It’s well known that 4G/LTE is a natural partner of SD-WAN, forming a key component in last-mile resiliancy strategies. But 4G/LTE also brings certain characaeristics that with the wrong SD-WAN solution can wrack up costs — as one of my customers learned the hard way.
We were recently called in to consult to a large manufacturer. The company had deployed 4G/LTE and SD-WAN on their own — and ended up being hit with a 4G/LTE bill of $96,000/month for a single site.
To understand what happened and how you can avoid tripping up your deployment, let’s take a closer look at the whats, hows, and whys you might plan for 4G/LTE when transitioning from MPLS to SD-WAN. Specifically, we’ll look at:
- MPLS’s last-mile availability problem
- How SD-WAN brings last-mile redundancy to all sites
- Limitations of last-mile redundancy
- 4G/LTE Defined
- Typical 4G/LTE last mile pricing and deployment
- What’s the risk of using 4G/LTE in the last mile?
- SD-WAN Experts recommendations for 4G/LTE
MPLS’s last-mile availability problem
As managed offerings, MPLS services are well known for being engineered for uptime. Typical MPLS service level agreements (SLAs) include 99.99% network uptime and actual MPLS availability will approach 99.999%.
But as a rule, those SLAs and uptime metrics only apply to the MPLS core. Last-mile availability is highly dependent on the type of last-mile connection. To achieve the same 99.99% uptime SLAs, for example, most MPLS providers will require redundant MPLS connections at the site. Given the relatively high cost of MPLS circuits, enterprises can often only justify redundant MPLS circuits for their largest or most critical sites.
Smaller or less critical locations end up being connected with individual MPLS circuits, possibly with Internet backup. Service across individual circuits can be disrupted by a misplaced backhoe, a routing problem in the MPLS provider’s network, or a range of other problems normally pertaining to the physical layer of the network. And with the Internet backup, failover is often not automatic and where it is, the time to converge noticeably disrupts network operation.
How SD-WAN brings last-mile redundancy to all sites
As we’ve noted, SD-WAN naturally addresses last-mile problems in many respects far better than MPLS. Good SD-WAN design all but assumes a site will be connected by at least two circuits — all Internet or a mix of MPLS and Internet. The SD-WAN software uses a combination of preconfigured policies and real-time analytics to steer incoming traffic across the optimum, last-mile connection.
With multiple connections, the possibility of a single point of failure is removed, and by using both connections IT never wastes bandwidth. Should one circuit be cut or experience a brownout, SD-WAN appliances will automatically switch traffic to the secondary connection, switching back to the primary connection when available.
Limitations of last-mile redundancy
At the same time, last-mile service redundancy can be trickier than you might initially expect. Purchasing separate Internet or MPLS circuits might still not guarantee a redundant last mile.
Too often buildings will have a single main-point-of-entry (MPOE) for the incoming fiber for telecom service. Multiple service providers deliver their services across this provider’s infrastructure. At other times, when there are multiple fibers entering a building they too often share a common ducting. In either case, an errant misplaced backhoe can sever the duct holding the fiber(s), disconnecting both circuits.
Determining last-mile diversity is possible but challenging. Buildings will differ block by block, town by town. All too often the sales reps and sales engineers of a service provider, and certainly an SD-WAN provider, will have direct knowledge of the infrastructure coming into a particular site. If you have the right contacts with the various cities and telcos, you can find out for a given site but it’s not simple.
An easier approach is to avoid the issue. By using different types of last-mile technologies you all but force diversity in the physical infrastructure. This won’t guarantee you core redundancy — the xDSL and cable connections might still terminate at the same router or involve the same backbone carrier — they will be all but assure you last mile redundancy.
Want to connect a small or home office with an SD-WAN? Use fiber and DSL connections, for example. Where that’s not possible or better capacity is needed, look at complementing wired connections with 4G/LTE as the tertiary connection. As a high-speed, last-mile wireless technology, 4G seemingly makes the perfect, backup.
4G, as defined by the ITU-R standards, describes networking services for mobile and stationary devices at speeds previously unheard of across 3G networks. 4G mobile services are expected to have peak line rates of 100 Mbits/s; stationary services should hit 1 Gbits/s, though I have never seen this in practice.
LTE or Long Term Evolution is a direction towards 4G speeds. ITU-R allowed those products that provided a substantial improvement over 3G but not quite 4G speeds to be named LTE. Long-term Evolution Advanced (LTE-A) gets even closer to 4G than LTE.
Typical 4G/LTE last-mile pricing and deployment
Without having to pull last-mile fibers, 4G/LTE deployment is very simple. A 4G/LTE router from vendors, such as Cradlepoint or Sierra, receives the 4G/LTE signal, sending it to an attached SD-WAN device using an Ethernet hand-off. (Some SD-WAN devices also come with internal 4G/LTE receivers.) The 4G/LTE carrier will provide the router with a SIM card to connect to their service.
4G/LTE service pricing involves a fee for a certain amount of pooled capacity between sites above which you pay by the number of gigabytes over the pool amount. Of course, rates will vary but in our experience, customers of SD-WAN Experts usually can get away with paying $25/month for 100 Mbits/s if purchased directly from the carrier and around $200/month if purchased through a managed network service provider, with a pool of 5GB.
What’s the risk of using 4G/LTE in the last mile?
As a metered service, 4G/LTE is fundamentally different than other last-mile solutions where you’re providing a set monthly fee regardless of the amount of bandwidth consumed. The more 4G/LTE bandwidth used, the higher your bill can be. Given that LTE is often a tertiary connection, only used when there’s a significant brownout or blackout in the other two last-mile networks, costing is normally negligible.
Except, that is, with the wrong SD-WAN solution. In order to switch between last-mile circuits, SD-WAN appliances must constantly sample the line to detect any brownouts or blackouts. With a metered service like 4G/LTE, the frequency at which you monitor that line makes an enormous difference. In the case of our customer, their SD-WAN devices did not account for last-mile differences, setting one heartbeat interval for all networks. The “fat” protocol sampled the line so frequently that over the month the total usage far exceeded plan capacity, increasing costs.
SD-WAN Experts’ recommendations for 4G/LTE
To avoid similar experiences, SD-WAN Experts recommends understanding how your SD-WAN solution will monitor last-mile connections. Prefer SD-WAN solutions (as for as last-mile availability is concerned) that can adjust sampling rates for each last mile.
When creating your RFP be sure to ask prospective SD-WAN vendors questions around LTE usage. Two questions we always ask — “How do you test if an LTE connection is up?” and “How much bandwidth is required to monitor 4G/LTE over a 24-hour period?”
Of course, there are numerous other factors as well to keep in mind. 4G/LTE isn’t the only last-mile technology nor even the only last-mile wireless backup solution. This says nothing about the broader issues around last-mile and service availability. If we can help address those and other issues, give us a shout.