Export of On-Path Delay in IPFIX

1. Export of On-Path Delay in IPFIX
draft-opsawg-ipfix-on-path-telemetry-02
Enabling a statistical network delay view, giving insights
where delay is being accumulated in the forwarding path
1
thomas.graf@swisscom.com
benoit.claise@huawei.com
alex.huang-feng@insa-lyon.fr
27. March 2023

2. On-Path Delay @ IPFIX
Draft Status
• Extended the introduction and the terminology section with
performance registry relevant information's.
• Corrected some small nits in the performance registry sections.
• Increased IPFIX entity data type sizes based on implementation tests
results.
• Corrected IPFIX entity data type semantic.
• Describing how IPFIX reduced-size encoding is applicable in new
operational consideration section.
• According input from Greg Mirsky detailing IOAM Application section.
• Removed nanosecond granularity.
• Added Implementation Status section.
2
Node based
Flow Aggregation
Apache Kafka
Message Broker
Timeseries DB
Pmacct
Data Collection
IOAM
nodes
Data-collection based
Flow Aggregation
Message Broker based
Consolidation
Data Base
Join

3. On-Path Delay @ IPFIX
Running Code at IETF 116 hackathon
• INSA Lyon showed
running open-source
code based on IOAM-
Trace-Type in FD.io VPP
(https://github.com/net
work-analytics/vpp-srh-
onpath-telemetry) and
Huawei based on IFIT
SRH TLV their
implementation in VRP.
• pmacct data collection
calculates
PathDelayMeanDeltaMi
croseconds by dividing
PathDelaySumDeltaMic
roseconds by
packetDeltaCount.
3
192.0.2.0/24 2a02:a90:4007:700 (VLAN 500)
.154
.157 .159
ipf-zbl1325-srv6lab-al-57
Lo0: 198.51.100.157
2001:DB8:6::1/128
SRv6 Locator: 2001:DB8:10:: 64/48
ipf-zbl1325-srv6lab-ms-58
Lo0: 2001:DB8:7::1/128
SRv6 Locator: 2001:DB8:17:: 64/48
ipf-zbl1325-srv6lab-ipcore-91
Lo0: 198.51.100.191
2001:db8:2::1/128
SRv6 Locator: 2001:db8:2::/48
ipf-zbl1325-srv6lab-ml-59
Lo0: 2001:DB8:8::1/128
SRv6 Locator: 2001:DB8:18:: 64/48
.155
.158
ietf-internal (Collector)
MGMT: 10.212.252.10
ens42.523
203.0.113.46/28
2001:db8:f:3::ffff/64
ens42.524
203.0.113.30/28
2001:db8:f:2::ffff/64
VRF-A10
RD: 65536:190
SRv6 SID: 2001:db8:1:0:40::
TenGigE0/0/0/0.524: 203.0.113.18/28
2001:db8:f:2::18/64
Lo10: 203.0.113.190/32
2001:db8:f::190/128
VRF-A10
RD: 65536:191
SRv6 SID: 2001:db8:2:0:40::
TenGigE0/0/0/0.524: 203.0.113.19/28
2001:db8:f:2::19/64
Lo10: 203.0.113.191/32
2001:db8:f::191/128
.156 .190
.191
203.0.113.16/28 2001:db8:f:2::/64 (VLAN 524)
VRF-A10
RD: 65536:157
SRv6 SID: 2001:db8:10:0:124::
GigabitEthernet0/7/0.523: 203.0.113.35/28
2001:2:1::35/64
Lo10: 203.0.113.157/32
2001:DB8:F::157/128
203.0.113.32/28 2001:db8:f:3::/64 (VLAN 523)
VRF-A10
RD: 65536:154
SRv6 SID: 2001:db8:9:0:124::
GigabitEthernet0/7/0.523: 203.0.113.34/28
2001:2:1::34/64
Lo10: 203.0.113.154/32
2001:DB8:F::154/128
MP-iBGP
ens224
192.0.2.1/24
2a02:a90:4007:700::1/64
ipf-zbl1327-srv6lab-al54
Lo0: 198.51.100.154
2001:DB8:3::1/128
SRv6 Locator: 2001:DB8:9:: 64/48
ipf-zbl1327-srv6lab-ms-55
Lo0: 2001:DB8:4::1/128
SRv6 Locator: 2001:DB8:14:: 64/48
ipf-zbl1327-srv6lab-ml-56
Lo0: 2001:DB8:5::1/128
SRv6 Locator: 2001:DB8:15:: 64/48
ipf-zbl1327-srv6lab-ipcore-90
Lo0: 198.51.100.190
2001:db8:1::1/128
SRv6 Locator: 2001:db8:1::/48
IPFIX 2a02:a90:4007::2:1/128
BMP 2a02:a90:4007::4:2/128
100GE1/0/4
400GE1/0/0
400GE2/0/0
100GE2/0/4
100GE0/5/0 100GE2/0/4
100GE1/0/4
400GE1/0/0
400GE2/0/0
100GE2/0/4
100GE1/0/4
100GE1/0/4
400GE2/0/0
400GE1/0/0
100GE2/0/4
400GE1/0/0
400GE2/0/0
 Delay τ

4. On-Path Delay @ IPFIX
Running Code at IETF 116 hackathon
4
(1) Shows SRv6 SID list
change of the traffic
engineered paths.
(2) Shows on which node
how much on-path
delay was being
measured.
(3) Shows the BGP
update/withdrawals
from the topology
change.
(4) Shows that Network
Anomaly Detection
detects the topology
and delay change and
the Max Concern
Score calculation.
1
2
3
4

5. On-Path Delay @ IPFIX
Next steps
• Do you recognize the problem statement?
• Network operators want to understand
• where delay with which network and device dimensions is being
accumulated
• at highest scale for a statistical network delay view.
• IEs in document defined are independent from how the delay is being
metered.
• Two vendors are validating technical feasibility. Others showing interest.
• Draft version -03 will contain data record and template examples.
5
thomas.graf@swisscom.com
benoit.claise@huawei.com
alex.huang-feng@insa-lyon.fr
27. March 2023
Node based
Flow Aggregation
Apache Kafka
Message Broker
Timeseries DB
Pmacct
Data Collection
IOAM
nodes
Data-collection based
Flow Aggregation
Message Broker based
Consolidation
Data Base
Join