an evolution of network telemetry - open compute project · 3/23/2018 6 carrier network oam...
TRANSCRIPT
An Evo lu t ion o f Network Te lemet ry
Tal Mizrahi
Marvell
Telemetry BoF
3/23/2018 3
Network Telemetry
Are we reinventing the wheel?
Self-driving Networks High reliability
Easy-to-manage
Software-defined Networks
High speed
Network Telemetry
Easy-to-troubleshoot
3/23/2018 4
Ping / Traceroute Reinventing the wheel?
3/23/2018 5
Old-School Passive Monitoring
Counters Per port
Queue State
Latency
Per flow
Per queue
… …
Reinventing the wheel?
3/23/2018 6
Carrier Network OAM Reinventing the wheel?
OAM Protocols
IETF ICMPv4
IEEE 802.1agITU-T Y.1731
IETF ICMPv6 IETF IPPM IP OAM
Higher
Layers
Layer 3
Layer 2
Layer 1
ITU-T Y.1711
MPLS OAM
IEEE 802.3ah
MPLS / PWE3
OAM
Ethernet
OAM
ITU-T G.8113.1MPLS-TP OAM
IETF MPLS-TP
OAM
IETF LSP-Ping
MPLS OAM
IETF PWE3
VCCVIETF
BFD
Active measurement / monitoring:
Control Message
Control Message
3/23/2018 7
Piggybacked Measurement Reinventing the wheel?
Telemetry is piggybacked onto data packets
IOAM / INT AM-PM
3/23/2018 8
Piggybacked Metadata – IOAM / INT Reinventing the wheel?
IOAM / INT Domain
Analytics Server
Switches push local
metadata into header:
delay, queue state, …
Telemetry
Info
IOAM (IETF)
INT (P4)
+SAI TAM
PacketsSent: 10,000
Packets Lost: 500
Packets Received: 9,500
3/23/2018 9Analytics Server
Data Center Network
Servers Servers
Reinventing the wheel?AM-PM: Alternate Marking –Performance Measurement
(RFC 8321)
Counts number of
packets received
Counts number
of packets sent
3/23/2018 10
Reinventing the wheel?Network Telemetry: An Evolution
Ping
TraceroutePassive
Monitoring
Carrier
OAM
IOAM / INT
AM-PM
NO !
Active Telemetry
3/23/2018 11
Marvell’s Network Telemetry Solutions
Passive Telemetry
CountersFlow-based
Probes IEEE 1588
Metadata
Grab-’n-GoTimestampingSelective
Probing
Per-hop Telemetry
Programmable
Logic
Packet
Metadata
Incoming
PacketModified
Packet
Incoming
Packet
Packet Processing
Packet
Metadata
Fixed
Programmable
3/23/2018 12
Metadata: Grab-’n-Go
Marvell Prestera
Packet
Processing
Packet
Metadata
Incoming
PacketOutgoing
Packet
3/23/2018 13
Timestamping Everything
✔ Periodic probing
✔ AM-PM
✔ TimeFlip
IEEE 1588 Synchronized Clock
Marvell Prestera
3/23/2018 14
Time.Sec Time.Frac1 * … ** … *
Periodic range
... timestampvalue
1 secondaction
timestamp field2 field3 field4 …
TCAM
Marvell Prestera Switch
header /
metadata
AM-PM using TimeFlip
Per-Flow Congestion Detection using AM-PM
loss and delay è congestion is detected
3/23/2018 15
3/23/2018 16
Selective Exporting
✔ Drop
✔ Congestion
✔ 1 of N
✔ Periodic
Analytics Server
Telemetry
Info
Marvell Prestera
IOAM / INT Domain
Analytics Server
3/23/2018 17
IOAM/INT: Periodic Exporting
Export one packet per second
Export last ms of every second
Telemetry
Info
Per flow / per port
IOAM / INT Domain
Analytics Server
3/23/2018 18
IOAM/INT: Adaptive Exporting
Event-driven:
✔ On drop
✔ On congestion
✔ On high rate
IOAM / INT Domain
Analytics Server
3/23/2018 19
Combining IOAM/INT with AM-PM
AM-PM as a trigger for exporting telemetry info.
Per-hop Telemetry
Challenge: export telemetry info without the expensive overhead of IOAM/INT.
3/23/2018 20
The Network Telemetry Toolset
CountersFlow-based
Probes IEEE 1588
Metadata
Grab-’n-GoTimestamping
Selective
Probing
Marvell Prestera
3/23/2018 22
References
[1] Mizrahi, T., Vovnoboy, V., Nisim, M., G. Navon, and A. Soffer, “Network Telemetry Solutions for Data Center and
Enterprise Networks”, Marvell white paper, 2018.
[2] Brockners, F., Bhandari, S., Pignataro, C., Gredler, H., Leddy, J., Youell, S., Mizrahi, T., Mozes, D., Lapukhov, P.,
Chang, R. and D. Bernier "Data Fields for In-situ OAM", draft-ietf-ippm-ioam-data-00, work in progress, 2017.
[3] C. Kim et al., “In-band network telemetry (INT)”, P4 consortium, 2015.
[4] Fioccola, G., Capello, A., Cociglio, M., Castaldelli, L., Chen, M., Zheng, L., Mirsky, G., and T. Mizrahi, “Alternate
Marking method for passive and hybrid performance monitoring”, RFC 8321, 2018.
[5] Mizrahi, T., Rottenstreich, O. and Y. Moses, “TimeFlip: Scheduling Network Updates with Timestamp-based TCAM
Ranges”, IEEE INFOCOM, 2015.