Single-homing — relying on one IP transit provider — is the cheapest way to connect your network to the internet. It's also the riskiest. One fiber cut, one BGP misconfiguration, one provider outage, and your business is offline.
Multi-homing solves that. By connecting your network to two or more transit providers, you get redundancy, better global routing, and real leverage on pricing. This guide explains how multi-homing works, what it costs, and how to deploy it without shooting yourself in the foot.
What Is Multi-Homing?
Multi-homing is the practice of connecting your autonomous system (AS) to two or more upstream networks — usually IP transit providers, sometimes a mix of transit and peering — and announcing your IP prefixes through all of them via BGP.
If one upstream fails, traffic automatically reroutes through the others. If one upstream has a bad route to a specific destination, BGP can choose a better path through another.
Why Multi-Home?
1. Redundancy and Uptime
Even the best transit providers experience outages — fiber cuts, router failures, BGP leaks, DDoS attacks on their infrastructure. Multi-homing isolates you from any single provider's failure.
2. Route Diversity
Different providers reach different destinations through different paths. A multi-homed network can pick the lowest-latency or most reliable path for each destination.
3. Pricing Leverage
When you have two transit contracts, neither provider can hold you hostage at renewal. You can shift traffic toward the cheaper provider and use the threat of doing so to negotiate.
4. Geographic and Backbone Diversity
Some providers are stronger in certain regions. Multi-homing with a North American Tier 1 and a European Tier 1, for example, gives you better latency on both sides of the Atlantic.
5. DDoS Resilience
If one provider is hit by a volumetric attack targeting your prefixes, you can withdraw the announcement from that provider and shift traffic to your other upstreams.
What You Need Before Multi-Homing
- Your own ASN (Autonomous System Number) from your regional internet registry (RIPE, ARIN, APNIC, LACNIC, AFRINIC).
- Provider-independent (PI) IP space — typically a /24 IPv4 minimum and a /48 IPv6 prefix. Provider-assigned (PA) space won't work, because you can't announce it from another provider.
- A BGP-capable router (Cisco, Juniper, Arista, MikroTik, FRR, BIRD, VyOS, etc.) with enough memory for full tables (~2 GB RAM per full table is a safe minimum).
- Two or more transit providers, ideally on physically diverse fiber paths and using different upstream carriers.
- RPKI ROAs for your prefixes, so other networks can validate your announcements.
Active/Active vs Active/Passive
Active/Active
You announce your prefixes to all transit providers simultaneously, and inbound and outbound traffic is load-balanced based on BGP best-path selection. This maximizes throughput and resilience but requires careful traffic engineering.
Active/Passive
You announce your prefixes to all providers, but use BGP techniques (AS-path prepending, MED, communities) to make one provider the primary and the other(s) backup. Simpler to manage but doesn't fully utilize capacity.
BGP Traffic Engineering Basics
Multi-homing is only as good as your BGP configuration. Key knobs:
Inbound Traffic Control
- AS-path prepending — repeat your ASN multiple times in the announcement to make a path look longer (less preferred).
- BGP communities — tag routes with provider-specific communities to influence how the upstream propagates them.
- More-specific announcements — announce a /25 or /26 to one provider while the /24 goes everywhere; longest-prefix match wins.
- Selective announcement — announce certain prefixes only to certain providers.
Outbound Traffic Control
- Local preference — set a higher local-pref on routes from your preferred provider.
- Weight (Cisco-specific) — same idea, but higher in the BGP decision process.
- MED (Multi-Exit Discriminator) — used between two routers connecting to the same provider.
- Route filtering — drop or prefer routes based on prefix list, AS-path, or community.
Common Multi-Homing Mistakes
- Becoming a transit network by accident. If you accept full tables from Provider A and announce them to Provider B, you're now transiting their traffic. Filter announcements outbound — only announce your own prefixes.
- No RPKI ROAs. Without ROAs, your prefixes are vulnerable to BGP hijacks and many networks may drop your routes.
- Announcing too-specific prefixes. Announcing a /25 globally is wasteful — many networks filter prefixes longer than /24 (IPv4) or /48 (IPv6).
- Single point of failure on your side. Two transit ports into the same router is still a single failure domain. Use two routers and diverse cross-connects.
- Forgetting IPv6. Plan IPv6 multi-homing alongside IPv4 — it's free with most providers.
- No monitoring. Multi-homing fails silently if you don't monitor BGP session state, prefix counts, and traffic distribution.
How to Choose Multiple Providers
Don't just buy two transit ports from the cheapest providers. Optimize for diversity:
- Different upstream backbones. If both your providers use the same Tier 1 upstream, that Tier 1's outage takes you both down.
- Different physical paths. Two fibers in the same conduit fail together. Ask for KMZ files or path documentation.
- Different routing policies. One provider with strong North American peering, one with strong European or Asian reach.
- Different sizes. One Tier 1 for global reach + one Tier 2 with strong regional peering is often the best mix.
Cost of Multi-Homing
Multi-homing roughly doubles your transit cost — but not your total cost. The redundancy benefit usually justifies it for any business where downtime costs more than a few hundred dollars per hour.
Practical cost-saving tactics:
- Mismatch your commits. Run a higher 95th percentile commit on your primary provider and a smaller commit on your backup, then engineer traffic accordingly.
- Use peering to offload. Free settlement-free peering at IXPs reduces what you push through paid transit.
- Negotiate aggressively. Tell each provider you have another quote — you almost always do.
Multi-Homing at Noded
Noded is built specifically for customers who want to multi-home easily. We support BYO IP space (bring your own /24 and /48), provide RPKI signing, and offer competitive 95th percentile pricing on a network with diverse Tier 1 upstreams and direct IXP peering.
Whether you're adding Noded as a second transit to an existing setup or building a multi-homed network from scratch, our network team can help you design announcements, communities, and traffic engineering. Get in touch to discuss your topology.
Frequently Asked Questions
Do I need my own ASN to multi-home?
Yes. Without your own ASN and PI IP space, you can't announce the same prefixes through multiple providers. Apply through your regional internet registry (RIPE, ARIN, etc.).
Can I multi-home with just one router?
Technically yes, but the router itself becomes your single point of failure. Best practice is two routers in different chassis, ideally in different racks or rooms.
How many transit providers should I have?
Two is the practical minimum for redundancy. Three is common for large networks (one Tier 1 + two Tier 2/regional). More than three is rarely worth the complexity.
What is RPKI and do I need it?
RPKI (Resource Public Key Infrastructure) is a system for cryptographically signing your BGP announcements. Yes, you need it — most major networks now drop or de-prefer RPKI-invalid routes.
Can I peer at an IXP instead of multi-homing with two transits?
Peering is a great complement to multi-homing, but peering alone doesn't reach the full internet. Use peering for traffic to peers, transit for everything else, and multi-home your transit for redundancy.