DD2491 p1 2008. Load balancing BGP. Johan Nicklasson KTHNOC/NADA

Similar documents
Load balancing and traffic control in BGP

Transitioning to BGP. ISP Workshops. Last updated 24 April 2013

Advanced BGP Policy. Advanced Topics

Load balancing and traffic control in BGP

Bell Aliant. Business Internet Border Gateway Protocol Policy and Features Guidelines

Border Gateway Protocol (BGP)

Exterior Gateway Protocols (BGP)

HP Networking BGP and MPLS technology training

Simple Multihoming. ISP/IXP Workshops

Multihoming and Multi-path Routing. CS 7260 Nick Feamster January

IPv6 over IPv4/MPLS Networks: The 6PE approach

E : Internet Routing

Inter-domain Routing Basics. Border Gateway Protocol. Inter-domain Routing Basics. Inter-domain Routing Basics. Exterior routing protocols created to:

APNIC elearning: BGP Attributes

BGP Multihoming. Why Multihome? Why Multihome? Why Multihome? Why Multihome? Why Multihome? Redundancy. Reliability

Introducing Basic MPLS Concepts

Outline. EE 122: Interdomain Routing Protocol (BGP) BGP Routing. Internet is more complicated... Ion Stoica TAs: Junda Liu, DK Moon, David Zats

Simple Multihoming. ISP Workshops. Last updated 30 th March 2015

BGP and Traffic Engineering with Akamai. Christian Kaufmann Akamai Technologies MENOG 14

CS551 External v.s. Internal BGP

BGP1 Multihoming and Traffic Engineering

Lecture 18: Border Gateway Protocol"

Border Gateway Protocol BGP4 (2)

Understanding Route Redistribution & Filtering

Internet inter-as routing: BGP

ISP Case Study. UUNET UK (1997) ISP/IXP Workshops. ISP/IXP Workshops. 1999, Cisco Systems, Inc.

Introduction to Routing

The Case for Source Address Routing in Multihoming Sites

Multihomed BGP Configurations

BGP Best Path Selection Algorithm

BGP Routing. Course Description. Students Will Learn. Target Audience. Hands-On

How To Make A Network Secure

IPv6 Addressing. ISP Training Workshops

basic BGP in Huawei CLI

Examination. IP routning på Internet och andra sammansatta nät, DD2491 IP routing in the Internet and other complex networks, DD2491

Juniper Exam JN0-343 Juniper Networks Certified Internet Specialist (JNCIS-ENT) Version: 10.1 [ Total Questions: 498 ]

BGP Multihoming Techniques

How To Understand Bg

BGP Basics. BGP Uses TCP 179 ibgp - BGP Peers in the same AS ebgp - BGP Peers in different AS's Private BGP ASN. BGP Router Processes

Using the Border Gateway Protocol for Interdomain Routing

Supporting Document PPP

Notice the router names, as these are often used in MPLS terminology. The Customer Edge router a router that directly connects to a customer network.

DESTINATION BASED RTBH FILTERING AT ATTACK ORIGINATING INTERNET SERVICE PROVIDER

For internal circulation of BSNLonly

Module 12 Multihoming to the Same ISP

Active measurements: networks. Prof. Anja Feldmann, Ph.D. Dr. Nikolaos Chatzis Georgios Smaragdakis, Ph.D.

BGP Multihoming Techniques

LHCONE Site Connections

Routing Protocol - BGP

Routing in Small Networks. Internet Routing Overview. Agenda. Routing in Large Networks

Address Scheme Planning for an ISP backbone Network

Introduction to The Internet

Introduction to The Internet. ISP/IXP Workshops

Hunting down a DDOS attack

Week 4 / Paper 1. Open issues in Interdomain Routing: a survey

Analyzing Capabilities of Commercial and Open-Source Routers to Implement Atomic BGP

Beginning BGP. Peter J. Welcher. Introduction. When Do We Need BGP?

Border Gateway Protocol Best Practices

Chapter 49 Border Gateway Protocol version 4 (BGP-4)

Understanding Virtual Router and Virtual Systems

Based on Computer Networking, 4 th Edition by Kurose and Ross

BGP and Traffic Engineering with Akamai. Caglar Dabanoglu Akamai Technologies AfPIF 2015, Maputo, August 25th

Enterprise Network Simulation Using MPLS- BGP

Understanding Route Aggregation in BGP

Introduction to MPLS-based VPNs

National Education Network. KAREN School Cluster High-level Design

Multihoming: An Overview

Network Level Multihoming and BGP Challenges

Doing Don ts: Modifying BGP Attributes within an Autonomous System

BGP Attributes and Path Selection

CS551 Multi-homing in BGP

Internet Routing Protocols Lecture 04 BGP Continued

How To Make A Network Plan Based On Bg, Qos, And Autonomous System (As)

BGP Terminology, Concepts, and Operation. Chapter , Cisco Systems, Inc. All rights reserved. Cisco Public

WAN Topologies MPLS. 2006, Cisco Systems, Inc. All rights reserved. Presentation_ID.scr Cisco Systems, Inc. All rights reserved.

ITRI CCL. IP Routing Primer. Paul C. Huang, Ph.D. ITRI / CCL / N300. CCL/N300; Paul Huang 1999/6/2 1

Can Forwarding Loops Appear when Activating ibgp Multipath Load Sharing?

IPv6 Address Planning

BGP Convergence in much less than a second Clarence Filsfils - cf@cisco.com

BGP Multihoming Techniques

B. Quoitin, S. Uhlig, C. Pelsser, L. Swinnen and O. Bonaventure

Fireware How To Dynamic Routing

Claudio Jeker. RIPE 41 Meeting Amsterdam, 15. January Using BGP topology information for DNS RR sorting

Network Architecture Validated designs utilizing MikroTik in the Data Center

APNIC elearning: BGP Basics. Contact: erou03_v1.0

Fast Re-Route in IP/MPLS networks using Ericsson s IP Operating System

Module 7. Routing and Congestion Control. Version 2 CSE IIT, Kharagpur

SRX High Availability Design Guide

BGP Multihoming Techniques

MPLS Implementation MPLS VPN

JUNOS Secure BGP Template

Masterkurs Rechnernetze IN2097

BGP route propagation. Internet AS relationships, Routing policy on Internet paths. Example of commercial relationship. Transit vs.

IPv6 Address Planning. MENOG 11 Amman 30 th September 9 th October 2012 Philip Smith

Quidway MPLS VPN Solution for Financial Networks

Transcription:

DD2491 p1 2008 Load balancing BGP Johan Nicklasson KTHNOC/NADA

Dual home When do you need to be dual homed? How should you be dual homed? Same provider. Different providers. What do you need to have in place to do dual homing? AS number PI vs. PA BGP?

Single provider You can do dual homing to the same provider. What kind of redundancy do we have in this setup? BGP does not load balance across multiple links.

Single provider What level of redundancy do we need? Do we need redundant routers? Do we need to connect to different POPs? What about the local loop? Can we load balance over redundant links? Maybe. In the previous example we could make use of an IGP to load balance packets over the two links. It is not likely that the provider wants to do that. A provider wants his edge to be as static as possible. What about different routers?

Single provider Redundant routers. We can loose 1 router and still be connected to the internet. If the ISP router dies our connections goes with it. The local loop may or may not be redundant.

Single provider Redundant routers and redundant POPs We can loose one of our routers and still have connectivity. We can loose one provider router/pop and still have connectivity. The local loop may or may not be redundant.

Single provider Dual homed to the same provider. We can have different levels of redundancy. Depending on our needs. Do we need BGP to dual home to the same provider? Even in the setup with 2 routers and 2 POPs we can use a static default route to get to the internet. We inject the default route into our IGP and the node will send it's traffic to the nearest exit point. Can the ISP load balance traffic to us?

Single provider If we use BGP on our connections to our ISP We must have an ibgp connection between our edge routers. We can still use a static default route to the internet. The provider can send us a default route via BGP. We can use MED, AS prepend or communities to try to get the ISP to send us traffic to different prefixes over different links.

Single provider If we get a full table from our ISP We need our hardware to handle ~230 000 prefixes. We can use policies to have the traffic leave on different links depending on the destination. If we don't make our IGP aware of those routes we could end up with suboptimal routing, depending on the network topology.

Single provider Addressing and AS numbers The provider will assign IP address space to us. We don't have to have our own AS number. We can use a private AS, that have to be assigned to us by our provider. Private AS numbers are 64512 to 65535. The provider have to remove private AS's from prefixes on their ebgp peerings.

Dual providers Our address space becomes an issue. We need PI (Provider independent) space We need a public AS number. How to get an AS number will be covered later. BGP is a must.

Dual providers

Dual providers If we should use IP space provided by ISP A, 10.1.1.0/24 That /24 is just a portion of the space provided to the ISP by the RIR. 10.1.0.0 /19 We get ISP B to announce our /24 (most ISPs will never announce part of another ISP aggregate).

Dual providers

Dual providers Which ISP will attract all our traffic? Longest prefix match One solution to this problem would be to have ISP A announce 10.1.1.0/24 and 10.1.0.0/19.

Dual providers

Dual providers Another solution to this problem is to get Provider Independent (PI) IP space from a RIR (Regional Internet Registry). To use PI space will also make it much easier to switch ISP. With PA space the ISP owns the IP addresses you use. If you move to another ISP the first one will make you return the borrowed space.

Dual providers Load balancing the egress Using BGP attributes and IGP cost you can prefer one prefix set over one ISP and another set over the other ISP. This will not balance the load equally over the two upstreams. If you monitor your traffic patterns you could try to balance the load more.

Dual providers Load balancing the ingress Is it possible to use MED when dual homing with two IPSs? Is it possible to use AS prepend? How about announcing more specific routes to attract traffic? You have to have a good dialog with your ISPs when you are doing any kind of traffic engineering.

Symmetry/asymmetry and the internet When you have more the one way to reach a destination symmetry can not be guaranteed. Some hardware dealing with state and flow needs symmetry to work properly.

Symmetry/asymmetry

Symmetry/asymmetry We have to have traffic leaving a firewall return over the same one. We could use AS prepend. We could advertise more specific routes What if the firewalls exchanged their current flow and state tables?

Questions?