Two Carriers or One Blended Pipe? Multihoming a Boise Rack in 2026

November 19, 2026 · 10 MIN READ

For most Boise racks pushing under 10 Gbps, one blended transit pipe from an operator running its own ASN beats juggling two carriers: one commit, one BGP session, upstream diversity handled for you. Run two carriers yourself only if you hold your own ASN and a /24, and have engineers who'll own routing at 3 a.m.

What's the Difference Between Blended Transit and Multihoming?

Multihoming, in the strict sense, means your network is reachable through two or more independent providers. You hold an ASN, you announce your own prefix to each upstream, and the global routing table sees multiple paths to you. When one carrier drops, BGP withdraws that path and the survivor carries everything.

Blended transit moves that architecture one hop upstream. Your provider does the multihoming, and you buy a single port. When you buy transit from IDACORE in Boise, your traffic leaves on AS36824: our ASN, our routers, our BGP sessions, across multiple upstreams. We choose exit paths, filter bogons, tune communities, and watch for the session that flaps at 2 a.m. You get a default route, or a full table if you want one, and a single commit to manage.

We didn't learn this from a vendor whitepaper. Our team founded and ran an ISP, held our own ASN, and peered at the Seattle Internet Exchange. Traffic engineering is something we do, not something we resell.

The honest tradeoff is that with blended transit, AS36824 sits in your path. You've diversified carriers, but you haven't diversified providers. Whether that matters depends on which failure you're protecting against.

Which failures does each model cover?

Failure scenario Blended transit (one port) Two carriers, your own ASN Hybrid (IDACORE + one direct carrier)
One upstream carrier's backbone degrades Covered: we route around it Covered: BGP withdraws the path Covered
Fiber cut on one upstream path Covered Covered Covered
Routing policy error at your transit provider Not covered Covered Covered
Your own edge router dies Not covered (needs dual routers) Not covered (needs dual routers) Not covered (needs dual routers)
Self-inflicted route leak or bad prepend Mostly avoided: we filter your announcements Your problem Partially your problem
ASN and IP space required No Yes Yes

That last row is where most teams make their decision. Notice the fourth row, too. Carrier diversity and hardware redundancy are separate problems. Two transit providers terminating on one router is still a single point of failure.

What Does Running Two Carriers Yourself Cost in 2026?

The port fees are the visible part. The rest of the bill hides in four places.

Address space and an ASN. You need an ARIN-assigned ASN and a provider-independent prefix. Anything longer than a /24 in IPv4 gets filtered by most of the internet, so a /24 is the floor. If you don't already have one, expect to spend thousands of dollars on the IPv4 transfer market. IPv6 is easy by comparison, since a /48 is cheap and plentiful. Plan to publish RPKI ROAs and IRR objects for both, or a growing share of networks will drop your routes.

Commits sized for failover. This is the cost people forget. Each leg has to carry 100% of your traffic when the other one fails. If your 95th percentile sits at 600 Mbps, you need a 1G commit on both sides. You end up paying for 2 Gbps of commitment so you can use 600 Mbps, with two contracts, two renewal cycles, and two account teams.

Routers that can hold full tables. The IPv4 global table is past a million prefixes in 2026, and IPv6 is climbing fast. Two full-table sessions need real RIB memory and FIB capacity with headroom. A cheap L3 switch won't hold up here.

People. Someone on your team has to understand why prepending doesn't always shift inbound traffic, set sane max-prefix limits, and catch a ROA with the wrong max-length before it blackholes half your customers. In our experience, the most common multihoming outage isn't a carrier failure. It's a self-inflicted mistake on the customer's edge.

How Does 95th Percentile Billing Change the Math?

Transit at IDACORE is billed on committed bandwidth at the 95th percentile. We sample your port every five minutes, sort the month's samples, throw out the top 5%, and bill on the highest remaining value. A 30-day month produces 8,640 samples, and the top 432 get discarded. That's about 36 hours of peak traffic you never pay for.

Here's why that matters for multihoming. When you run two carriers, each link is billed independently. A six-hour failover dumps all your traffic onto the surviving link, but those six hours fit comfortably inside its 36-hour burst window. Your bill barely moves. With a blended pipe, a carrier failure upstream never touches your port's utilization at all, because we absorb the reroute.

Committed bandwidth Monthly price Effective rate
100 Mbps $170/mo $1.70/Mbps
1 Gbps $600/mo $0.60/Mbps
10 Gbps $2,000/mo $0.20/Mbps
100 Gbps $11,050/mo $0.111/Mbps

These are IDACORE's published rates, on-net in Boise and at IDACORE North in Coeur d'Alene. There's no per-GB meter running anywhere on that table.

What Does This Look Like for a Real Boise Deployment?

Take a Treasure Valley SaaS company running its production stack in IDACORE Boise. Their gear draws 4 kW, so colocation is $1,200/month at $300/kW, billed on power draw regardless of how many U they occupy. Their 95th percentile lands around 600 Mbps, averaging about 350 Mbps across the month, with month-end reporting spikes near 900 Mbps.

Blended option: a 1G commit from IDACORE at $600/month. Colocation plus transit comes to $1,800/month. No ASN, no /24, no full-table router. They take a default route over one BGP session and they're done.

Two-carrier option: two separate 1G commits from direct carriers, since each has to absorb the 900 Mbps peak alone. Add the ASN, a /24 from the transfer market, routers sized for two full tables, and the engineering hours to run it all. For a team without dedicated network staff, the people cost alone outweighs the second port.

For comparison, the hyperscaler version: 350 Mbps average works out to roughly 113 TB of egress a month. At typical hyperscaler internet egress list rates of $0.05 to $0.09 per GB, that's about $5,700 to $10,200 a month, and the bill grows every time the product succeeds. The same traffic on IDACORE transit is $600. Sound familiar?

This works at small scale, too. IDACORE Boise rents a single rack unit, and you pay for the power your server draws, not for the space. A 200W 1U server costs $60/month at $300/kW, and it gets the same blended transit as a full cabinet.

When Should You Run Two Carriers or a Hybrid?

Blended transit is the right default. It stops being the right answer in a few specific cases:

  • You already own an ASN and PI space and want the same prefix announced from Boise and from a DR site such as IDACORE North, about 45 minutes from Spokane.
  • A contract or regulator requires provider diversity, not just carrier diversity. Healthcare and financial customers run into this.
  • You're above 10G and a policy failure at a single provider is a material business risk.
  • You need per-prefix traffic engineering toward specific networks for latency reasons. From Boise, we see 14ms to Salt Lake City, 22ms to Portland, and 23ms to Seattle, and some workloads care which path gets there.

For most of these, the smart build is a hybrid: IDACORE transit as your primary, already blended across multiple upstreams, plus one direct contract with another of the seven on-net carriers in the building (Zayo, Lumen/Level 3, Cogent, CenturyLink, Syringa, Cable One, Hurricane Electric). You get provider diversity, and your primary leg is already diverse underneath.

A minimal FRR config for that hybrid looks like this. It uses documentation ASNs and prefixes, with 64496 as your ASN and 64500 as the second carrier:

ip route 198.51.100.0/24 Null0
!
router bgp 64496
 bgp router-id 198.51.100.1
 neighbor 203.0.113.1 remote-as 36824
 neighbor 203.0.113.1 description IDACORE-transit
 neighbor 192.0.2.1 remote-as 64500
 neighbor 192.0.2.1 description carrier-b
 !
 address-family ipv4 unicast
  network 198.51.100.0/24
  neighbor 203.0.113.1 route-map IDACORE-IN in
  neighbor 203.0.113.1 route-map IDACORE-OUT out
  neighbor 203.0.113.1 maximum-prefix 1500000 90
  neighbor 192.0.2.1 route-map CARRIERB-IN in
  neighbor 192.0.2.1 route-map CARRIERB-OUT out
  neighbor 192.0.2.1 maximum-prefix 1500000 90
 exit-address-family
!
ip prefix-list OURS seq 10 permit 198.51.100.0/24
!
route-map IDACORE-IN permit 10
 set local-preference 200
route-map CARRIERB-IN permit 10
 set local-preference 100
route-map IDACORE-OUT permit 10
 match ip address prefix-list OURS
route-map CARRIERB-OUT permit 10
 match ip address prefix-list OURS
 set as-path prepend 64496 64496

Local preference sends outbound traffic via IDACORE. The double prepend nudges inbound traffic the same way. Treat prepending as a hint, not a command, because networks with their own local-pref policy will ignore it. The strict outbound route-maps are the line that matters most. Announce only your prefix, every time, to everyone.

Frequently Asked Questions

How much does IP transit cost in Boise, Idaho?
IDACORE Boise sells IP transit from its own ASN on committed tiers with 95th percentile billing: 100 Mbps for $170/month ($1.70/Mbps), 1 Gbps for $600/month ($0.60/Mbps), 10 Gbps for $2,000/month ($0.20/Mbps), and 100 Gbps for $11,050/month ($0.111/Mbps). Pricing is published, there are no per-GB egress charges, and transit is on-net at both IDACORE Boise and IDACORE North in Coeur d'Alene.

Do I need my own ASN to multihome a colocated server?
Not for blended transit. IDACORE carries your traffic on its own ASN, AS36824, and handles upstream diversity, so you need no ASN, no portable IP space, and no full-table router. True multihoming across two separate providers does require your own ASN plus a provider-independent prefix: at least a /24 for IPv4 or a /48 for IPv6, because smaller announcements get filtered across much of the internet.

How does 95th percentile billing work for IP transit?
Your provider samples port utilization every five minutes, sorts the month's samples, discards the top 5%, and bills on the highest remaining value. In a 30-day month that's 8,640 samples with the top 432 dropped, which works out to about 36 hours of peak traffic that never shows up on the invoice. Short failover events and traffic spikes usually fall inside that window, so the bill stays predictable in a way per-GB egress never is.

Which carriers are available at IDACORE Boise?
IDACORE Boise has seven on-net carriers: Zayo, Lumen/Level 3, Cogent, CenturyLink, Syringa, Cable One, and Hurricane Electric. Most customers buy IP transit directly from IDACORE on its own ASN, which gives them upstream diversity through a single commit. Customers who want to multihome with their own ASN can pair IDACORE transit with a direct contract from one of the other on-net carriers in the same building.

Can I run BGP with IDACORE and a second carrier at the same time?
Yes. If you hold your own ASN and a /24, you can announce it to IDACORE's AS36824 and to a second on-net carrier in Boise, then steer outbound traffic with local preference and nudge inbound traffic with AS-path prepending. That hybrid gives you provider-level diversity, with one leg already blended across multiple upstreams. It also means you own the routing policy, the router capacity, and the on-call rotation for both sessions.

Still deciding between one commit and two carriers for your Boise rack? Send us your traffic graphs and we'll tell you which model fits, including when the answer is a hybrid. You'll be talking with the people who run AS36824, and a 1G commit is $600/month with no per-GB egress surprises. Get a transit sizing review from an IDACORE network engineer.

Ready to Implement These Strategies?

Our team of experts can help you apply these ip transit techniques to your infrastructure. Contact us for personalized guidance and support.

Get Expert Help