Branch offices come in many sizes and purposes – from small to large, and from critical functions to a simple home office. The enterprise needs a network that can adapt, offering availability levels to meet each type of office requirements. What are your options?
MPLS networks have been the backbone of enterprise networks for years. Although MPLS circuits are considerably more expensive than general Internet circuits, businesses have relied on MPLS networks for their dependability. MPLS networks are known and relied upon for high uptime, with a target of “five-nines” (99.999%) uptime. Service level agreements (SLAs) guarantee latency, packet delivery, and availability. With an outage, the MPLS provider resolves the issue within a stated period or pays the requisite penalties.
Software-defined wide area networking (SD-WAN) is making organizations rethink their WAN infrastructure. Instead of connecting a location with one highly-available MPLS connection, SD-WAN can connect a location with multiple, less-reliable broadband Internet connections, selecting the optimum connection per application. Ultimately, the goal is to deliver just the right performance and uptime characteristics by taking advantage of the inexpensive public Internet.
Reliability at a Price
MPLS services remain significantly more expensive than Internet services. At customer premise-based data centers, traditional Internet connectivity might offer a 2x – 4x price/bit benefit over MPLS, while at colocation facilities, the price/bit benefits are typically in the 10x – 50x range.
Adding MPLS bandwidth is a lengthy, costly process, requiring configuration changes, and additional hardware taking anywhere between 3-6 months. Waiting on critical additional MPLS bandwidth results in project delays and lost revenue. Because of the high costs, redundancy is often too expensive, leaving companies to connect locations with a single circuit. Internet backup may be used but that adds complexity to the network.
MPLS networks are not infallible, and outages do occur from events such as accidental cable cuts. Another factor affecting performance and uptime in an MPLS network, the last mile, may involve more than one carrier to create the network. The carrier who delivers the last mile varies by location and may not be the carrier providing the MPLS service. Oftentimes, SLAs are limited to the backbone and not the last mile where outages are likely to occur. Performance and uptime could be unpredictable if the last mile carrier does not meet the expectations of the MPLS carrier.
SD-WAN High Uptime Strategies
SD-WAN created more flexibility and the ability to overcome the high bandwidth costs of MPLS services by integrating Internet transports (such as cable, DSL, fiber, and 4G) into the WAN and forming a virtual overlay across all transports. With features like load balancing and measuring the real-time transport quality of each circuit, SD-WAN provides the high uptime businesses demand by using a mix of Internet connections.
Connecting an MPLS service with an individual line means possible line failures from cable cuts, router misconfigurations, and other cabling infrastructures. With SD-WAN, active/active load balancing configuration protects against such failures by using redundant active lines to connect locations to the SD-WAN. When one line fails, traffic fails over to the alternate connection.
The equation for calculating network availability of a location using SD-WAN with multiple circuits shows that the combined availability of multiple circuits in parallel is always higher than the availability of its individual circuit; SD-WAN can compete with MPLS in high availability requirements.
Site Availability = 1-((1-Service A Availability)*(1-Service B Availability)*(1-Service N Availability))
Network availability and downtime for individual consumer grade 99% availability circuits and the parallel combinations:
| Component | Availability | Downtime |
| X | 99.0% (2-nines) | 5256 min/year (7.3 hours/mo) |
| Two X circuits operating in parallel | 99.99% (4-nines) | 52.6 min/year (4.4 minutes/mo) |
| Three X circuits operating in parallel | 99.9999% (6-nines) | 0.526 min/year (2.6 seconds/mo) |
| Four X circuits operating in parallel | 99.999999% (8-nines) | 0.00526 min/year (0.026 seconds/mo) |
By adding circuits in a load balanced configuration with redundant components for high availability, uptime is increased with each additional circuit. With this method, it’s possible to reach five 9s with services that individually offer less than five 9s uptime. Adding LTE or cellular access at a location eliminates the risk of a line cut by the local loop, or last mile provider, enhancing availability with increased fault-tolerance.
Being able to mix and match circuit types and quantity allows each branch office to meet the availability requirements determined by the organization. Some examples of connectivity an organization may choose to meet requirements without overspending:
Critical branch – Redundant fiber with local SLA
Regional branch – A mix of DIA and broadband
Small branch – Redundant broadband
Organizations rely on MPLS for consistent response time for real-time applications such as voice and video. To provide a similar level of consistency, SD-WAN networks automatically detect blackouts and brownouts. When latency and packet loss increases, it can failover active sessions to use a better performing circuit. Look for an SD-WAN solution that provides Fast Session Failover that occurs quickly – in the 100-200ms range. Real-time traffic like voice and video will lose their sessions or experience jitter and delay if the failover takes too long.
Cato Networks SD-WAN includes technology like global, affordable, SLA-backed backbone with over 30 PoPs worldwide fully meshed over multiple tier-1 IP transit providers. Strategies like active/active failover, Application QoS, Policy-Based Routing (PBR), and Forward Error Correction (FEC), give SD-WAN from Cato Networks the high uptime organizations need.
Subscribe to Cato’s blog for the latest topics related to SD-WAN.