Differentiated Ethernet Service Verification

1
Differentiated Ethernet Service Verification

• Alex Luo / David Agresti
• Sr. Product Marketing Manager
• for Heynen

2
Summary
With the introduction of Ethernet services the
need came up to qualify links in the same way as
for ITU G.82x recommendations in PDH and SDH
networks. RFC2544 became a kind of benchmarking
standard as it is offering Throughput and Latency
measurements. The launch of new services like
Video over IP and Voice over IP requests new ways
to qualify the Ethernet link. From now on, the
link cannot just be seen as a simple pipe
offering a certain kind of bandwidth. Quality of
Service (QoS) is the headline. What is needed to
offer these services and how can we qualify them?
3
Content

• Differentiated Ethernet Service and Quality of
  Service (QoS)
• RFC2544 - Traditional Ethernet Throughput and
  Latency Verification
• VLAN and CoS
• IP and ToS
• Frame Delay Parameters
• QA

4
Quality of Service (QoS)

• Best-effort network or service doesnt support
  QoS.
• Refers to resource reservation control mechanism
  rather than the achieved service quality.
• Provides high quality communication over a
  best-effort network by over-provisioning the
  capacity sufficient for the expected peak traffic
  load.
• Provides different priority to different
  applications, users, or data flows, especially
  for real-time streaming multimedia such as delay
  sensitive voice over IP and IPTV services in
  networks where the capacity is a limited
  resource.
• Bit rate, delay, jitter, packet loss probability
  and/or bit error rate can be guaranteed.

5
Key Measurements of QoS

• Packet Loss
• Depends on the state of the network and routing
  devices, dropped packets may be re-transmitted
  causing severe delays in overall performance.
• Delay
• Packet gets held up in long queue, takes more
  direct routes to avoid congestion. Excessive
  delay can render application unusable.
• Jitter
• Packets from the source will reach with
  different delays varies with position in the
  queue and path between routes. Jitter can
  seriously affect the quality of streaming audio
  and/or video.
• Out-of-Sequence Delivery
• Different packets may take different routes
  results different delay and reaches destination
  in different order than they were sent. Video and
  voice quality is dramatically affected by both
  latency and lack of synchronization.
• Error
• Packets are misdirected, mis-combined or
  corrupted. Packets may be re-transmitted causing
  severe delays in overall transmission.

6
Protocols of QoS

• Resource Reservation Protocol (RSVP)
• IntServ model used by applications to request
  and reserve resource through a network.
• Differentiated Services (DiffServ)
• DiffServ model includes VLAN 802.1Q in MAC layer
  and differentiated service code point (DSCP) in
  IP layer.
• Class of Service (CoS) 802.1Q and 802.1p
• Three-bit field with Layer 2 VLAN.
• Type of Service (ToS)
• A byte in the IP header.
• Multi-protocol Label Switching (MPLS)
• Three-bit field for priority is in experimental
  stage.

7
RFC 2544

• Benchmarking Methodology for Network Interconnect
  Devices
• Not designed for test Ethernet services, but
  often used as such
• Six standardized, semi-automated tests
• Requires Loopback at far end for round trip
  measurement
• Layer 1 / DWDM Hard loop
• Layer 2 Soft loop with 2nd test set
• Frame Sizes
• Standard 64, 128, 256, 512, 1024, 1280, 1518
• With VLAN 68, 1522
• Other 65, 4096, 9000, etc.
• Pros
• Automated testing increases efficiency
• Results recognized (or expected) by customers
• Cons
• Tests do not always line up with service
  agreements, class of service, etc.
• Tests can take longer than is practical

8
RFC 2544 - Throughput

• Objective
• What is the maximum rate without losing any
  frames?
• Procedure
• Test service at 100
• If you lose 1 or more frames, reduce rate and try
  again
• If you lose no frames, move on to next frame size
• Pros
• Can batch a number of tests together into a
  single, automated process
• Provides maximum throughput, if less than 100
• Cons
• Very sensitive to frame loss
• Bad idea of routed, wireless, or other lossy
  networks
• A particular service may fail for all test rates
• Assumes 100 of Line Rate dedicated to service
• On a 10G link, if a particular service is policed
  to a maximum rate of 1 Gbps, then Throughput test
  will fail all test rates above 10.

9
RFC 2544 - Latency

• Objective
• How long does it take a frame to get across the
  network?
• Tested as a roundtrip delay measurement
• Procedure
• Send frames at throughput rate
• Put timestamp into test frame before sending
• When you receive the test frame, compare the
  incoming time with its timestamp
• Average over a number of trials to obtain a final
  result
• Pros
• The latency (or roundtrip delay) of the network
  is a key metric and has a direct impact on any
  real-time services, such as voice and video
• Cons
• Takes 4½ hours if you follow the standard for all
  frame rates
• Some networks can vary in latency hour-to-hour or
  day-to-day

10
RFC 2544 - Frame Loss Rate

• Objective
• How many frames are lost at various rates?
• Procedure
• Send frames at 100, 90, etc.
• Measure percentage of frames lost
• If two test rates in a row have no frame loss,
  move on to next frame size
• If 100 and 90 have 0.00 frame loss, you can
  skip the lower rates
• Pros
• Handy guide for understanding how a device
  performs under different loads
• Can compare to device specifications provided by
  supplier
• Cons
• Rarely provides useful service information
• On a perfect system (already tested and passed at
  100 throughput), test is a waste of time

11
RFC 2544 - Other Tests

• Back-to-back Frames
• How many frames can be sent at 100 throughput
  before losing any?
• System Recovery
• How long does it take to stop losing frames once
  I reduce the rate?
• Reset
• How long does the device take to reboot?
• Notes
• Apply to device benchmarking used in lab
  environment or by network equipment makers

12
VLAN and CoS
Preamble 7 Bytes
Preamble
SFD
Start of Frame Delimiter 1 Byte
Destination Address
Destination MAC Address 6 Bytes
Source Address
Source MAC Address 6 Bytes
TPID
Tag Protocol ID 2 Bytes
TCI
Tag Control Information 2 Bytes VLAN ID,
Priority and CFI
68 to 1522 bytes
Type/Length
Ethertype/Length 2 Bytes
Payload
Payload 46 to 1500 Bytes long
FCS / CRC
Frame Check Sequence 4 Bytes
13
Class of Service (CoS)

• A 3-bit field within VLAN (Layer 2) tag to
  specify a priority value that can be used by
  Quality of Service disciplines to differentiated
  traffic
• Traffic is managed by grouping similar types of
  application like email, streaming video, voice,
  file transfer etc.
• A level of service in bandwidth and delivery time
  is not guaranteed a best-effort method
• Simpler to manage and more scalable as a network
  grows in structure and traffic volume

14
VLAN Fields
12 Bits

• TPID (2 Bytes) Set to 8100 (hex) means that
  802.1Q is used
• User Priority (3 bits) (0 to 7)
• Defined by 802.1p or vendor-specific Class of
  Service (CoS) standard
• CFI (1 bit) Indicates ordering of the bits of
  the VLAN ID
• Set at 0 for Ethernet
• VID (12 bits) VLAN Identifier (0 to 4095)
• 4095 (FFF) is a reserved value
• 0 indicates no VLAN
• 1 is a reserved value for Ethernet bridges

User Priority
C F I
VID or VLAN ID
TCI
3 Bits
1
15
VLAN Membership

• By PORT All stations connected to a port will
  have the same VLAN ID
• Fast traffic forwarding
• Easy to maintain for network administrator
• User tied to geographic location
• By MAC address Each MAC address (source) will be
  associated with specific VLAN ID
• Offers more flexibility
• MAC address lookup takes more processing time
• Network administrator must manually assign VID to
  each MAC address
• By IP address (Layer 3 based VLAN) IP address
  associated with VID
• Flexibility and easy configuration
• IP address lookup requires more processing time

16
Q-in-Q MAC-in-MAC

• C-Tag Customer Tag
• S-Tag Service Tag
• VLAN Priority (0-7) based on the class of service
• Ideal for distinguishing Voice, Video, Data, and
  Management traffic
• B-MAC Backbone MAC Header
• The Access MAC frame becomes the payload for a
  Backbone MAC frame
• C-MAC Customer MAC Header
• Q-in-Q IEEE 802.1ad
• MAC-in-MAC IEEE 802.1ah

User
C-MAC
C-Payload
Enterprise CPE
C-MAC
C-Tag
C-Payload
Carrier Access
C-MAC
S-Tag
C-Tag
C-Payload
Carrier Core
B-MAC
C-MAC
S-Tag
C-Tag
C-Payload
17
VLAN Tagged Frame Lengths

• Minimum Legal Frame Size
• 64 bytes 4 per VLAN Tag
• 1 Tag 68 bytes
• 2 Tags 72 bytes
• Maximum Legal Frame Size
• 1518 bytes 4 per VLAN Tag
• 1 Tag 1522 bytes
• 2 Tags 1526 bytes
• Network interface cards and older switches may
  drop large tagged frames as "oversized"

18
CoS-Based Ethernet Service Level Agreement

• E-Line Service
• 4 Classes of Service
• CoS determined via 802.1p CoS ID
• Common type of SLA used with CoS-based IP VPNs

19
LAN MAN Networks with Q-in-Q
Customer A
VLAN 25
Provider Access Networks
VLAN 20
VLAN 20
VLAN 20
Customer B
Outer VLAN 200
Outer VLAN 100
VLAN 10
Provider Core Network
VLAN 10
Outer VLAN 100
Outer VLAN 200
VLAN 25
Customer A
VLAN 10
VLAN 20
VLAN 20
Provider Access Network
Customer B
20
Q-in-Q The Good and Bad

• Security
• Each customer is assigned a unique VLAN ID
  through the providers network
• Fully transparent
• Customer VLANs treated as data through providers
  network
• Customer can use any VLAN ID values
• No fear of overlap with other customers
• Service Policing
• Provider sets priority for each customer
• Limited QoS and traffic engineering

• Proprietary Implementations
• Cisco, Extreme Networks, Juniper, etc.
• Scalability issues
• Limited to 4095 customers
• Places strain on MAC learning capacity of
  backbone switches
• No signaling
• Manual provisioning of new services

21
Type of Service (ToS)

• A byte in the IPv4 header (Layer 3) defined in
  different RFCs (RFC 791, 1122, 1349, 2474)
• Modern definition is a 6-bit Differentiated
  Service Code Point (DSCP) and 2-bit Explicit
  Congestion Notification field
• Router support may become a must in the future
• Routing table maintains a ToS value for each
  route Default value is 0 if ToS is not specified
  or supported

22
IP Header ToS RFC 1349

• Precedence (3 bits) RFC 791
• 000 Routine
• 001 Priority
• 010 Immediate
• 011 Flash
• 100 Flash / Override
• 101 CRIITC/ECP
• 110 Inter-network Control
• 111 - Network Control
• Type of Service (4 bits)
• 0000 Normal Delay
• 0001 Minimize Monetary Cost
• 0010 Maximize Reliability
• 0100 Maximize Throughput
• 1000 Minimize Delay

Precedence
ToS
0
3 Bits
4 Bits
1
23
IP Header ToS RFC 2474

• Precedence (3 bits) RFC 791
• 000 Routine
• 001 Priority
• 010 Immediate
• 011 Flash
• 100 Flash / Override
• 101 CRIITC/ECP
• 110 Inter-network Control
• 111 - Network Control
• D Low Delay Request
• T High Throughput Request
• R High Reliability Request
• ECN (2 bits)
• Unused

Precedence
D
T
R
ECN field
3 Bits
2 Bits
1
1
1
24
Frame Delay Parameters

• Frame Delay Variation
• Inter Packet Delay Variation (IPDV) MEF10.1
  uses an IPDV approach, with a modification that
  it takes the absolute value of difference of
  adjacent observations
• Packet Delay Variation (PDV) The PDV approach
  subtracts the minimum delay from each observation
• Mean Frame Delay
• Many Service providers use the mean in Service
  Level Specification
• Frame Delay Range
• Provides a simple variation constraint thats
  easily explained
• Addresses aspects of attribute combination user
  relevance
• Provides a basis for comparison with ITU-T FDV
• One-way nature of Delay and Loss measurement
• Will result in consistent treatment of multipoint
  and point-to-point cases.

25
Frame Delay Performance (MEF 10.1)
26
Inter-Frame Delay Variation (MEF 10.1)
27
Frame Delay Range (MEF 10.1)

• Frame Delay Range is independent of the actual
  Frame Delays
• Different absolute delays have identical Frame
  Delay Ranges
• Frame Delay Range is normalized by subtracting
  the minimum. Objective is specified in terms of
  Y percentile - X percentile (min)

28
Differentiated Service Measurement

• Assign each stream a different
• VLAN Priority
• Type of Service field in IP Header
• DiffServ and DSCP Differentiated Services Code
  Point
• Assign each stream a different bandwidth
• Data
• Voice
• Video
• Management

• Aggregate statistics All Traffic
• Utilization
• FCS/CRC Errors
• Packet Distribution
• Per-Stream statistics
• Lost Frames
• Delay
• Delay Variation
• Out-of-Sequence Frames
• Error Frames
• Frame Rate

29

• Thank You
• QA