BRKNMS 3043

Advanced Performance Measurement for Critical IP Traffic with Cisco IOS IP Service Level Agreements
Michael Geller, BRKNMS-3043

ArchitectManaged and Cloud Services
July 13, 2011
BRKNMS-3043
2011 Cisco and/or its affiliates. All rights reserved.
Cisco Public
Thanks to My Co-Author!
Many thanks to Hanlin Fang, Product Manager for IP SLA for her partnership in putting this presentation together!
We want to hear from YOU!
BRKNMS-3043
Cisco Public
Rules of the Game!

Silence your phone, pda, pager, mp3 player
At CiscoLive! your evaluation is extremely important Please remember to wear your badge at all times Please visit the World of Solutions
You can ask questions any time
BRKNMS-3043
Cisco Public
Meet the Engineer

To make the most of your time at Networkers at Cisco Live 2011, schedule a Face-to-Face Meeting with top Cisco Engineers. Designed to provide a big picture perspective as well as in-depth technology discussions, these faceto-face meetings will provide fascinating dialogue and a wealth of valuable insights and ideas. Visit the Meeting Center reception desk located in the Meeting Centre in World of Solutions.
BRKNMS-3043
Cisco Public
Prerequisites
Before attending this session, familiarities with Cisco IOS IP Service Level Agreements (IP SLAs) is essential Configuration and generic features will not be covered Only new or advanced topics, as well as design recommendations will be covered
BRKNMS-3043
Cisco Public
Objectives
This session targets network performance measurement only
Understand the internals New features update Performance and scalability considerations How to get the most of IP SLAs Future and IP SLAs strategic vision
BRKNMS-3043
Cisco Public
This Is Not
An introduction to IP SLAs
Recommendations on QoS configuration A talk on backend network management applications A speculation on upcoming features A marketing document
BRKNMS-3043
Cisco Public
Agenda
Reminder
IP SLAs Accuracy Performance and Scalability New Features

Design Recommendations Get the Most Out of IP SLAs IP SLAs Initiative
BRKNMS-3043
Cisco Public
Reminder
IP SLAs in an active probing and monitoring feature in Cisco IOS Wide protocol and applications coverage: UDP, TCP, ICMP, HTTP, DNS, DHCP, FTP
Microsecond granularity Use it through SNMP or CLI
Already in Cisco IOS (available on most platforms and interfaces type)
BRKNMS-3043
Cisco Public
IP SLAs Naming History

Used to be called RTR, renamed SAA in 12.0(5)T, now the official product name is IP SLAs. The newer IP SLAs Engine 2 is a major code rewrite to improve speed and memory usage over Engine 1. Introduced initially in 12.2(15)T2, 12.3(3) and 12.2(25)S, and is therefore present in all later trains. First phase of new ip sla CLI appears originally in 12.3(14)T, next phase for 12.4(6)T. SNMP MIBs are unchanged.
The latest Engine 3 started with 15.1(1)T, currently in T-train only -- Time
Engine:
Engine 1 RTR SAA
Engine 2
Engine 3 IP SLAs
Feature Name:
CLI:
rtr
ip sla mon.
ip sla
10
BRKNMS-3043
Cisco Public
IP SLA Capability Overview

Domain
TCP/IP
UDP Jitter
VoIP
UDP Jitter (+VoIP g711, g729) VoIP RTP (DSP required) VoIP H.323 and SIP Call Setup Delay VoIP H.323 and SIP Gatekeeper Delay
Network Service
HTTP
MPLS
LSP Ping
Video
Video Operation on 3K
Metro-Ethernet
Ethernet Echo (802.1ag)
UDP Echo
UDP Path Echo TCP Connect ICMP Echo ICMP Path Echo
DNS
DHCP FTP
LSP Trace
LSP AutoDiscovery and Auto-Schedule
Ethernet Jitter
Ethernet MEP VLAN Auto-Discovery and AutoScheduling
(ECMP Tree Trace) VCCV PWE3 Echo
Y.1731 on 7600
ICMP Jitter
Core Value Features

Flexible Operation Schedule SNMP and CLI Set and Get Support
RT Threshold Alerts + Automatic Reaction Probes

QoS Integration (with Engine 3) Auto IP SLA with Endpoint Auto Discovery and Registration
Hourly Aggregate Statistics History (Up to 24hrs)

Cisco IOS, IOS-XR, and Linux Operating System Support
BRKNMS-3043
Cisco Public
11
UDP Jitter Operation

Measures the delay, delay variation (jitter), corruption, misordering and packet loss by generating periodic UDP traffic One-way results for jitter and packet-loss. If clocks are synchronized and IOS is at least 12.2(T), one-way delay is also measured. Detect and report out-of-sequence and corrupted packets
Since 12.3(4)Talso with MOS and ICPIF score for voice clarity estimation.
This operation always requires IPSLA responder
BRKNMS-3043
Cisco Public
12
UDP JitterMeasurement Example

Send Packets
P2
ST2
i1
STx = sent tstamp for packet x.
Receive packets
P1
ST1
P2 IP Core
RT2
i2
P1
IPSLA
RT1 Responder
RTx = receive tstamp for packet x.
Reflected packets
Reply to packets
P1
AT1
ATx = receive tstamp for packet x.
i4
P2
AT2
P1
RT1+d1
i3
P2
dx = processing time spent between packet arrival and treatment.
RT2+d2
Each packet contains STx, RTx, ATx, dx and the source can now calculate: JitterSD = (RT2-RT1)-(ST2-ST1) = i2-i1 JitterDS = (AT2-AT1)-((RT2+d2)-(RT1+d1)) = i4-i3
BRKNMS-3043
Cisco Public
13
Jitter CalculationBeware!
t=0
i1=20 ms
t=20
p2
lat = 55 ms
i1=20 ms
t=40
p3
lat = 50 ms
send
lat = 50 ms
p1
t=25
i2=15 ms p2
i2=25 ms p1
receive
p3
IPSLA Jitter RTP Stream Jitter
i2-i1 = +5 ms Packet too late: 5 ms
i2-i1 = -5 ms Packet on-time: 0 ms
If you count positive and negative jitter, you are penalized twice. Counting only positive jitter is enough.
BRKNMS-3043
Cisco Public
14
UDP Jitter Operation (Example)

Simulating G.711 VoIP call Use RTP/UDP ports 16384 and above, the packet size is 172 bytes (160 bytes of payload + 12 bytes for RTP) Packets are sent every 20 milliseconds Marked with DSCP value of 8 (TOS equivalent 0x20)
ip sla 1 udp-jitter 10.52.130.68 16384 \ num-packets 1000 interval 20 tos 0x20 frequency 60 request-data-size 172 ip sla schedule 1 life forever start-time now
B
A
A = 20 ms B = 20 s (1000 x 20 ms) C = 40 s (60 s 20 s)

BRKNMS-3043 Cisco Public
15
UDP Jitter Example (New CLI Phase I)

Differences Between CLIs:
rtr 1 type jitter dest-ipaddr 10.52.130.68 dest-port 16384 \ num-packets 1000 interval 20 request-data-size 172 tos 20 frequency 60 rtr schedule 1 life forever start-time now ip sla monitor 1 type jitter dest-ipaddr 10.52.130.68 dest-port 16384 \ num-packets 1000 interval 20 request-data-size 172 tos 20 frequency 60 ip sla monitor schedule 1 start-time now ip sla 1 udp-jitter 10.52.130.68 16384 \ num-packets 1000 interval 20 request-data-size 172 tos 20 frequency 60 ip sla schedule 1 life forever start-time now
BRKNMS-3043
Cisco Public
16
UDP Jitter Output

etychon-1#sh ip sla statistics 1 Round trip time (RTT) Index 1 Latest RTT: 1 ms Latest operation start time: *10:33:11.335 PST Fri Oct 7 2005 Latest operation return code: OK RTT Values Number Of RTT: 20 RTT Min/Avg/Max: 1/1/4 ms Latency one-way time milliseconds Number of Latency one-way Samples: 20 Source to Destination Latency one way Min/Avg/Max: 1/1/2 ms Destination to Source Latency one way Min/Avg/Max: 1/1/3 ms Jitter time milliseconds Number of Jitter Samples: 19 Source to Destination Jitter Min/Avg/Max: 4/4/4 ms Destination to Source Jitter Min/Avg/Max: 3/3/3 ms Packet Loss Values Loss Source to Destination: 0 Loss Destination to Source: 0 Out Of Sequence: 0 Tail Drop: 0 Packet Late Arrival: 0 Voice Score Values Calculated Planning Impairment Factor (ICPIF): 0 Mean Opinion Score (MOS): 0 Number of successes: 5 Number of failures: 3 Operation time to live: 3166 sec
BRKNMS-3043
Cisco Public
17
UDP Jitter with VoIP MOS Score

Newly introduced in Cisco IOS 12.3(4)TAdvanced feature set
Modified jitter operation reports both Mean Opinion Score (MOS) and Calculated Planning Impairment Factor (ICPIF)
Those results are estimates and should be used for comparison only and should not be interpreted as reflecting actual customer opinions
Supported Codecs:
G.711 A Law (g711alaw: 64 kbps PCM compression method) G.711 mu Law (g711ulaw: 64 kbps PCM compression method) G.729A (g729a: 8 kbps CS-ACELP compression method)
Note: this is not a real RTP voice stream, but it has the same characteristics. For real RTP stream generation, check IP SLAs VoIP RTP operation.
BRKNMS-3043
Cisco Public
18
VoIP Operation: Sample Configuration

Operation parameters autoconfigured to simulate a G729a codec
1000 packets, interval 20 ms (default values) Operation frequency will be randomized between 40 and 60 seconds
ip sla 30 udp-jitter 192.1.3.2 16001 codec g729a ip sla group schedule 30 30-31 schedule-period 1 frequency range 40-60 start-time now life forever
BRKNMS-3043
Cisco Public
19
IP SLA Video Operation at a Glance

Video Operation (VO)
VO Partners and NMS
ActionPacked , LiveAction, ServOne Collaboration Manager: for CTS traffic (March 2011) Cisco Works LMS 4.1 (Apr 2011), 4.0 Patch (~Jan 2011)
One of key IP SLAs operations Simulate real video application traffic based on application profiles Pre-packaged traffic profiles: IPTV, Tele-Presence, Video Surveillance Use case: Pre-deployment assessment Post-deployment trouble shooting Platforms: Today: 3K WIP: 4K, 6K, ISR G2 In roadmap: ASR 1K, etc
Is my network ready for 100 HD Desktop Cameras, 30 IPVSC and a new Telepresence room?
BRKNMS-3043
Cisco Public
20
Summary
IP SLAs is a Cisco IOS Feature
Active monitoring with synthetic operations sending additional traffic in the network.
Detailed results like availability, delay, loss, and jitter per direction and MOS score.
Easy to use, available on many platforms.
BRKNMS-3043
Cisco Public
21
Agenda
Reminder
IPSLA Accuracy Performance and Scalability New Features

Design Recommendations Get the Most Out of IPSLA IPSLA Initiative
BRKNMS-3043
Cisco Public
22
IPSLA AccuracyICMP Echo Probe

ICMP Echo Probe Sender Responder
(90% Process Load)
With unloaded receiver, IPSLA measures 15.0 ms

With high CPU load on the receiver: 58.5 ms!! Any System Will Report Wrong Results when Excessive CPU Time Is Spent on the Receiver Between the ICMP Echo Request and Echo Reply
Fortunately, We Have a Solution

BRKNMS-3043
Cisco Public
23
Processing Time Measurement

When running the responder, we have a clear advantage, because
A mechanism to measure the processing time spent on the receiving router is in place, inserting a timestamp when the responder receives and send the packet
Receive timestamp done at interrupt level, as soon as the packet is dequeued from the interface driver; with absolute priority over everything else
With IPSLA, this mechanism is implemented for both UDP Echo and UDP Jitter operations
BRKNMS-3043
Cisco Public
24
UDP Echo Operation (with IPSLA Responder)

T1 T2
Sender
T5 T4
T3 Responder
Processing Delay on the Source: Tps = T 5-T4 Processing Delay on the Destination: Tpd = T 3-T2
Round Trip Time Delay: T = [] = T 2 - T1 + T 4 - T3
We have no control of queuing delay on the source and destination, but this is experienced by real traffic too, and must be accounted as such
BRKNMS-3043
Cisco Public
25
IPSLA Accuracy: UDP Echo Probe

UDP Echo Probe Sender Responder
(90% Process Load)
With unloaded receiver: 15.0 ms

With 90% CPU receiver: 15.3 ms The IPSLA Responder Processing Delay Will Be Subtracted from the Final Results
BRKNMS-3043
Cisco Public
26
Absolute Accuracy Tests

To validate IPSLAs accuracy, we wanted to compare its results with another measurement device
Weve used the following topology:
Agilent RouterTester Measurement Reference
IPSLA Measurement
Cisco 7200 Cisco 7200
RouterTester Measurement
PacketStorm Impairment Generator
BRKNMS-3043
Cisco Public
27
Test Results
Release used: 12.3(7)T Advanced Enterprise on a Cisco 7200 VXR with NPE400 RouterTester and IPSLA sending packets at the same rate All results obtained for delay and jitter are in sync with Agilents result at 1 ms
Accuracy is preserved under CPU load, but spikes may happen during high-frequency interrupt events, like writing to NVRAM (write memory) Better accuracy is sometimes possible, but is dependant upon implementation details (hardware + IOS image + configuration).
BRKNMS-3043
Cisco Public
28
IPSLA Accuracy: ICMP vs. UDP

As seen beforefor RTT accuracy, always use UDP Echo or jitter with IPSLA responder Only in this case, processing time spent on the sender and responder routers will be subtracted Results more accurate regardless of the sender and receiver CPU process load
BRKNMS-3043
Cisco Public
29
Summary
IP SLAs uses a special timestamping mechanism at interrupt level and its accuracy preserved even under high CPU load The absolute tested accuracy is 1 ms. In other words, when it says 35 ms, it could be somewhere between 34 ms and 36 ms.
BRKNMS-3043
Cisco Public
30
Agenda
Reminder

BRKNMS-3043
Cisco Public
31
Cisco IOS IP SLAs Performance: CPU Load by Platform

(Jitter Probe Running Eng 2+2000 Active Jitter Oper Cisco IOS 12.4(PI3)T)
Oper/ Second 4 8 12
Pkts/ second 200 400 600
Oper/ Minute 240 480 720
2800
2811 2851 2691 3745
3845
3825
1841
3
6 8
3
5 7
1
2 3
2
3 4
1
1 2
0
1 2
2
3 5
3
4 6
16
20 24
800
1000 1200
960
1200 1440
10
13 15
9
11 13
4
4 5
5
6 8
2
3 4
2
3 4
7
8 10
8
10 11
28
32
1400
1600
1680
1920
18
20
14
16
6
7
9
10
4
5
4
5
12
14
13
15
36
40
1800
2000
2160
2400
23
24
18
20
8
9
11
12
5
6
6
6
16
17
17
18
44
48
2200
2400
2640
2880
27
29
21
21
10
11
14
15
7
7
7
8
19
21
20
22
52
56 60
BRKNMS-3043
2600
2800 3000
3120
3360 3600
32
34 36
22
22 23
Cisco Public
12
13 14
16
17 18
8
9 9
8
9 9
23
26 27
23
24 26
32
Cisco IP SLAs Performance: UDP-Jitter

UDP-Jitter Probe Running Engine 3Cisco IOS 15.1(4)M
Default Parameters: Frequency (60secs), Request Size (32bytes), Packet Interval (20ms), Number of Packets (10)
1921 Operations (Total) Operations/Second Packets Per Second

Operations/Min CPU Usage
2921 2000 33.3 333.3

2000 ~8%
3925 3000 50 500.0

3000 ~8%
3945 4000 66.7 667.0

4000 ~8%
3945E 5000 83.3 833.3

5000 ~1%
1000 16.7 166.7

1000 ~6%
Each configuration being different, use those numbers with care: they are only an indication .
No SNMP polling were performed to gather the operation results .
BRKNMS-3043
Cisco Public
33
IP SLA Performance: UDP-Jitter for VoIP

UDP-Jitter Probe for VoIP (G.729a) running Engine 3: Cisco IOS 15.1(4)M
Default Parameters: Frequency (60secs), Codec Packet Size (32bytes), Codec Interval (20ms), Codec Number of Packets (1000)
1921
Operations (Total) Operations/Second Packets Per Second Operations/Min CPU Usage 150 2.5 2500.0 150 ~59%
2921
225 3.75 3750.0 225 ~61%
3925
275 4.58 4583.3 275 ~43%
3945
400 6.7 6733.3 400 ~54%
3945E
900 15.0 15000.0 900 ~43%
Each configuration being different, use those numbers with care: they are only an indication .
No SNMP polling were performed to gather the operation results
BRKNMS-3043
Cisco Public
34
Summary
Under normal conditions and with reasonable targets, a performance issue with IP SLAs is unlikely Memory usage is reasonable, and should never be a problem on any platform. Compared to Engine 1, both performance and memory usage have been improved on IPSLA Engine 2 and 2+
BRKNMS-3043
Cisco Public
35
Agenda
Reminder

BRKNMS-3043
Cisco Public
36
Template-Based Configuration
ip sla auto template type ip udp-jitter my-jitter-template parameters request-data-size 64 num-packets 1000 ip sla auto endpoint-list type ip my-endpoint-list ip-address 10.0.0.2-3 port 5566
ip sla auto schedule my-master-scheduler frequency 45 start-time now ip sla auto group type ip my-ipsla-group schedule my-master-scheduler template udp-jitter my-jitter-template destination my-endpoint-list
BRKNMS-3043
Cisco Public
37
What operations have been created?

hub1#sh ip sla auto summary-statistics group type ip my-ipsla-group IP SLAs Auto Group Summary Statistics Legend sno: Serial Number in current display oper-id: Entry Number of IPSLAs operation type: Type of IPSLAs operation n-rtts: Number of successful round trips in current hour of operation rtt (min/av/max): The min, max and avg values of latency in current hour of operation avg-jitter(DS/SD): average jitter value in destination to source and source to destination direction pak-loss: accumulated sum of source to destination and destination to source packet loss in current hour Summmary Statistics:
Auto Group Name: my-ipsla-group Template: my-jitter-template Number of Operations: 2 sno oper-id type n-rtts rtt avg-jitter packet (min/avg/max) (DS/SD) loss 1 1058464225 udp-jitter 732 1/1/5 ms 1/1 ms 0 2 1894530068 udp-jitter 3419 1/1/162 ms 1/1 ms 0
BRKNMS-3043
Cisco Public
38
QoS Integration (Example)

Observation: Need to send the same operation in each class. Problem: Provision the same operation multiple times is lengthy, error prone,
and counter productive. Solution: Discover the QoS classes on the outgoing interface and automatically instantiate probes.
class-map voice-traffic match dscp EF
QoS Class definition

class-map data-traffic match dscp AFnn
policy auto-measure class voice-traffic measure type ip-sla group voice-traffic-probes-grp class data-traffic measure type ip-sla group udp-jitter-probes-grp
How to measure in each class?

39
BRKNMS-3043
Cisco Public
End-Point Auto Registration

ip sla auto endpoint-list type ip my-ep-list-auto Ip sla auto discovery Ip sla auto group type ip my-group-auto schedule my-master-schedule template udp-jitter my-jitter-template destination my-ep-list-auto Hub
Hub to Spoke-1 ip sla operation-ID-1 udp-jitter 10.10.10.2 5000 Hub to Spoke-2 ip sla operation-ID-2 udp-jitter 20.20.20.2 5000 Hub to Spoke-3 ip sla operation-ID-3 udp-jitter 30.30.30.2 5000
172.17.0.5
spoke-3
30.30.30.2 10.10.10.2
ip sla responder auto-register 172.17.0.5 endpoint-list my-ep-list-auto
spoke -1
20.20.20.2 spoke-2 ip sla responder auto-register 172.17.0.5 endpoint-list my-ep-list-auto
ip sla responder auto-register 172.17.0.5 endpoint list myep-list-auto

BRKNMS-3043
Cisco Public
40
Registered endpoints
hub1#sh ip sla auto group Group Name: my-ipsla-group-auto Description: Activation Trigger: Immediate Destination: my-endpoint-list-auto Schedule: my-master-scheduler Measure Template: my-jitter-template(udp-jitter) IP SLAs auto-generated operations of group my-ipsla-group-auto sno oper-id type dest-ip-addr/port 1 1400050412 udp-jitter 10.10.10.2/5000 2 1584779241 udp-jitter 20.20.20.2/5000 3 1930415937 udp-jitter 30.30.30.2/5000
BRKNMS-3043
Cisco Public
41
New Operation in IP SLA: VO

Uses
Availability
Network Performance Monitoring
VoIP Monitoring
Service Level Agreement (SLA) Monitoring
Network Assessment
Multiprotocol Label Switching (MPLS) Monitoring
Trouble Shooting
Measurement Metrics
Latency Packet Loss Network Jitter Dist. of Stats Connectivity
Operations
Jitter FTP DNS DHCP DLSW ICMP UDP TCP HTTP LDP H.323 SIP RTP
Video
Defined Packet Size, Spacing COS and Protocol

Cisco IOS Software IP SLA Source
IP Server
IP Server MIB Data Active Generated Traffic to Measure the Network

Destination Cisco IOS IP SLA Software Responder
42
Cisco IOS Software IP SLA

BRKNMS-3043
Cisco Public
IPSLA Video Operation

Convenient for pre-deployment assessment, pre-event testing and post-event troubleshooting.
Is my network ready for 100 HD Desktop Cameras, 30 IPVSC and a new Telepresence room? More bandwidth needed? Deploy PfR? QoS needed?
Fully integrated with IPSLA control and scheduling framework Extension to current IPSLA CLI and MIB interface to allow easy integration with NMS products
Switch D
Router B Switch A
BRKNMS-3043
Router C
Cisco Public
43
IPSLA Video Operation Embedded Traffic Simulator
New
IPSLA known in industry for jitter, ICMP, etc. probes Most probes measure experience without affecting user traffic (hopefully) Need traffic to stress test network
X
IPSLA VO provides Realistic representation of arbitrary video (RTP) traffic Packet sizes, burstiness, traffic rate, etc. pre-packaged profiles: IPTV, Video Surv, CTS Extensible via data file Custom profile generation from packet capture
Cisco Public
BRKNMS-3043
44
Pre-Deployment Planning
Objective
Enable clientless deployment and capacity planning - How many streams at bandwidth x at this time of day can we expect to support - What delay/loss impact does the addition of an extra stream at bandwidth X
Solution Value
Clientless pre-deployment and provisioning for network readiness assessment and traffic modeling
Remote Site
Operations System (OSS) or Application

BRKNMS-3043
Cisco Public
45
Video Configuration: Video Op. (Cont.)

router(config)#
ipsladev3750e-3(config-ip-sla)#video ? Hostname or A.B.C.D Destination IP address or hostname ipsladev3750e-3(config-ip-sla)#video 192.168.1.4 ? <1-65535> Port Number
ipsladev3750e-3(config-ip-sla)#video 192.168.1.4 4336 ? source-ip Source address ipsladev3750e-3(config-ip-sla)#video 192.168.1.4 4336 source-ip 192.168.1.3 ? source-port Source Port
The required parameters for video
BRKNMS-3043
Cisco Public
46
Video Configuration: Video Op. (Cont.)

router(config)#
ipsladev3750e-3(config-ip-sla)#$6 source-ip 192.168.1.3 source-port 3228 ? profile traffic profile type to be configured
ipsladev3750e-3(config-ip-sla)#$p 192.168.1.3 source-port 3228 profile ? IPTV IP Television traffic (2.6 Mbps) IPVSC IP video surveillance camera traffic (2.2 Mbps) TELEPRESENCE Cisco Telepresence 1080P traffic (6.6Mbps)
ipsladev3750e-3(config-ip-sla)#$p 192.168.1.3 source-port 3228 profile IPVSC
BRKNMS-3043
Cisco Public
47
IP SLA Video Show Configuration

ipsladev3750e-3#show ip sla configuration 111 IP IP SLAs Infrastructure Engine-III Entry number: 111 Owner: Tag: Operation timeout (milliseconds): 5000 Type of operation to perform: video Video profile name: IPVSC Video duration (seconds): 20 DSCP: cs5 Target address/Source address: 192.168.1.4/192.168.1.3 Target port/Source port: 4336/3228 Vrf Name: Control Packets: enabled Schedule: Operation frequency (seconds): 900 (not considered if randomly scheduled) Next Scheduled Start Time: Start Time already passed Group Scheduled : FALSE Randomly Scheduled : FALSE Life (seconds): 3600 Entry Ageout (seconds): never Recurring (Starting Everyday): FALSE Status of entry (SNMP RowStatus): Active Threshold (milliseconds): 5000 Distribution Statistics: Number of statistic hours kept: 2 Number of statistic distribution buckets kept: 1 Statistic distribution interval (milliseconds): 20 Enhanced History:
BRKNMS-3043
Cisco Public
48
IP SLA Video Show Statistics

ipsladev3750e-1#show ip sla statistics 1 IPSLAs Latest Operation Statistics
IPSLA operation id: 1 Type of operation: video Latest operation start time: 10:50:53 PST Fri Feb 25 2011 Latest operation return code: OK Packets: Sender Transmitted: 2034 Responder Received: 1994 Latency one-way time: Number of Latency one-way Samples: 1894 Source to Destination Latency one way Min/Avg/Max: 0/29/31 milliseconds NTP sync state: SYNC Inter Packet Delay Variation, RFC 5481 (IPDV): Number of SD IPDV Samples: 1847 Source to Destination IPDV Min/Avg/Max: 0/1/3 milliseconds Packet Loss Values: Loss Source to Destination: 60 Out Of Sequence: 33 Number of successes: 1 Number of failures: 0 Operation time to live: 3578 sec
BRKNMS-3043
Cisco Public
49
Debugging Commands & Steps

Available DEBUG commands:
Sender debug
ipsladev3750e-3#debug ip sla trace ipsladev3750e-3#debug ip sla error
This will show both IP SLA debug and platform debug. Basic familiarity with IP SLA debug is expected 3K platform video debug will also be enabled
Responder Debug
ipsladev3750e-3#debug ip sla trace 0 ipsladev3750e-3#debug ip slaerror 0
BRKNMS-3043
Cisco Public
50
IP SLAsMPLS Health Monitor

Automatically create and delete IP SLAs LSP ping or LSP traceroute operations based on network topology Works on the MPLS L3 layer, under the IP layer. Discovers MPLS issues even when IP routing is working ok. Dramatically reduces troubleshooting time, and cost associated to maintenance of MPLS networks.
Other PEs are discovered using BGP next-hop, and operations configured accordingly.
Requires 12.2(27)SBC and later. New capability for Metro Ethernet on 7600: Y.1731
BRKNMS-3043
Cisco Public
51
Cisco IP SLA For Metro Ethernet

Performance Management
Cisco IP SLA
IP and MPLS
IP SLA for Metro Ethernet: Echo Probe Jitter Probe
NEW 15.1(2)S
CFM
ETH-LM LMM Probe ETH-DM DMM Probe ETH-DM 1DM Probe
Y.1731
Embedded Policy Management
Cisco IP SLAs Embedded Policy Management

Scheduling Automation / Policy Alerts / Data Collection
In-band Performance Management Tool for Ethernet

Delay, Delay Variation and Packet Loss measurement Built in CFM principles
Automatic Discovery of Probe Endpoints

Using entries on CFM CCM database
BRKNMS-3043
Cisco Public
52
Y.1731 Performance Monitoring

Supported Features
Two-way delay measurement One-way delay measurement Single-ended loss measurement
Unsupported Features
Dual-ended loss measurement
Meant only for point-to-point scenarios

Allows per CoS delay or loss measurements BRKSPG-2202 E-OAM session covering Ethernet performance mgt
BRKNMS-3043
Cisco Public
53
Cisco Performance Management Solution Strategy
Refer to url below for the following topics

http://wwwin.cisco.com/ios/tech/collateral/EOAM-modulePerformance.ppt
IP SLAs for Metro Ethernet Overview IP SLAs for Metro Ethernet and Y.1731 PM comparison
BRKNMS-3043
Cisco Public
54
Agenda
Reminder

BRKNMS-3043
Cisco Public
55
Reasonableless Test
Dont overdo it, your metrics must be:
Attainable Measurable
Relevant Controllable Mutually Acceptable
Understandable Cost Effective
Use a limited but relevant number of indicators. Better is the enemy of good: good is good enough.
BRKNMS-3043
Cisco Public
56
Typical SLA Requirements

Traffic Type
Maximum Packet Loss Maximum One-Way Latency
Max. Jitter
VoIP
(land line quality)
1%
120 ms
30 ms
Videoconferencing Streaming video (one way video)
1%
200 ms
50 ms
2%
5s
N/A
(assuming the receive buffer is large enough)
BRKNMS-3043
Cisco Public
57
Real-Time vs. Periodic Reporting

Real-Time Reporting Periodic Reporting
Confirmation of status
Historical reports
Potential problems Notification Nature of problem
Objectives vs. Estimates Anticipation: potential impact, things to avoid

Change in service levels
BRKNMS-3043
Cisco Public
58
Cisco IOS IPSLA Uses and Metrics

Data Traffic Minimize packet loss Maximize bandwidth Verify Quality of Service (QoS) Packet loss Latency per QoS
VoIP
Minimize delay, packet loss, jitter
Service Level Agreement Measure delay, packet loss, jitter One-way
Availability
Connectivity testing
Streaming Video Minimize delay, packet loss
Requirement
Jitter Packet loss Latency MOS Voice Quality Score
Jitter Packet loss Latency One-way Enhanced accuracy NTP
Connectivity tests to IP devices
Jitter Packet loss Latency
IPSLAs Measurement
BRKNMS-3043
Cisco Public
59
Video Per-Application Latency, Jitter, Loss Targets

General Guidelines
Application Streaming Video Video Conferencing TelePresence
Latency < 1000ms < 150 ms < 150 ms
Jitter < 100 ms <30 ms < 10 ms
Loss (VoD) < 0.1% NA NA
Loss (Live) < 0.05% < 0.10% < 0.05%
Digital Signage IPTV Video Surveillance
< 1000 ms < 1000 ms < 1000 ms
< 100 ms < 100 ms < 100 ms
< 0.1% < 0.1% < 0.1%
0% 0% < 0.05%
for reference
BRKNMS-3043
Cisco Public
60
Class of Service
One operation instance to measure each class of service Same operation type for all classes Traffic coloring from within IP SLA with TOS/DSCP configuration
Bear in mind the corner case with locally generated and colored traffic on some distributed platforms Workaround is to use a Shadow Router
BRKNMS-3043
Cisco Public
61
Why Use a Shadow Router?

A shadow router is a dedicated box for IP SLAs. But why? If your Provider Edge (PE) router is already overloaded (> 60% CPU at interrupt level) If your PE lacks memory
If your PE is a distributed platform If you want to separate measurement from forwarding Upgrade freely for the latest and greatest IP SLA features without disturbing the traffic, then Use a shadow router (router dedicated to IPSLA)
BRKNMS-3043
Cisco Public
62
Shadow Router Configuration

A shadow router is typically a dedicated router located near a ideal measurement point.
A point-of-presence (POP) is an ideal location.
Point-of-Presence Shadow PE
It can be connected to the PE via various methods: direct IP connection, tunnels, dot1q
CE
BRKNMS-3043
Cisco Public
63
How to Probe?
Full mesh
Full mesh between same-customer CPEs Partial mesh Composite SLAs
BRKNMS-3043
Cisco Public
64
Full Mesh
Nodes Operation
2
3
1
3
4 5 6
7
6 10 15
21
2 n
8
100
28
4950
Number of operations is proportional to the square of the number of nodes

Does not scale
BRKNMS-3043
Cisco Public
65
Full Mesh CE-to-CE [Example]
CE PE
CE
Core
PE
PE
CE
Accurate: direct measurement from end-to-end, best user-perspective view Expensive: for n nodes, requires n(n-1)/2 operations In certain cases, it might be difficult to poll the results with SNMP on the CE
BRKNMS-3043
Cisco Public
66
Partial Mesh
London
Amsterdam San Jose
Full mesh is not always desirable, while partial mesh dramatically reduces the number of operations.
Measurement points can be based on traffic matrix, traffic importance For instance, try a coverage objective for 80% of the traffic
Raleigh
Paris
Brussels
To build a traffic matrix, use NetFlow.

Cisco Public
BRKNMS-3043
67
Composite SLA for Delay [Example]
CE
CE
PE
Core
PE
PE
Easy: Total delay can be easily calculated by adding the measured delay along the path CE Flexible: You can split the measurement for Core Edge, and total
Measurements are less accurate, as each measurement carry its own error tolerance (typically 1 ms per measurement)
BRKNMS-3043
Cisco Public
68
Composite SLA for Packet Drop [1/2]

A trivial solution might is to consider the sum of drop probabilities; this is conservative A more accurate approach is to invert the probability of a successful packet delivery If x is the loss probability across section x, then the total loss probability is:
1... x = 1 - [(1 - 1 ).(1 - 2 )

BRKNMS-3043
(1 - n )]
69
Cisco Public
Composite SLA for Packet Drop [2/2]

Example: We Have Three Sections with Various Drop Probabilities:
r1
1 = 0.05
r2
2 = 0.06
r3
3 = 0.12
r4
First solution (approximation): 0.05+0.06+0.12=0.23 (23%) Second solution (exact): 1-[(1-0.05)x(1-0.06)x(1-0.12)]=0.21416 (21.4%)
BRKNMS-3043
Cisco Public
70
Composite SLA for Jitter

2 ms 4 ms
Jitter = 2 + 4 + 3 = 9 ms ?
3 ms
Can We Add a Jitter Value to a Jitter Value?
Short answer: No! This is not a valid approach to calculate total jitter based on measured jitter, because we dont know how to do it (jitter is not additive)
Too many factors: positive jitter, negative jitter, percentile-95 of jitter, average jitter,
Youd better measure it, not calculate it
BRKNMS-3043
Cisco Public
71
Summary
PE-PE, PE-CE or CE-CE, full-mesh or partial-mesh is all your decision! IPSLA can run on almost any existing Cisco router. When this is not possible/desirable then a shadow router is recommended Composite SLAs are a good idea while end-to-end jitter results are not required
BRKNMS-3043
Cisco Public
72
Agenda
Reminder

BRKNMS-3043
Cisco Public
73
Common Questions
How should I configure my operations to accurately measure jitter/delay/packet loss? How many packets should be sent per operation? How frequently? What percentage of by bandwidth should be dedicated for measurement?
BRKNMS-3043
Cisco Public
74
Spectrum of Test
This is the proportion of time during which the network is under test
A small spectrum of test means a small probability of catching an event

For example: running a test for 20 seconds every 60 seconds is equivalent to a 33% spectrum of test
BRKNMS-3043
Cisco Public
75
Spectrum of Test
This Event Was Missed
Network Is Under Test
Delay
Time
BRKNMS-3043
Cisco Public
76
Spectrum of Test
Fault Is Detected
Network Is Under Test
Delay
Time
BRKNMS-3043
Cisco Public
77
Number of Packets
The more packets sent:
The larger the population The more diluted are the results
At identical frequency, the longer the operation, and the wider the test spectrum. Example of result dilution with the same spectrum, but a bigger number of packets per operation.
Non-diluted:
Diluted:
BRKNMS-3043
Cisco Public
78
Frequency
The operation frequency, as well as operation duration, have a direct impact on the Spectrum of Coverage
Increasing the frequency will increase your spectrum of coverage, and increase the bandwidth consumed but will not change the accuracy
BRKNMS-3043
Cisco Public
79
Interval
The interval is the space between two consecutive probe packets
Long intervals (hundreds of ms) are for trends, and will lead to higher jitter results
Short intervals (low tens of ms) are for very precise measurement, limited in time; the jitter is expected to be smaller in that case
BRKNMS-3043
Cisco Public
80
Interval Effect of Jitter

Longer Intervals Ultimately Measures Bigger Jitter, Because of Coarse Granularity:
Delay
Time
Jitter
BRKNMS-3043
Cisco Public
81
Interval Effect of Jitter

Shorter Intervals Measurements Are More Granular, and Hence Give Less Jitter:
Delay
Time
Jitter
BRKNMS-3043
Cisco Public
82
Interval and Jitter

Compare different jitter measurements only if the measurement intervals are identical Short interval is more accurate, but more expensive: use it occasionally to have a true application-like jitter Long interval is less accurate, but consumes less bandwidth: use it to expand your test spectrum and keep an eye on your jitter trends
BRKNMS-3043
Cisco Public
83
Packet Size
The main effect of packet size is to modify the Serialization Delay
On fast links, this is negligible compared to the propagation delay, so the packet size has little or not effect but to consume bandwidth
Use small packets of fast links, like on core network
Use realistic packets for low-speed access links, where the serialization delay is a factor we need to count
BRKNMS-3043
Cisco Public
84
Summary
The Design Will Have to Accommodate Some Tradeoffs, You Can Choose Two Out of Three:
Low Bandwidth Consumption
High Measurement Accuracy
Large Spectrum of Coverage
BRKNMS-3043
Cisco Public
85
Agenda
Reminder

Design Recommendations Get the Most Out of IPSLA IPSLA Initiative and Roadmap
BRKNMS-3043
Cisco Public
86
IP SLA Partners
Cisco Network Management Solutions
MPLS Diagnostics Engine (MDE) Performance Visibility Manager (PVM) MPLS Network Monitoring Network Performance
Unified Service Monitor (USM)

Unified Operations Manager (UOM) Internetworking Performance Monitor (IPM)
Unified Communications - Service Quality

Unified Communications - VoIP Monitoring Enterprise Performance Monitoring
Partner
SolarWinds
Product
Orion VoIP Monitor
Strategy
SMB/ Mid-Market
InfoVista
Fluke (Crannog, Visual, +1) NetQoS
VistaView
Response Watch, Enterprise Monitor NetVoyant
SP / Enterprise
Enterprise Enterprise
CA
Wired City IBM/Proviso
E-Health
IT Monitor
SP / Enterprise
Industrial/Vertical SP
BRKNMS-3043
Cisco Public
87
Smarter NetworksCisco ANY SLAs

-Application and Business aware remediation - Intelligent Network Service operations, benefit business-critical applications - Network bottleneck identification, reduce network downtime - Invoke notification or correction actions
Preemptive (in the future)
Business Value
- Network Availability assessment, reduce deployment time - Deploy new applications and services with complete confidence
Proactive
Predictive
- End-2-end network performance Visibility - Network Health Awareness
Reactive
Operational Excellence
BRKNMS-3043
Cisco Public
88
ANY SLAs 4.0 Feature Planning

PI 17
SLAsVO
ISR G2 with DSP Pre-defined traffic profile One-way delay, jitter, and packet loss Integration with MT via EEM RTP traffic simulation
PI 18
SLAs 4.0
IPv6* on Operations, Engines, MIBs, and APIs Feature Enhancements: autodest port, IP Addr based report, etc
PI 18
SLAs Responder 4.0
PI19
SLAs4.1
Responder on Linux OS Package Responder Licensing and User Guide
Multicast support on Udp jitter VRF support on IPv6
Y1731 Throughput
Define CLIs to enable ondemand throughput testing
Extensible SLAs 4.0

Open API for new probe generation (phase I) API User Guide Intelligent Responder with device info measurement
SLAs 4.1 VO (CP-xxxxx)

MSI VO Responder Support VO integration with MT VO IPv6 Support VO multicast support VO Signaling support VO TCP traffic support
SLAs X-OS 4.0

IPSLAs on Nexus 7K
Roadmap Items are subject to change,
BRKNMS-3043
Cisco Public
89
References
Cisco IOS IPSLA home page
http://www.cisco.com/go/ipsla
For questions related to Cisco IP SLAs that cannot be handled by the Technical Assistance Center (TAC), feel free to write an email to:
ask-ipsla@cisco.com
BRKNMS-3043
Cisco Public
90
Summary and Conclusion

IPSLA is a Cisco IOS feature available today to actively measure and report many network metrics. It is easy to use, and is supported by many existing network management applications. We also have MPLS OAM, Gatekeeper Registration, H323/SIP Call Setup operation, and many other new features.
Stay tuned. We have an ambitious roadmap for new features like better voice measurements, multicast, Ethernet OAM and were always listening your suggestions!
BRKNMS-3043
Cisco Public
91
Q&A
BRKNMS-3043
Cisco Public
92
Visit the Cisco Store for Related Titles http://theciscostores.com
BRKNMS-3043
Cisco Public
93
Recommended Reading
ISBN 1-58705-198-2
BRKNMS-3043
ISBN 0-12370-549-5
Cisco Public
94
Complete Your Online Session Evaluation

Receive 25 Cisco Preferred Access points for each session evaluation you complete. Give us your feedback and you could win fabulous prizes. Points are calculated on a daily basis. Winners will be notified by email after July 22nd.
Complete your session evaluation online now (open a browser through our wireless network to access our portal) or visit one of the Internet stations throughout the Convention Center.
Dont forget to activate your Cisco Live and Networkers Virtual account for access to all session materials, communities, and on-demand and live activities throughout the year. Activate your account at any internet station or visit www.ciscolivevirtual.com.
BRKNMS-3043
Cisco Public
95
NMS Sessions Offered

Session Monday: Title
(1 of 2)
BRKNMS-1942 Managing Infrastructure as a Service (IaaS) for Cloud Environment BRKNMS-2032 Rapid and Repeatable Service Delivery Through Automation BRKNMS-3021 Advanced Cisco IOS Device Instrumentation Tuesday: BRKNMS-1032 Network Management KPI's Introduction to Network Performance Measurement with Cisco IOS BRKNMS-1204 IP Service Level Agent BRKNMS-1532 Introduction to Accounting Principles with NetFlow and NBAR BRKNMS-2031 SYSLOG Design, Methodology and Best Practices
BRKNMS-2501 Enterprise QoS Deployment, Monitoring and Management
BRKNMS-1800 Enhancing Troubleshooting with Embedded Automation

BRKNMS-3043
Cisco Public
96
NMS Sessions Offered (2 of 2)

Session
Wednesday:
Title
BRKNMS-2031 SYSLOG Design, Methodology and Best Practices BRKNMS-2035 Ten Cool LMS Tricks to Better Manage Your Network BRKNMS-2499 Operating and Managing Converged Enterprise Architectures
BRKNMS-2640 Advanced DHCP and DNS Deployments
Advanced Performance Measurement for Critical IP Traffic with BRKNMS-3043 Cisco IOS IP Service Level Agreements
Thursday: BRKNMS-2006 Energy Management BRKNMS-2030 Onboard Automation with Cisco IOS Embedded Event Manager BRKNMS-2658 Securely Managing Your Networks and SNMPv3 BRKNMS-3132 Advanced NetFlow BRKNMS-1035 The NOC at CiscoLive
BRKNMS-3043
Cisco Public
97
BRKNMS-3043
Cisco Public
98
Thank you.
BRKNMS-3043
Cisco Public
99

BRKNMS 3043

Încărcat de

Informații document

Descriere originală:

Titlu original

Drepturi de autor

Formate disponibile

Partajați acest document

Partajați sau inserați document

Opțiuni de partajare

Vi se pare util acest document?

Este necorespunzător acest conținut?

Drepturi de autor:

Formate disponibile

BRKNMS 3043

Încărcat de

Drepturi de autor:

Formate disponibile

Advanced Performance Measurement for Critical IP Traffic with Cisco IOS IP Service Level Agreements

Michael Geller, BRKNMS-3043

2011 Cisco and/or its affiliates. All rights reserved.

We want to hear from YOU!

2011 Cisco and/or its affiliates. All rights reserved.

Rules of the Game!

2011 Cisco and/or its affiliates. All rights reserved.

Meet the Engineer

2011 Cisco and/or its affiliates. All rights reserved.

2011 Cisco and/or its affiliates. All rights reserved.

2011 Cisco and/or its affiliates. All rights reserved.

2011 Cisco and/or its affiliates. All rights reserved.

IP SLAs Accuracy Performance and Scalability New Features

2011 Cisco and/or its affiliates. All rights reserved.

Microsecond granularity Use it through SNMP or CLI

Already in Cisco IOS (available on most platforms and interfaces type)

2011 Cisco and/or its affiliates. All rights reserved.

IP SLAs Naming History

Engine 1 RTR SAA

IP SLA Capability Overview

(ECMP Tree Trace) VCCV PWE3 Echo

Core Value Features

RT Threshold Alerts + Automatic Reaction Probes

Hourly Aggregate Statistics History (Up to 24hrs)

2011 Cisco and/or its affiliates. All rights reserved.

UDP Jitter Operation

2011 Cisco and/or its affiliates. All rights reserved.

UDP JitterMeasurement Example

STx = sent tstamp for packet x.

dx = processing time spent between packet arrival and treatment.

IPSLA Jitter RTP Stream Jitter

i2-i1 = +5 ms Packet too late: 5 ms

i2-i1 = -5 ms Packet on-time: 0 ms

2011 Cisco and/or its affiliates. All rights reserved.

UDP Jitter Operation (Example)

A = 20 ms B = 20 s (1000 x 20 ms) C = 40 s (60 s 20 s)

2011 Cisco and/or its affiliates. All rights reserved.

UDP Jitter Example (New CLI Phase I)

UDP Jitter Output

UDP Jitter with VoIP MOS Score

2011 Cisco and/or its affiliates. All rights reserved.

VoIP Operation: Sample Configuration

2011 Cisco and/or its affiliates. All rights reserved.

IP SLA Video Operation at a Glance

2011 Cisco and/or its affiliates. All rights reserved.

Easy to use, available on many platforms.

2011 Cisco and/or its affiliates. All rights reserved.

IPSLA Accuracy Performance and Scalability New Features

2011 Cisco and/or its affiliates. All rights reserved.

IPSLA AccuracyICMP Echo Probe

(90% Process Load)

With unloaded receiver, IPSLA measures 15.0 ms

Fortunately, We Have a Solution

Processing Time Measurement

2011 Cisco and/or its affiliates. All rights reserved.

UDP Echo Operation (with IPSLA Responder)

2011 Cisco and/or its affiliates. All rights reserved.

IPSLA Accuracy: UDP Echo Probe

(90% Process Load)

With unloaded receiver: 15.0 ms

2011 Cisco and/or its affiliates. All rights reserved.

Absolute Accuracy Tests