Between Domain Traffic Engineering

Presentation Transcript

Between Domain Traffic Engineering Principles, Applications and Case Studies

Who We Are Josh Wepman Applications Engineer/Snake Oil Salesman Ixia NetOps jaw@ixiacom.com Joe Abley Toolmaker/Engineer/Token Canadian MFN PAIX jabley@mfnx.net

What We Are Talking About Inter-space Measurement, Analysis and Control Improving Connectivity With whom? Where? At what speed?

What we are NOT discussing MPLS DiffServ RSVP CR-LDP All sorts of different words with bunches of capital letters that have gotten to be connected with "movement designing… "

Goals For The Afternoon Methods and Concepts on the best way to "improve" between space availability Depending on your identity, "improve" will have distinctive implications Finding approaches to decrease effect of disappointment in associate or travel systems a.k.a. "increasing unwavering quality" WARNING: Some operational multifaceted nature may emerge! Put on your hazard delicate glasses...

Presentation Outline Inter-Domain TE Goals Definition Inter-area TE Measurement Applying Data to Address Your Goals Eliciting Control and the Feedback-Loop Conceptual Examples Who is Doing This Stuff? Real_Live_Network Examples No Questions? Great!

Iteration-1 – Conceptual Define Goals , Measure, Analyze, Refine Goals, Action What is it you have to finish? Between Domain TE Goals Definition

Examples of Goals Need to offload my "NSFnet" peering joins outbound (clog administration) Need to extend my between area peering joins cluefully (development) Need to discover a few people to give my administrations to (deals) That's correct, I said it… offer stuff!!!

Adjusting Your Assumptions Be set up to modify your presumptions in view of measured information! What you wanted to do, and what you wind up doing may change considerably. Try not to fear - this is genuine system information! Sign ought to increment as substantial system information gets to be accessible and counseled

Data Needs… What information sets are required? Stream trade information BGP directing information Active estimation information SNMP Some open apparatuses accessible (cflowd, zebra, ping, scotty, and so on) Some business items accessible…

Inter-area TE Measurement Also Known As: Getting great, issue/objective particular information!

Assumed Network Model Hierarchical Network Model Ingress/Egress Network administrations are isolated from Transit Services Works in other system models (as we will appear), however this is the thing that we are concentrating on...

Hierarchical Network Model Core Network Services Core1 Core2 Peer1 Peer2 LocalASN RemoteASN AS2 AS3 AS4 AS9

Types of Data to Measure Routing Data Focus here is BGP Traffic Data Flow-send out V5 is the center here Active Measurement Performance Data Ping/Traceroute/One-way delay/Jitter

Routing Data Routers for the most part do this well Core competency by plan (Routers route...) Different information sets are accessible for estimation IBGP (Good in the event that you are taking a gander at the entire framework, looking outbound or utilizing a level system demonstrate) Route-Reflection (Often required for inbound examination, can make some intricacy in level netowrk models) EBGP (Good to see your neighbor's perspective of you) Choose the right one to gauge in view of your needs/objectives

Routing Data – In/Outbound Core Network Services IBGP versus Course Reflection Core1 Core2 Collector Peer1 Peer2 Data LocalASN RemoteASN AS2 AS3 AS4 Routes AS9

Routing Data – In/Outbound When your objective is outbound portrayal, and your estimation point is the leave point for movement, IBGP is your person/young lady/other. Courses are constantly outside, and along these lines dependably spread (sans race and strategy obviously) "Conventions detest being humanized" When your objective is inbound portrayal, and your estimation point is the section point for activity, Route-Reflection must be utilized. Best way to get inner courses "neatly"

Route Data – Full Mesh (digression) Value of full work observing… Historical course following Policy benchmarking Tracking med-determination issue Identifying catastrophes the FIRST run through cluefully Don't simply sit tight for it to happen once more! If it's not too much trouble For everybody's purpose! Marginally off theme, yet beautiful darn essential!

Route Data – Full Mesh (pic) Core1 Core2 Core1 Core2 Collector Core2 Core1 Core2 Core1

Traffic Accounting Data Also Known As: Flow-fare NetFlow Cflow A MAJOR torment in the AS!

The Quick Skinny on Flow Packet and Byte counters per one of a kind arrangement of activity traits Measured from vital switches per input interface Which interfaces relies on upon your characterized objectives/needs... Made some amazing progress in the most recent couple of years In a few regards… 

Flow Data Inbound - Easy Core Network Services Core1 Core2 Collector Peer1 Peer2 Data LocalASN RemoteASN AS2 AS3 AS4 Routes AS9

Flow Data Outbound - Easy Core Network Services Core1 Core2 Collector Peer1 Peer2 Data LocalASN RemoteASN AS2 AS3 AS4 Routes AS9

Flow Data Outbound - Harder AS2 AS4 Core AS6 Core AS3

Flow Data Outbound - Harder Since stream send out information is inbound just, all potential feeder connects in a non-various leveled, blended administrations gadget must be represented keeping in mind the end goal to catch all activity outbound Issue: How would you know what information coming in center link4 is destined for the neighborhood outer connection? Course Reflection is awful here! Can twofold number! Issue exacerbated by complex strategy

18 Words or less on stream information Micro-administration of systems in light of streams == BAD Macro-administration of systems in view of streams == GOOD

Operational Challenges (1) Keep this as a top priority! Gilb's Law: "Anything can be measured in a way that is better than not measuring it by any means."

Operational Challenges (2) ACLs versus information send out in the colossal monster! Tested NetFlow on the GSR is generally conveyed to the LCs ACL > SNF > PIRC > IP Coloring > BGP Policy bookkeeping > FR Traffic policing which is not FR movement forming Apparently this progressions in 12.0(18)S

Operational Challenges (3) Some arrivals of JUNOS have bugs where just stream information from the most elevated numbered ifIndex gets sent out Check for PR20159

Operational Challenges (4) On rapid interfaces, as well as can be expected practically do is test at some proportion < 1:1 If you have to tally bytes, this will acquaint mistakes If you require with analyze tests, ensure the examples are standardized This does NOT mean duplicate by interim! Absence of ebb and flow look into on measurable legitimacy of stream information in view of tests Last research around 1993 Research originates before generous HTTP movement

Operational Challenges (5) The Gilb-Wepman Construct: "The aggregate P.I.T.A. consider experienced through the procedure of system estimation is far not exactly the aggregate P.I.T.A calculate experienced through arranging and designing a system without system estimations." P.I.T.A = Pain In The Ass those without clients might be new to this term

Performance Data Active estimation Round-excursion versus one-way mrtg and connect usage Important, however not a portion of our cases Short on time unfortunately… Helps in objective choice and re-determination Bottom line – is it better or more terrible?

Applying Data to your Goals What to do with this information? Activity Accounting Data connected to Routing information? Activity Load per <something> characteristic or course The attention here is on movement details (byte and parcel rates) per AS-PATH

AS-PATH/Traffic-information tables Traffic stack per AS-PATH makes a tree of activity connections (101) X-bits/sec (101,1234) Y-bits/sec (101,1234,9995) Z-bits/sec 101 - > 1234 - > 9995 X+Y+Z - > Y+Z - > Z Addresses the center mile AS's rather than conventional first or last ASN. Permits "TO" (source/sink) and "THROUGH" (travel) values rather than just "TO" values.

Data Aggregation - Time Aggregate information over time allotments (large scale level view) Long term averages Short term benchmarks obviously, fleeting signifies "~long term". Smaller scale administration of systems in light of streams BAD!

Data Aggregation - Interfaces Aggregate over the arrangement of interfaces that speak to your issue proclamation What interfaces am I keen on? Could be interface particular (one) Can be switch specific (many) Can be area wide (all) Can be N of M interfaces (a few) Pretty basic…

What to do with this? What does one do once they have this information?

Eliciting Control and The Feedback Loop Sit down, Josh Begone with your Snake Oil It's an ideal opportunity to beat on a few switches

Assumptions about your Routing Architecture Routes to outside systems are in BGP Your IGP lets you know how to discover the NEXT_HOP addresses in BGP We select leave focuses for activity in view of BGP way choice, not some other bizarre thing If you're steering strategy varies essentially from this, you have a bigger number of issues than estimation can tackle

Fixing Outbound Traffic Mark approach on BGP courses at the place where you learn them General arrangement - incline toward peering joins over costly travel joins, lean toward private peering joins over open peering joins Specific arrangement - briefly keep away from NAP X for movement to AS Y, favor AS C to achieve remote system D

Tweakable Knobs LOCAL_PREF MED AS_PATH Check your merchant's BGP way determination sudden death round rundown, and picked an arrangement of handles that gives you the sort of control your arrangement directs

Control of Outbound Traffic Danger, Will Robinson! Helpdesk telephone may ring Small change, delay, check, log, stop, inhale, rehash Exit determination is a sensibly exact science

Fixing Inbound Traffic Controlling inbound movement stream is about attempting to impact the BGP p