11 December, 20001 Transport Performance Metric MIB draft-ietf-rmonmib-tpm-mib-01.txt Russell Dietz RMONMIB Working Group 49 th IETF
Post on 19-Jan-2018
216 Views
Preview:
DESCRIPTION
Transcript
11 December, 2000
1
Transport Performance Metric MIB
draft-ietf-rmonmib-tpm-mib-01.txtRussell Dietz
rdietz@hifn.comRMONMIB Working Group
49th IETF – San Diego, CA 2000
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
2
Application Performance…
Client Server
Work Request
Work Result
?
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
3
Why transport metrics?
Client Server
Work Request
Work Result?
Sub RequestSub RequestSub Request
Sub Request
Sub ResultSub Result
Sub ResultSub Result
Trouble-Trouble-shootingshooting
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
4
Web (HTTP) Example
Client Server
HTTP Connect
HTTP ACK
*form*.jdbc-sybase
*order-entry*finish
*form*.jdbc-sybase
re-directedfinish-ack
entry-ackform graphic
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
5
NetMeeting (H.323) Example
Client Server
H.323 Connect
H.323 (Q.931)
Q.931 (Set)H.245 (start)H.245 (alive)
H.245 (alive)
rtsp (audio)rtsp (video)
H.245 (video)Q.931 (H.245)
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
6
Categories of Metrics Jitter Exchange
Response Application
Response
Connection Connection
Sequence Connection
Window Routing
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
7
Jitter Metrics Message Jitter…
starts with the abstraction of a communicated message… a sequence of adjacent data packets transferred in a given direction, concluded by the transfer of one or more adjacent data packets in the other direction.
measures the inter-packet gaps only for data packets within messages. It does not measure the gaps between the messages themselves.
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
8
Message Jitter… example, an application that is downloading a
series of large graphics. The client first requests a graphic file download from the server. The server then sends the graphic to the client. The client then requests the next graphic, and so on. Within each download, the observed inter-packet gaps are of interest. The inter-packet gap from the last packet of one download and the first packet of the next is of less interest These later gaps are excluded.
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
9
Message Jitter
Data Data Data Data
Data Data Data Data
Server -> Client Data Message
Data
Client -> Server Data Msg
* **
Client->Server Inter-pkt Gaps
Client -> Server Data Message
Metric Units Description
N Inter-Packet Gaps Count of the # of Inter-Packet Gaps measured for Data from the Client to the Server
(x) uSec Sum total of the Delta Times in these Inter-Packet Gaps
(X2) uSec2 Sum total of the Delta Times squared in these Inter-Packet Gaps
Xmax uSec The maximum Delta Time of Inter-Packet Gaps measured
Xmin uSec The minimum Delta Time of Inter-Packet Gaps measured.
(IX) Sum total of the data point # times the Delta Times in these Inter-Packet Gaps
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
10
Jitter Metrics Stream Jitter…
does not include message considerations. Rather, measures the inter-packet gaps for all packets of the data stream for a given transfer direction.
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
11
Stream Jitter… example is where an H.323 Voice-over-IP
(VoIP) data-stream is transferring an audio data-stream over RTP from one end-point to another. In this case, all inter-packets gaps observed would be of interest.
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
12
Stream Jitter
Data Data Data Data
Data Data Data Data
Server -> Client Data Message
Data
Client -> Server Data Msg
* **
Client->Server Inter-pkt Gaps
Client -> Server Data Message
Metric Units Description
N Inter-Packet Gaps Count of the # of Inter-Packet Gaps measured for Data from the Client to the Server
(x) uSec Sum total of the Delta Times in these Inter-Packet Gaps
(X2) uSec2 Sum total of the Delta Times squared in these Inter-Packet Gaps
Xmax uSec The maximum Delta Time of Inter-Packet Gaps measured
Xmin uSec The minimum Delta Time of Inter-Packet Gaps measured.
(IX) Sum total of the data point # times the Delta Times in these Inter-Packet Gaps
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
13
H.323 – Stream Jitter
Client Server
H.323 Connect
H.323 (Q.931)
Q.931 (Set)H.245 (start)H.245 (alive)
H.245 (alive)
rtsp (audio)rtsp (video)
H.245 (video)Q.931 (H.245)
Stream Jitter
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
14
Response Metrics Exchange Responses…
starts with the abstraction of a communicated exchange-message. An exchange-message is considered to start with a series of adjacent data packets transferred in a given direction (exchange message request). The end of such a message is defined to be the transfer of one or more adjacent data packets in the other direction (exchange message response). Matters of sequencing or retransmission (duplicate) detection are excluded from consideration in the analysis of data packets when observing exchange-message related packets.
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
15
Exchange Response Example
CS CS CS
SC SC SC
LastCS
Data
CS E-SERT
CS S-S ERT
CS CS CS
SC SC SC
ChangeDir
CS->SC
CS S-E ERT
StartCS
Data
SC S-S ERT
SC E-SERT
CS
LastSC
Data
ChangeDir
SC->CS
CS E-SERT
CS S-S ERT
SC S-E ERT
LastCS
Data
StartSC
Data
StartCS
Data
ChangeDir
SC->CS
LastSC
Data
CS S-E ERT
SC S-S ERT
SC E-SERT
ChangeDir
CS->SC
StartSC
Data
StartCS
Data
DefaultDir
(C->S)
Req. Network Delay+
Server DelayServer Delay
Req. Network Delay+
Server Delay+
Resp. Network Delay
Client Delay(to nextrequest)
Exchange Message #1 Exchange Message #2
LEGENDCS = Client->Server Data PacketSC = Server->Client Data Packet
CS S-S ERT = Client->Server Start-to-Start Exchange Response TimeCS E-S ERT = Client->Server End-to-Start Exchange Response TimeCS S-E ERT = Client->Server Start-to-End Exchange Response TimeSC S-S ERT = Server->Client Start-to-Start Exchange Response TimeSC E-S ERT = Server->Client End-to-Start Exchange Response TimeSC S-E ERT = Server->Client Start-to-End Exchange Response Time
Req. Network Delay = SUM(CS S-S RT) - SUM(CS E-S RT);Resp. Network Delay = SUM(CS S-E RT) - SUM(CS S-S RT);
Server Overhead = SUM(CS E-S RT);
Client Overhead = SUM(SC E-S RT);
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
16
Exchange ResponseStart to Start
Data Data Data Data
Data Data Data Data
Server -> Client Data Message
Client->Server Start-StartExchange Response Time
Client -> Server Data Message
Metric Units Description
N Client-> Svr Exchange Msg Requests
Count of # Client<->Server request/response pairs measured for exchange messages from the Client to the Server
mSec Sum total of the Start-to-Start Delta Times in these Exchange Response Times
(x2) mSec2 Sum squared total of the Start-to-Start Delta Times in these Exchange Response Times
Xmax mSec The maximum Start-to-Start Delta Time of these Exchange Response Times
Xmin mSec The minimum Start-to-Start Delta Time of these Exchange Response Times
(IX) Sum total of the data point # times the Start-to-Start Delta Time
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
17
HTTP – Exchange Response
Client Server
HTTP Connect
HTTP ACK
*form*.jdbc-sybase
*order-entry*finish
*form*.jdbc-sybase
re-directedfinish-ack
entry-ackform graphic
Exchange Response
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
18
Application Response Example
CS#1S
SA
CS#2
CS#3
SC#1
SC#2
SC#3
LastCS
Data
CS E-SART
CS S-S ART
CS#6
CS#4
CS#5
SC#4
SC#5
SC#6
ChangeDir
CS->SC
CS S-E ART
StartCS
Data
SC S-S ART
SC E-SART
CS#7
LastSC
Data
ChangeDir
SC->CS
CS E-SART
CS S-S ART
SC S-E ART
LastCS
Data
StartSC
Data
StartCS
Data
ChangeDir
SC->CS
LastSC
Data
CS S-E ART
SC S-S RT
SC E-SRT
ChangeDir
CS->SC
StartSC
Data
Start & LastCS
Data
DefaultDir
(C->S)
Req. Network Delay+
Server DelayServer Delay
Req. Network Delay+
Server Delay+
Resp. Network Delay
Client Delay(to next request)
CS#2
CS#3
ReTxIgnored
ReTxIgnored
LEGENDS = SYNSA = SYN-ACKCS = Client->Server Data PacketSC = Server->Client Data Packet
CS S-S ART = Client->Server Start-to-Start Application Response TimeCS E-S ART = Client->Server End-to-Start Application Response TimeCS S-E ART = Client->Server Start-to-End Application Response TimeSC S-S ART = Server->Client Start-to-Start Application Response TimeSC E-S ART = Server->Client End-to-Start Application Response TimeSC S-E ART = Server->Client Start-to-End Application Response Time
Req. Network Delay = SUM(CS S-S ART) - SUM(CS E-S ART);Resp. Network Delay = SUM(CS S-E ART) - SUM(CS S-S ART);
Server Overhead = SUM(CS E-S ART);
Client Overhead = SUM(SC E-S ART);
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
19
Application ResponseStart to Start
Data Data Data Data
Data Data Data Data
Server -> Client Application Response
Client->Server Start-StartApplication Response Time
Client -> Server Application Request
Data Data
Server -> Client Retransmitted Datas
Data Data
Client->Server Retransmitted Datas
Metric Units Description
N Client->Svr Application Message Requests
Count of the # Client->Server request/response pairs measured for application message from the Client to the Server
mSec Sum total of the Start-to-Start Delta Times in these Application Response Times
(x2) mSec2 Sum squared total of the Start-to-Start Delta Times in these Application Response Times
Xmax mSec The maximum Start-to-Start Delta Time of these Application Response Times
Xmin mSec The minimum Start-to-Start Delta Time of these Application Response Times
(IX) Sum total of the data point # times the Start-to-Start Delta Time
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
20
HTTP – Application Response
Client Server
HTTP Connect
HTTP ACK
*form*.jdbc-sybase
*order-entry*finish
*form*.jdbc-sybase
re-directedfinish-ack
entry-ackform graphic
Application Response
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
21
More Metrics… Jitter Exchange
Response Application
Response
ConnectionConnection Connection Connection
SequenceSequence Connection Connection
WindowWindow RoutingRouting
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
22
TPM MIB – Changes in 01 Basic edits for minor corrections
Removal of some ‘statistics’ for complete ‘reports’
Usage of xxxClientID in place of xxxClientServerAddress
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
23
APM – Overall Framework
TPM StudyAPM Study
Micro-flow
Aggregation
Aggregation
Flow
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
26
TPM MIB - Framework APM MIB used to determine application
flows/aggregations to report micro-flows in TPM report table.
TPM Controls used to create performance metric reports on flows/aggregations.
All reports are in the form of statistics Drill-down for performance anomalies
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
27
TPM MIB Tables tpmReportControlTable - Study Control tpmMetricTable - Metric Selection
tpmReportTable - Detail Reports tpmCurrentTable - Current Aggregations tpmExceptionTable - Exception Reports
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
28
APM/TPM Report Table Linkage
apmReportControl – N
tpmReportControl – N
apmReportControlIndex - N
apmReportEntry – N, r
tpmReportEntry – N, r, m, f..n
apmReportIndex - r
tpmReportEntry – N, r, m+1, f..n
apmReportEntry – N, r+1
tpmReportEntry – N, r+1, m, f..n
apmReportIndex - r+1
tpmReportEntry – N, r+1, m+1, f..n
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
29
APM/TPM Current Table Links
apmReportControl – N
tpmReportControl – N
apmReportControlIndex - N
apmCurrentEntry – N, t
tpmCurrentEntry – N, t, m, f..n
apmTransactionID - t
tpmCurrentEntry – N, t, m+1, f..n
apmCurrentEntry – N, t..n
tpmCurrentEntry – N, t..n, m, f..n
apmTransationID – t..n
tpmCurrentEntry – N, t..n, m+1, f..n
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
30
APM/TPM Exception Table Links
apmExceptionTable – E
tpmExceptionControl – E
apmExceptionIndex - E
apmTransactionResponsivenessNotification –
E
tpmExceptionReportEntry – E, t, m, f..n
apmTransactionID - t
tpmExceptionReportEntry – E, t, m+1, f..n
apmTransactionUnsuccessfulNotification –
E
tpmExceptionReportEntry – E, t+1, m. f..n
apmTransactionID – t+1
tpmExceptionReportEntry – E, t+1, m+1, f..n
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
33
APM/TPM – Micro-flows tpmReportTable
tpmReportControlIndex protocolLocalDirIndex – Application (micro) protocolLocalDirIndex – Network tpmReportServerAddress tpmReportClientID tpmReportMetricIndex – Metric and
Protocol tpmCurrentApmTransactionID
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
34
APM/TPM Micro-flowsProtoc
olClient Server Protocol Metric
WEBWEB JimJim AmazonAmazon WEBWEB ResponsResponsee
WEB Jim Amazon HTTP RT-Delay
WEB Jim DNS-1 DNS RT-Delay
WEB Jim D-Click HTTP RT-Delay
WEB Jim FTP-A FTP RT-Delay
11 December, 2000
RMONMIB WG - 49th IETF San Diego, CA 2000
35
Updates in process Enhance ‘linkage’ with APM features
for additional ‘drill-down’ Integration of metrics from PMCAPS Integration of metric selection Document describing usage of
metrics for specific application groupings
11 December, 2000
36
Transport Performance Metric MIB
draft-ietf-rmonmib-tpm-mib-01.txtRussell Dietz
rdietz@hifn.comRMONMIB Working Group
49th IETF – San Diego, CA 2000
top related