ethernet Frames
Ethernet Link (Layer 2) EstablishmentPhysical Layer has to be active:
• This means the optical signal has to be received• Ethernet Synchronization (for example, 1.25 Gbps) achieved• Physical Link Established (Auto-negotiation must be
achieved)
If Physical Layer healthy, Ethernet frames can be send to verify traffic and throughput
Layer 2 traffic (Ethernet frames) can be transported over the link
Each Ethernet node has a unique address, called a MAC address, assigned to it by the manufacturer
All traffic originating on that node has its MAC address populated into a source address field in the Ethernet frames
All traffic destined for that node will have its MAC address populated into the destination address field by the originating node
Datalink
Network
Transport
Application
Presentation
Session
Physical
Same frame regardless of rate - 10Mbps, 100Mbps, 1Gbps or 10Gbps LAN
Variable Frame Size – 64 bytes to 1518 bytes (excluding Preamble and SFD)
Must have an integer number of bytes in every frame
Note – testing practices• Undersized frames less than 64 bytes are considered errors• Jumbo frames larger than 1518 bytes are considered valid frames
Data (46-1500)L/T (2)SA (6)DA (6) FCS (4)SFD (1)Preamble (7)
Ethernet Frames
Data (46-1500)L/T (2)SA (6)DA (6) FCS (4)SFD (1)Preamble (7)
Preamble - allows physical layer to detect carrier and acquire sync (7 bytes of alternating 1’s and 0’s)
SFD - Start of Frame Delimiter - identifies the beginning of a frame (1 byte -10101011)
Ethernet Frame – More Detail
Addresses• DA - Destination Address • SA - Source Address• Each source address is unique• First 3 bytes identify manufacturer of equipment, assigned by IEEE
Length/Type Field• 802.3 frame - indicates length of data field (<=1500)• Ethernet Type II (DIX) frame - indicates type of frame data (>=1536)
Data (46-1500)L/T (2)SA (6)DA (6) FCS (4)SFD (1)Preamble (7)
Data (46-1500)L/T (2)SA (6)DA (6) FCS (4)SFD (1)Preamble (7)
Fields used for FCS calc
Ethernet Frame – FCS Calculation• Data – the payload -> next layer up (IP, IPX, etc.)• FCS - Frame Check Sequence - A 32-bit cyclic redundancy check
performed on the frame for error detection• FCS - a primary measurement to verify that your Ethernet pipe is good!
Used to channelize traffic (pipe with in a pipe)Can also be used to differentiate and prioritize traffic
Contains 4 fields:• Tag Protocol Identifier - Fixed at hex 81-00
for Ethernet• User Priority - Used to prioritize different
types of traffic• Canonical Frame Identifier - used only in
Token Ring• VLAN ID - Specifies the VLAN group (value is
0 to 4095)
Data (46-1500)L/T (2)SA (6)DA (6) FCS (4)SFD (1)Preamble (7) VLAN (4)
VLAN (Virtual Local Area Network)
1 Gigabit trunk
IPTVHigh Priority
Traffic
Internet BrowsingLow Priority Traffic
2 bytes 3 bits 1 bit 12 bits
VLAN IDCFIPriorityTPI
Auto-Negotiation ProcessMechanism used to negotiate the speed and duplex configuration of the link between two partners
Only applies to 10/100 and GigE - not to 10GigE
If Auto-Neg is ON (enabled):• Each node advertises capabilities to the other• Each node will configure themselves to the highest common denominator of
their partner’s capabilities and their own• Key parameters advertised:
• Half Duplex/Full Duplex • Flow control capabilities
If Auto-Neg is OFF (Disabled):• Link will come up IDLE
IMPORTANT - Both link partners must have settings for Auto-negotiation matched (either enabled or disabled on both ends), otherwise a negotiation mismatch will result and link will not be established
Flow Control Process
Mechanism used to control network congestion and prevent traffic from being dropped
Applies to 10/100, GigE and 10GigE interfaces
PAUSE Frames• Uses a defined MAC control frame format • Contains a time field that defines PAUSE time• Receipt of a PAUSE frame causes Ethernet device to cease transmission
for PAUSE time
Service getting more complex:• Point to point or switched• 10/100 and Gigabit interfaces• Not always a full Gigabit - configurable bandwidth in STS increments• Flow control may be enabled
SONET RingOC-48 or OC-192
Customer GigE Switch
EthernetLink A
EthernetLink B
SONET Link
Cisco 15454
Ethernet over SONET (EoS) Service
Relationship between SONET mapping and maximum throughput
Transport Circuit Size % GE Interface Speed Bandwidth in Mbps
STS1 5 50
STS3c 15 150
STS6c 30 300
STS9c 45 450
STS12c 60 600
STS24c 100 1000
STS48c 100 1000
Ethernet over SONET (EoS) Service
Turn-up, Acceptance, and Troubleshooting Ethernet/IP
Transport Networks
Concord Communications214 South Main Street
Concord, NH 03301Tel: 603-228-8627 Fax: 603-224-3719
www.ccairep.com
Agenda
Testing a network• Basic structure of an Ethernet/IP Network• Parameters to test
• Throughput, Frame Loss, Latency, and Jitter• Limits of basic testing
Transport Testing• Testing a bit Transparent Link• Testing an Ethernet over SONET Link• Testing a packet Switched Link
Ethernet Transport Testing DemoJDSU Ethernet Turn-up Product
CE PE PEPB PB CE
Access Backbone, Core WAN, Metro, Satellite
Access
Basic structure of an Ethernet/IP Network
EdgeTransport
Equipment
Transmissiontechnique
Switchingtechnique
EdgeProcessingEquipment
Packet orientedEthernet
Purpose of test is to check the backbone for errors
LAN/NE LAN/NE
Testing Responsibility of Transport Groups
Application Layer
End-to-End Transport Layer
CE PE PEPB PB CE
DWDMSONETSatellite
Ethernet/IPEthernet/IP
Responsibility of Transport Groups
Test is run from one LAN to another, placing a tester where the enterprise router used to be and testing the transport
layer
LAN/NE LAN/NE
Throughput – Do I get my full pipe?
Audio
Affected by • The max bandwidth link set up between two locations (e.g. SONET NE)• Network Overhead • Whether network elements are setup properly
Customer Complaints• For VoIP if the customer can only place min # of call at a time successfully• For Data if the customer states that during peak hours the internet slows
downDetailed Comments• Throughput is the key result of a test
• Frame loss, latency & jitter should be checked at max throughput for transport quality
• Best way to test throughput: • Generate a constant load at max throughput rate.
Round Trip Time (Latency) – Network DelayAffected by • Priority queuing and traffic shaping problems• Network Congestion• Length and type of Link
Customer Complaints• For VoIP when the customer complains of Over talk and echo• For Data if the customer complains of extreme load times for web pages
Detailed Comments• Key time to test round trip time is when throughput is at max on the link, to
ensure that buffering and live traffic is not going to drastically delay the Round Trip Time
• Because a ping is done using IGMP (Internet Group Management Protocol) routers handle these at a lower priority than traffic, pinging produces an inaccurate latency measurement
14 3
3
2
1 4TIMESTAMP TIMESTAMP
TIMESTAMP TIMESTAMPTIMESTAMP TIMESTAMP
TIMESTAMP2 TIMESTAMP
Packet Loss Parameter
Core IP Network/Gateway
Three Packets In Two Packets OutHome
Affected by • Buffer overruns in network components• Network Congestion• Bad network elements
Customer Complaints• VoIP packet clicks & fuzziness or even dropped calls in extreme situations• Again for Data complainants of extreme load times (TCP retransmissions)• Complains of no throughput it may be due to excessive packet loss
Detailed Comments• The best way to check packet loss is by having a sequence number in
every packet to see on the receiving side if any packets were lost or erred in transit
• Note this test should be run at max throughput as most packet loss problems can’t be seen unless the network and queues are being stressed
BERT versus packet oriented testing
1110111101011111010000100000111111101011
1110111101011111010000100000111111101011
Bit errors Throughput
4 (10%) (0%)
Packet based testing is the future of switched links though BERT is the past and can still be useful on bit transparent links
40 bit sequence of 4 packets
10 (25%) (75%)
Ethernet frames/packets are dropped if they contain a single errorAssuming 1518 byte packets, a BER of 8e-5 can lead to a 99.99% throughput or a 0% throughput rate depending on when the errors are distributed!!!
Bottom Line is a Packet Error Rate is key in the IP world, not the BER rate like we are all used to
Packet JitterPacket Jitter
Packets Equally spaced
Variation in packet spacing (jitter)
Packet Jitter = variation in arrival time between packetsPacket Jitter = variation in arrival time between packetsAffected by • Variation in network traffic levels• Network Congestion• Priority queuing and traffic shaping problems in combined VoIP / data
networksCustomer Complaints• For VoIP packet clicking and popping noises to the end user• For VoIP garbled speech
Detailed Comments• Note Jitter is important for testing in networks that use VoIP or IPTV. It is
NOT generally critical in data only networks
Transport Techniques
Two key things to remember about Ethernet
1. Ethernet frames that are errored (FCS errored frames) will be discarded by Ethernet-aware devices.
2. Devices with Ethernet Intelligence (switches, routers) can not be hard-looped like a physical layer (T1, etc) device.
• When performing testing on a Layer 2+ Ethernet network or element, you cannot use hard loop-backs and expect to get the traffic back
Testing links – Bit Transparent Link
The most important features:Ethernet access rate and BB transport rate must be of the same.Ethernet link is transferred bit transparent.Per Ethernet link a separate transport channel(fiber, opt. channel, SDH VCG).Very small delay time.No packet switching; No Store & Forward; No prioritizationBit errors and errored packets are forwarded.
on bit level
Dark fiberOpt. channelSDH VCG
Ethernet link
1 Gbps
Idles
Frame
1 Gbps
Idles
FrameFiber Patch
Opt. MUX
NewSDH/GFP-T
GFP-T: Transparent Generic Framing ProcedureVCG: Virtual Concatenation Group
CE - Access BackbonePE - Adaptation
Test links –Packet Switched Networks
The most important features:Ethernet access rate und BB transport rate can be different.Ethernet link is transferred packet based.For identification the PE inserts a tag/label.Multiple Ethernet links can be transferred in a single BB trunk.Store & Forward causes delay and delay variation.With packet switching; probably with prioritizationBit errors and errored packets are not forwarded.
on packet levelEthernet link
100 Mbps
Idles
Frame
FCS
Stor
e&Fo
rwar
d
L2/3
Pac
ket
Switc
hing
+ VP
N
Packet loss
CE - Access BackbonePE - Adaptation
pro Eth. Link < = 100 MbpsTotal link 1 GigE or 10 GigE
Frame
How to test EoS Service:Hard loop generally no longer certaintyDepends on network element FST-2802 generates traffic at customer premiseFST-2802 in CO loops traffic back
Verify:
Connectivity – “Link”
Through Put – “Ethernet
frames”
QoS – “no FCS errors”
SONET RingOC-48 or OC-192
Testing Ethernet over SONET (EoS)
EthernetLink A
EthernetLink B
SONET Link
Cisco 15454
Loopback Mode
FST-2802
FST-2802
Turn-up of multiple services (VoIP, Video, Data)
Test goal “Multiple Services”:• Prove 100% availability of the transmission bandwidth as a whole and divided
on the individual service classes (VoIP, Video, Data).• Service-related (VoIP, video, data) QoS proof is required.• Control of the bandwidth management (Prioritization, bandwidth rate
limitation)• QoS parameters per service: Throughput, Packet Loss, Round Trip Time,
Packet Jitter
Test method:• Measurement setup is possible only with two testers. Physical loop is not
permitted.• Generate Multiple Streams with service-typical load profiles, in order to
stimulate the test link in accordance with the later services.• RFC-2544 is not meaningful.
• RFC-2544 checks general performance and not service-related performance. • E.g. if one service is limited by low prioritization level, the coordination of the zeroing
process is rather complex.
Ethernet Transport Testing Demo
JDSU FST-2802 JDSU
TB-8000
GigE 1.25 Gbps
Single Mode FiberConnection
High-end Modular PlatformSingle or Dual Port (10/100/100 + Optical GigE)Single port 10GigEL1/L2/L3 TestingRFC2544Remote ControlOther Transport Technologies (SONET/SDH, T/E-Carrier etc.)
High-end Modular PlatformSingle or Dual Port (10/100/100 + Optical GigE)Single port 10GigEL1/L2/L3 TestingRFC2544Remote ControlOther Transport Technologies (SONET/SDH, T/E-Carrier etc.)
Point solutionSingle PortCable DiagnosticL2/L3 TestingRFC2544Loopback DeviceLow Cost
Point solutionSingle PortCable DiagnosticL2/L3 TestingRFC2544Loopback DeviceLow Cost
Modular PlatformSingle Port TermDual Port ThruCable DiagnosticL2/L3/L4 TestingRFC2544Multiple Streams VOIP/Video Test ApplicationsRemote ControlOther Access Technologies (E1/T1, DSL Copper)
Modular PlatformSingle Port TermDual Port ThruCable DiagnosticL2/L3/L4 TestingRFC2544Multiple Streams VOIP/Video Test ApplicationsRemote ControlOther Access Technologies (E1/T1, DSL Copper)
TB-8000TB-8000 SmartClassSmartClassHST-3000HST-3000
JDSU Ethernet Turn-up Product - Summary
First Ethernet SolutionSingle or Dual PortEthernet and FCL1/L2/L3 TestingRFC2544Multiple StreamsGraphical ResultsPDF PrintoutRandom Frame Size IP/VLAN DiscoveryRemote Control
First Ethernet SolutionSingle or Dual PortEthernet and FCL1/L2/L3 TestingRFC2544Multiple StreamsGraphical ResultsPDF PrintoutRandom Frame Size IP/VLAN DiscoveryRemote Control
FST-2802FST-2802