Jdsu Truespeed Automated Rfc-6349 T Testing 1a323f
This document was ed by and they confirmed that they have the permission to share it. If you are author or own the copyright of this book, please report to us by using this report form. Report 445h4w
Overview 1s532p
& View Jdsu Truespeed Automated Rfc-6349 T Testing as PDF for free.
More details 6h715l
- Words: 2,293
- Pages: 36
TrueSpeedTM Automated RFC 6349 T Testing Guylain Barlow, Senior Product Line Manager
Agenda § Who should test T? § Why Test T? § What is RFC 6349? § T Performance Issues and the TrueSpeed Solution
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
2
T Testing Applications Customer Site
T-BERD/MTS 5800 T-BERD
Ethernet Provider
Customer Aggregation
T-BERD/MTS 8000
Backhaul Network T-BERD/MTS 8000
T-BERD/MTS 6000
T-BERD/MTS 6000
1. Service providers offering Ethernet Business services 2. Service providers offering Ethernet Backhaul services 3. Mobility service providers deploying 3G or 4G/LTE © 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
3
Why is T Testing Important? § Running RFC2544, Y.1564 or other Layer 2/3 installation tests is always the first step § But even when these Layer 2/3 tests “”, end-customers can still complain that the “network is slow” and the cause of poor application performance (i.e. FTP, web browsing, etc.) § Lack of T testing is a turn-up gap because end-customer applications are transported using T in Layer 4 § Save up to 30% OPEX costs by eliminating or quickly resolving painful end-customer finger pointing scenarios
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
4
Which Applications use T?
7 6 5
on Applicati Presenta
tion
Session
4
t Transpor
3
Network
2 1
HTTP, FTP, E-mail, Sharepoint, Communicator, Facebook, YouTube, etc
Datalink Physical
T
RFC 6349 TrueSpeedTM
IP
Y.1564 SAMComplete & RFC 2544
Ethernet
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
5
Case Study: Customer Issue
100 Mbps Service ~20 msec RTT § § §
Layer 2 traffic testing was done twice, Test results = “” End-customer performance dipped to 50-60 Mbps at various times of the day (end-customer running “iperf” based speed tests) TrueSpeedTM test was run from network provider demarc in Tampa to Atlanta – T throughput was rock solid at nearly 100 Mbps (95 Mbps layer 4) NO DROPs – Finger was clearly pointed at customer data center network; investigation proceeded in the customer arena (provider was “cleared of guilt”) à larger T window was required © 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
6
Case Study: Network Provider Issue
300 Mbps Service ~75 msec RTT
§ § §
Layer 2 traffic testing was done twice, Test results = “” End-customer file transfer performance was low (150 Mbits/sec) TrueSpeedTM (30-40 T sessions with 64 KB window) stressed network buffers – Problem was reproduced with bursty traffic (traffic drop) – Network provider isolated to mis-configured device at California site à needed larger egress buffers
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
7
Brief Overview of RFC 6349
8
RFC 6349: T Throughput Test Methodology • IETF RFC 6349 specifies the methodology for measuring end-to-end T Throughput in a managed IP network. • JDSU and two (2) Network Providers co-authored this RFC “0”. Run traditional RFC2544 to the integrity of the network at Layers 2 and 3 before conducting T testing. 1. Path MTU Detection (per RFC4821) – network MTU with active T segment size testing to ensure T payload does not get fragmented 2. Baseline Round-Trip Time and Bandwidth – Predict optimum T window size to automatically calculate the T Bandwidth Delay Product (BDP) 3. Single and Multiple T Connection Throughput Tests – T window size predictions to enable automated “full pipe” T testing © 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
9
RFC 6349 Metrics: T Efficiency § T retransmissions are a normal behavior in network communications. • But what is the “efficiency” of a network transfer? • Time spent transmitting “good” payload versus retransmitting it. § The T Efficiency metric is the percentage of Bytes that did not have to be retransmitted and is defined as: T efficiency=
Transmitted Bytes - Retransmitted Bytes x 100 Transmitted Bytes
§ As an example, if 100,000 Bytes were sent and 1,000 had to be retransmitted, the T Efficiency would be calculated as: 101,000 - 1,000 101,000
x 100 = 99%
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
10
RFC 6349 Metrics: Buffer Delay § T throughput is also affected by increase in RTT, which can be caused by network congestion or buffer delay. § The Buffer Delay Percentage is defined as: Buffer Delay % = Average RTT during Transfer - Baseline RTT Baseline RTT
x 100
§ Example: If the baseline RTT for a network path is 2 ms and the average RTT increases to 3 ms during the test. The Buffer Delay Percentage would be calculated as: 3 – 2 x 100 = 50% 2
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
11
JDSU TrueSpeed = RFC 6349 Automated § Reduce operating expenses up to 30% by minimizing truck rolls § Complete fast, repeatable, automated T tests in less than 5 minutes § results with an intuitive graphical interface that is easy to use by technicians at all skill levels § Closes the testing gap! Solve the tough, costly problems that you miss with RFC 2544 and Y.1564
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
12
T Performance Issues and the TrueSpeed Solution
13
T Window Problem: Misconfigured E Host Provider Network
Customer Premise
100 M Ethernet Service 20 msec round trip Customer Laptop
“Long Fat Network” Firewall, Proxy, etc.
Problem
Description
Location / Responsibility
Window size not optimized
High latency and/or High BW links require end host T WINDOW SETTINGS TO BE OPTIMAL for full T throughput
End hosts, Firewalls, Proxies
How does TrueSpeedTM detect this?
WALK THE WINDOW test reports the actual throughput vs. End-Customer ideal for various Window Sizes
© 2014 JDS Uniphase Corporation
|
How can the problem be fixed? End-customer must tune the end host device T Window size
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
14
Latency and T Window Size Relationship § The T Window size is the amount of data that a sender will place onto the network before an ACK is required from the receiver T3 Link with 25 ms RTT, (latency = 12.5 ms) Internet
Sender with Window = 17K
#1
#2
#3
#12
* sending stops
ACK takes 12.5 ms to reach Sender
Receiver ACK
• Window is “full” after ~ 3ms; no more data can be sent till the ACK is received • Twelve packets “in flight” = ~17KB • Optimum window is obviously much greater © 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
15
The Bandwidth Delay Product (BDP) § Calculate Ideal T Window based on the Bandwidth Delay Product (BDP) • BDP = RTD * Link BW / 8 • Using the T3 example (45 Mbps link with 25 msec RTD) • BDP = 25 msec * 45 Mb/s / 8 = ~140 KBytes (Ideal T Window)
§ If T Window size = 17KBytes • would only achieve maximum of ~5.5 Mbps T throughput
§ Most applications ride on T (i.e. FTP, HTTP) • Rather than UDP • FTP and HTTP performance are directly related to proper T performance • Test T first, tune the T window, then test FTP and HTTP
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
16
DEMO 1: TrueSpeed Window Size Testing Egress Port 100 Mbps
100 Mbps
20
TrueSpeed: 4 Window Sizes
§ BDP = BW x RTT / 8 = 100 Mbps x 0.020 / 8 = ~256,000 bytes § TrueSpeed “Walk the Window” test will be conducted to demonstrate T throughput for 32 KB, 64 KB,128 KB, and 256KB T Windows
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
17
Traffic Policers, Traffic Shapers and T
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
18
Bursty vs. Constant Traffic Applications
7 6 5
4
Pr
on Applicati n esentatio Session
Bursty Traffic
Constant Traffic
HTTP, FTP, E-mail, Sharepoint, Communicator, Facebook, YouTube, Call setup
Voice and Broadband Video
T
t Transpor
UDP Traffic
3 2 1
Traffic
Network Datalink Physical
Time
© 2014 JDS Uniphase Corporation
Time
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
19
Constant Traffic vs. Burst Traffic on 1GigE Link 100Mb per second Constant Traffic on 1GigE Link 1GigE
Gap Time
Constant
Time Interval
1GigE
Burst
Still 100Mb per second, but transmitted as a burst for a fraction of the time interval Gap Time
Time Interval
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
20
Network Policers • Traffic is always sent at 1Gbps line rate • NTE with policer – 100M CIR (Committed Information Rate) and 64KB CBS (Committed Burst Size) • Policers meter on the ingress port, they don’t buffer Ø count bytes and discard frames when the SLA is violated • Common Ethernet Mobile Backhaul scenario NTE Policing at 100M CIR 64KB CBS 1Gbps Line Rate From Customer E
Ingress Port
Network side 10Gbps port CIR = 100M CBS = 64KB
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
21
Constant Traffic – 100 Mbps
Policer Config:
Tokens
CIR = 100 Mbps CBS = 64 KB
Leaky bucket algorithm 1Gbps Line Rate
1Gbps Line Rate CIR = 100 Mbps
Frame
CBS = 64 KB
Ethernet Switch or Router
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
22
Burst of Frames Equal to CBS
Policer Config:
Tokens
CIR = 100 Mbps CBS = 64 KB
1Gbps Line Rate
1Gbps Line Rate CIR = 100 Mbps
Frame
CBS = 64 KB
Ethernet Switch or Router
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
23
Burst of Frames Larger than CBS!
Policer Config:
Tokens
CIR = 100 Mbps CBS = 64 KB
1Gbps Line Rate Frame
1Gbps Line Rate
X
CIR = 100 Mbps CBS = 64 KB
Ethernet Switch or Router
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
24
Demo 2: TrueSpeed Testing into a Policer with 64KB CBS Policer: 100 Mbps CIR 64 KB CBS
Gig E
GigE
410.5 ms Delay
TrueSpeed: up to 64 T sessions: s browsing web, emailing, FTPing, etc.
• In the demo, the Near End T-BERD will send T traffic to the Far End TBERD which is essentially a T “speed test” Ø 2 T sessions, each with a 64KB window will be used to generate an aggregate 128 KB T window Ø 128KBytes = Bandwidth Delay Product (BDP) © 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
25
Traffic Shaping to Improve T Performance
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
26
What is Traffic Shaping? § Traffic shaping will buffer and queue the excess packets that are above a predefined traffic rate (i.e. CIR) Shaped WAN Traffic
Unshaped LAN Traffic Shaper: 100 Mbps
GigE
GigE
E Router
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
27
Traffic Shaping in Action
Shaper Config: CIR = 100 Mbps CBS = 64 KB
1Gbps Line Rate
1Gbps Line Rate 256 KB
64 KB
64 KB
Shaper Queue
64 KB
CIR = 100 Mbps CBS = 64 KB
64 KB
Ethernet Switch or Router
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
28
Recap: Shaper vs. a Policer
Policing
Shaping
Traffic shaping will buffer and queue the excess packets that are above a predefined traffic rate
Traffic policing will drop the excess packets that are above a predefined traffic rate
29 © 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
29
Best Practice: Shaper at Customer Edge
MPLS
LAN E Router with Shaping
CE Router with Policing
MPLS TBERD with Integrated Shaper!!
CE Router with Policing © 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
30
Shaper Results Recap
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
31
The Future: TrueSpeed Network Function Virtualization (NFV)
Service Performance Challenges 1. Consumers complain they are not getting the bandwidth they are paying for a. They run a speed test from an off-network site and get unreliable results Test #1
Finger pointing begins!
Test #2
b. Carrier SLAs cannot be correlated to customer results
Test #3
2. Different Workgroups use different methods of testing performance a. Speed Test servers, FTP servers, iPerf servers b. Lack of standards based methods
Additional finger pointing! 3. Small Business Retail customers complain they have throughput problems a. No way of correlating service performance results
Problem escalation & revenue impact!
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
33
RFC 6349 – the TrueSpeed™ Advantage SpeedTest or RFC 6349? § SpeedTest • • • •
Well known way of testing speed of a service Performance varies when server is off-network Non-standard test Results do not help troubleshoot problem
§ RFC 6349 • Standardized Test Methodology -
By standard body that created T/IP & Internet (IETF)
• Apples-to-apples verification with customer (consumer, business or retail) by all internal groups • Provides information (actionable) need to repair -
-
-
T Efficiency (Indicator of retransmissions) §
Buffer setup problems
§
Congestion
Buffer Delay (Indicator of variable latency) §
Congestion
§
Multiple paths for traffic
T Window size §
Is windows scaling turned on?
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
34
Test & Measurement – Transition to Virtualization The path to “Network Function Virtualization” will take a long time , test and measurement tools will evolve into NFV as well.
Today
CER
PER
E Ethernet NID
Tomorrow
CER
PER
TrueSpeedTM VNF SW RFC 6349
E Ethernet NID x86 Server
CER
Where it’s all headed
PER
TrueSpeedTM VNF SW RFC 6349
E Ethernet NID x86 Server © 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
35
Questions and Answers Session
36
Agenda § Who should test T? § Why Test T? § What is RFC 6349? § T Performance Issues and the TrueSpeed Solution
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
2
T Testing Applications Customer Site
T-BERD/MTS 5800 T-BERD
Ethernet Provider
Customer Aggregation
T-BERD/MTS 8000
Backhaul Network T-BERD/MTS 8000
T-BERD/MTS 6000
T-BERD/MTS 6000
1. Service providers offering Ethernet Business services 2. Service providers offering Ethernet Backhaul services 3. Mobility service providers deploying 3G or 4G/LTE © 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
3
Why is T Testing Important? § Running RFC2544, Y.1564 or other Layer 2/3 installation tests is always the first step § But even when these Layer 2/3 tests “”, end-customers can still complain that the “network is slow” and the cause of poor application performance (i.e. FTP, web browsing, etc.) § Lack of T testing is a turn-up gap because end-customer applications are transported using T in Layer 4 § Save up to 30% OPEX costs by eliminating or quickly resolving painful end-customer finger pointing scenarios
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
4
Which Applications use T?
7 6 5
on Applicati Presenta
tion
Session
4
t Transpor
3
Network
2 1
HTTP, FTP, E-mail, Sharepoint, Communicator, Facebook, YouTube, etc
Datalink Physical
T
RFC 6349 TrueSpeedTM
IP
Y.1564 SAMComplete & RFC 2544
Ethernet
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
5
Case Study: Customer Issue
100 Mbps Service ~20 msec RTT § § §
Layer 2 traffic testing was done twice, Test results = “” End-customer performance dipped to 50-60 Mbps at various times of the day (end-customer running “iperf” based speed tests) TrueSpeedTM test was run from network provider demarc in Tampa to Atlanta – T throughput was rock solid at nearly 100 Mbps (95 Mbps layer 4) NO DROPs – Finger was clearly pointed at customer data center network; investigation proceeded in the customer arena (provider was “cleared of guilt”) à larger T window was required © 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
6
Case Study: Network Provider Issue
300 Mbps Service ~75 msec RTT
§ § §
Layer 2 traffic testing was done twice, Test results = “” End-customer file transfer performance was low (150 Mbits/sec) TrueSpeedTM (30-40 T sessions with 64 KB window) stressed network buffers – Problem was reproduced with bursty traffic (traffic drop) – Network provider isolated to mis-configured device at California site à needed larger egress buffers
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
7
Brief Overview of RFC 6349
8
RFC 6349: T Throughput Test Methodology • IETF RFC 6349 specifies the methodology for measuring end-to-end T Throughput in a managed IP network. • JDSU and two (2) Network Providers co-authored this RFC “0”. Run traditional RFC2544 to the integrity of the network at Layers 2 and 3 before conducting T testing. 1. Path MTU Detection (per RFC4821) – network MTU with active T segment size testing to ensure T payload does not get fragmented 2. Baseline Round-Trip Time and Bandwidth – Predict optimum T window size to automatically calculate the T Bandwidth Delay Product (BDP) 3. Single and Multiple T Connection Throughput Tests – T window size predictions to enable automated “full pipe” T testing © 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
9
RFC 6349 Metrics: T Efficiency § T retransmissions are a normal behavior in network communications. • But what is the “efficiency” of a network transfer? • Time spent transmitting “good” payload versus retransmitting it. § The T Efficiency metric is the percentage of Bytes that did not have to be retransmitted and is defined as: T efficiency=
Transmitted Bytes - Retransmitted Bytes x 100 Transmitted Bytes
§ As an example, if 100,000 Bytes were sent and 1,000 had to be retransmitted, the T Efficiency would be calculated as: 101,000 - 1,000 101,000
x 100 = 99%
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
10
RFC 6349 Metrics: Buffer Delay § T throughput is also affected by increase in RTT, which can be caused by network congestion or buffer delay. § The Buffer Delay Percentage is defined as: Buffer Delay % = Average RTT during Transfer - Baseline RTT Baseline RTT
x 100
§ Example: If the baseline RTT for a network path is 2 ms and the average RTT increases to 3 ms during the test. The Buffer Delay Percentage would be calculated as: 3 – 2 x 100 = 50% 2
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
11
JDSU TrueSpeed = RFC 6349 Automated § Reduce operating expenses up to 30% by minimizing truck rolls § Complete fast, repeatable, automated T tests in less than 5 minutes § results with an intuitive graphical interface that is easy to use by technicians at all skill levels § Closes the testing gap! Solve the tough, costly problems that you miss with RFC 2544 and Y.1564
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
12
T Performance Issues and the TrueSpeed Solution
13
T Window Problem: Misconfigured E Host Provider Network
Customer Premise
100 M Ethernet Service 20 msec round trip Customer Laptop
“Long Fat Network” Firewall, Proxy, etc.
Problem
Description
Location / Responsibility
Window size not optimized
High latency and/or High BW links require end host T WINDOW SETTINGS TO BE OPTIMAL for full T throughput
End hosts, Firewalls, Proxies
How does TrueSpeedTM detect this?
WALK THE WINDOW test reports the actual throughput vs. End-Customer ideal for various Window Sizes
© 2014 JDS Uniphase Corporation
|
How can the problem be fixed? End-customer must tune the end host device T Window size
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
14
Latency and T Window Size Relationship § The T Window size is the amount of data that a sender will place onto the network before an ACK is required from the receiver T3 Link with 25 ms RTT, (latency = 12.5 ms) Internet
Sender with Window = 17K
#1
#2
#3
#12
* sending stops
ACK takes 12.5 ms to reach Sender
Receiver ACK
• Window is “full” after ~ 3ms; no more data can be sent till the ACK is received • Twelve packets “in flight” = ~17KB • Optimum window is obviously much greater © 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
15
The Bandwidth Delay Product (BDP) § Calculate Ideal T Window based on the Bandwidth Delay Product (BDP) • BDP = RTD * Link BW / 8 • Using the T3 example (45 Mbps link with 25 msec RTD) • BDP = 25 msec * 45 Mb/s / 8 = ~140 KBytes (Ideal T Window)
§ If T Window size = 17KBytes • would only achieve maximum of ~5.5 Mbps T throughput
§ Most applications ride on T (i.e. FTP, HTTP) • Rather than UDP • FTP and HTTP performance are directly related to proper T performance • Test T first, tune the T window, then test FTP and HTTP
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
16
DEMO 1: TrueSpeed Window Size Testing Egress Port 100 Mbps
100 Mbps
20
TrueSpeed: 4 Window Sizes
§ BDP = BW x RTT / 8 = 100 Mbps x 0.020 / 8 = ~256,000 bytes § TrueSpeed “Walk the Window” test will be conducted to demonstrate T throughput for 32 KB, 64 KB,128 KB, and 256KB T Windows
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
17
Traffic Policers, Traffic Shapers and T
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
18
Bursty vs. Constant Traffic Applications
7 6 5
4
Pr
on Applicati n esentatio Session
Bursty Traffic
Constant Traffic
HTTP, FTP, E-mail, Sharepoint, Communicator, Facebook, YouTube, Call setup
Voice and Broadband Video
T
t Transpor
UDP Traffic
3 2 1
Traffic
Network Datalink Physical
Time
© 2014 JDS Uniphase Corporation
Time
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
19
Constant Traffic vs. Burst Traffic on 1GigE Link 100Mb per second Constant Traffic on 1GigE Link 1GigE
Gap Time
Constant
Time Interval
1GigE
Burst
Still 100Mb per second, but transmitted as a burst for a fraction of the time interval Gap Time
Time Interval
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
20
Network Policers • Traffic is always sent at 1Gbps line rate • NTE with policer – 100M CIR (Committed Information Rate) and 64KB CBS (Committed Burst Size) • Policers meter on the ingress port, they don’t buffer Ø count bytes and discard frames when the SLA is violated • Common Ethernet Mobile Backhaul scenario NTE Policing at 100M CIR 64KB CBS 1Gbps Line Rate From Customer E
Ingress Port
Network side 10Gbps port CIR = 100M CBS = 64KB
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
21
Constant Traffic – 100 Mbps
Policer Config:
Tokens
CIR = 100 Mbps CBS = 64 KB
Leaky bucket algorithm 1Gbps Line Rate
1Gbps Line Rate CIR = 100 Mbps
Frame
CBS = 64 KB
Ethernet Switch or Router
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
22
Burst of Frames Equal to CBS
Policer Config:
Tokens
CIR = 100 Mbps CBS = 64 KB
1Gbps Line Rate
1Gbps Line Rate CIR = 100 Mbps
Frame
CBS = 64 KB
Ethernet Switch or Router
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
23
Burst of Frames Larger than CBS!
Policer Config:
Tokens
CIR = 100 Mbps CBS = 64 KB
1Gbps Line Rate Frame
1Gbps Line Rate
X
CIR = 100 Mbps CBS = 64 KB
Ethernet Switch or Router
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
24
Demo 2: TrueSpeed Testing into a Policer with 64KB CBS Policer: 100 Mbps CIR 64 KB CBS
Gig E
GigE
410.5 ms Delay
TrueSpeed: up to 64 T sessions: s browsing web, emailing, FTPing, etc.
• In the demo, the Near End T-BERD will send T traffic to the Far End TBERD which is essentially a T “speed test” Ø 2 T sessions, each with a 64KB window will be used to generate an aggregate 128 KB T window Ø 128KBytes = Bandwidth Delay Product (BDP) © 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
25
Traffic Shaping to Improve T Performance
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
26
What is Traffic Shaping? § Traffic shaping will buffer and queue the excess packets that are above a predefined traffic rate (i.e. CIR) Shaped WAN Traffic
Unshaped LAN Traffic Shaper: 100 Mbps
GigE
GigE
E Router
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
27
Traffic Shaping in Action
Shaper Config: CIR = 100 Mbps CBS = 64 KB
1Gbps Line Rate
1Gbps Line Rate 256 KB
64 KB
64 KB
Shaper Queue
64 KB
CIR = 100 Mbps CBS = 64 KB
64 KB
Ethernet Switch or Router
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
28
Recap: Shaper vs. a Policer
Policing
Shaping
Traffic shaping will buffer and queue the excess packets that are above a predefined traffic rate
Traffic policing will drop the excess packets that are above a predefined traffic rate
29 © 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
29
Best Practice: Shaper at Customer Edge
MPLS
LAN E Router with Shaping
CE Router with Policing
MPLS TBERD with Integrated Shaper!!
CE Router with Policing © 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
30
Shaper Results Recap
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
31
The Future: TrueSpeed Network Function Virtualization (NFV)
Service Performance Challenges 1. Consumers complain they are not getting the bandwidth they are paying for a. They run a speed test from an off-network site and get unreliable results Test #1
Finger pointing begins!
Test #2
b. Carrier SLAs cannot be correlated to customer results
Test #3
2. Different Workgroups use different methods of testing performance a. Speed Test servers, FTP servers, iPerf servers b. Lack of standards based methods
Additional finger pointing! 3. Small Business Retail customers complain they have throughput problems a. No way of correlating service performance results
Problem escalation & revenue impact!
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
33
RFC 6349 – the TrueSpeed™ Advantage SpeedTest or RFC 6349? § SpeedTest • • • •
Well known way of testing speed of a service Performance varies when server is off-network Non-standard test Results do not help troubleshoot problem
§ RFC 6349 • Standardized Test Methodology -
By standard body that created T/IP & Internet (IETF)
• Apples-to-apples verification with customer (consumer, business or retail) by all internal groups • Provides information (actionable) need to repair -
-
-
T Efficiency (Indicator of retransmissions) §
Buffer setup problems
§
Congestion
Buffer Delay (Indicator of variable latency) §
Congestion
§
Multiple paths for traffic
T Window size §
Is windows scaling turned on?
© 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
34
Test & Measurement – Transition to Virtualization The path to “Network Function Virtualization” will take a long time , test and measurement tools will evolve into NFV as well.
Today
CER
PER
E Ethernet NID
Tomorrow
CER
PER
TrueSpeedTM VNF SW RFC 6349
E Ethernet NID x86 Server
CER
Where it’s all headed
PER
TrueSpeedTM VNF SW RFC 6349
E Ethernet NID x86 Server © 2014 JDS Uniphase Corporation
|
JDSU CONFIDENTIAL AND PROPRIETARY INFORMATION
35
Questions and Answers Session
36
Related Documents 2w1qw
Jdsu Truespeed Automated Rfc-6349 T Testing 1a323f
December 2021 0
Automated Hardware Testing Using Python 5i3t2a
November 2019 35
Equipos Jdsu 6y5v6p
November 2019 70
-friendly Dashboard For Automated Testing 2j662
October 2022 0
Jdsu Hst-3000 Rfc-2544 Ethernet Testing Guide 2i2x1q
November 2019 12
Jdsu Drive-test Casestudy 3t1i1s
December 2021 0More Documents from "Takis" m3ww
504uu
November 2021 0
Jdsu Truespeed Automated Rfc-6349 T Testing 1a323f
December 2021 0
Pavis Patrick Postdramatic Theatre Pdf 2h3qb
May 2022 0
Structural Design Manual For Culverts 1i676m
November 2020 0
Project, Saltwater Dummy Load._pdf 22i61
December 2019 30