Guide, 12 Nov 2026, read 12/13 Nov 2026
Only a few SASE vendors sell or manage the last-mile circuit, so plan dual links yourself
The circuit is the usual cause of branch outages
In both of my migrations, the most common branch outage was the ISP circuit, not the SASE service. A vendor can publish five nines for its PoPs and a branch can still be down for six hours because a cable was cut. So the last mile deserves its own section in the design, with a named owner.
The terms matter: last mile is the circuit from your site to the ISP network. First mile is the same segment viewed from the user. The SASE service starts at the PoP.
Three levels of vendor involvement
From the vendors' pages I find three levels. First, vendors that source and manage the circuits. Second, vendors that monitor circuits and work with the ISP. Third, vendors that only steer traffic across links you buy.
| Vendor | Last-mile involvement as published | Failover as published |
|---|---|---|
| Open Systems | Sources, contracts and manages ISP lines under one contract; optional 24x7 Line Operations Service | 99.99% end-to-end availability with Last-Mile Connectivity (vendor figure) |
| Cato Networks | Monitors last-mile ISPs 24x7 at every site and works with the ISP on outages (managed service) | Socket at the site; link detail none found on 12 Nov 2026 |
| Aryaka | Last Mile Services: procurement, management, monitoring | Repair time and availability follow the third-party provider; 99.99% end to end needs ANAP HA and dual ISP at both sites |
| Netskope | No managed circuit service found on 12 Nov 2026 | Built-in 4G/5G with access to 400+ cellular networks |
| Check Point | No managed circuit service found on 12 Nov 2026 | Fails over in under a second to MPLS, 5G or broadband links (vendor) |
| Arista VeloCloud | No managed circuit service found on 12 Nov 2026 | Aggregates links with DMPO; broadband, MPLS, LTE and satellite listed |
| iboss | No managed circuits found on 12 Nov 2026 | Multiple WAN ports with active/active load balancing and active/passive failover |
| Barracuda | No managed circuits found on 12 Nov 2026 | Multi-uplink SD-WAN; optional USB LTE modem for failover, load balancing or primary link |
What dual links need to be real
Two circuits from the same carrier on the same street entrance are one circuit with two bills. Ask for different carriers or at least different physical paths, and ask the carrier for the building entrance. Add LTE or 5G as a third path for sites where an hour of downtime costs real money.
Netskope states its built-in 4G/5G offers access to 400+ cellular networks, and Cisco states Meraki MX models include integrated LTE. Barracuda says a USB LTE modem can serve as a failover, a load-balanced path or a primary link. These are options, not a design. A cellular failover has limits: lower bandwidth, higher latency and carrier-grade NAT that can break inbound services.
Test the diversity claim before you pay for it. Ask the carrier for the route map or at least the central office each circuit terminates in. Two circuits that share a central office share a single point of failure that no SD-WAN policy can steer around. Also check the power: a second link on the same unprotected power feed fails together with the first.
Reading the managed-circuit claims
Open Systems states that because it sources and manages first- and last-mile ISP lines, its end-to-end SLAs cover site to PoP as well as PoP to PoP. That is a different promise from a PoP-only SLA. It also means you rely on its carrier choices in each country.
Aryaka is more careful in its text. Its SLA sets an end-to-end uptime target of 99.99% for site pairs with ANAP HA and dual ISP on the Last Mile Service at both sites. Its services terms state that repair time and availability for circuits follow the third-party provider, and Aryaka is not responsible for an ISP failing to restore service. So a managed last mile does not remove the carrier from the fault chain.
What monitoring buys you
Cato's 24x7 last-mile ISP monitoring is a stated managed service. Monitoring shortens the time to know it is the ISP. It does not shorten the time the ISP takes to fix it. Fortinet's data sheet describes digital experience monitoring with first-mile and last-mile checks. That can give you the evidence for an ISP escalation.
Ask each vendor whether the monitoring is on by default and whether it covers every site or only the ones you flag.
Procurement notes
If you buy circuits from the SASE vendor, keep the right to move a circuit to your own contract. If you buy them yourself, get one contact at the carrier who owns the ticket. Align contract end dates across carriers so a renewal does not leave one site on a month-to-month price. And test failover on a live site before sign-off: pull the primary link on a quiet Sunday and time the recovery.
Write the failover target into the design: for example, a session should survive a primary link loss with no more than a few seconds of interruption. Check Point states sub-second failover and Open Systems states 99.99% end to end, but those are vendor figures, and only a test on your own site confirms them.
Finally, keep an asset sheet per site: circuit ID, carrier, bandwidth, contract end date, demarcation location and the carrier support number. When the branch is down, the person on call should not have to look for any of it.
Sources
- Open Systems, SD-WAN backbone, read 12 Nov 2026
- Open Systems, connect China, read 12 Nov 2026
- Cato Networks, services, read 12 Nov 2026
- Aryaka, service level agreement, read 12 Nov 2026
- Aryaka, services terms, read 12 Nov 2026
- Netskope, Converged Gateway, read 12 Nov 2026
- Check Point, Internet Access, read 12 Nov 2026
- Arista, VeloCloud SD-WAN Edge 7x0 datasheet, read 12 Nov 2026
- iboss, SD-WAN, read 12 Nov 2026
- Barracuda, SecureEdge features, read 12 Nov 2026
- Cisco, Meraki security appliances, read 12 Nov 2026
- Fortinet, FortiSASE data sheet, read 12 Nov 2026
Last reviewed 12 Nov 2026.