Cloud adoption continues to accelerate as organizations seek the very real business benefits of on-demand access to applications, services, data, storage and communications tools. Almost all companies now use some form of cloud, with the vast majority adopting a multi-cloud strategy.
However, many are experiencing mixed results. Only about half say they are realizing the value they expected from their cloud investments, according to PwC’s recent U.S. Cloud Business Survey. Very often, other issues in the technology stack prevent companies from fully leveraging the possibilities the cloud creates.
As we’ve noted in previous posts, a common challenge involves connecting branch offices and remote workforces with the cloud resources they need. Traditional wide-area networks (WANs) that use multi-protocol label switching (MPLS) services to provide connectivity have proven to be less than ideal for delivering cloud apps and services.
Introduced in 1997 by the Internet Engineering Task Force, MPLS quickly became the go-to connectivity option for branch offices. The routing technique labels data packets by type, making it easier to route time-sensitive traffic such as voice and video onto low-latency network connections. That was a marked improvement over previous routing conventions that required packets to move through a series of routers with IP address inspection at each hop.
Of course, that was well before the popularization of cloud computing concepts. MPLS WANs are closed networks that create secure point-to-point connections between the data center and branch offices using tunneling technology. That technique is not well suited for handling large quantities of cloud traffic.
MPLS WANs are set up in a hub-and-spoke model in which each branch location has a dedicated connection to the data center. With no direct access to cloud and Internet resources from the branch, that traffic must be backhauled to the data center for security inspections before being redistributed. This technique, known as hairpinning or tromboning, creates latency issues that significantly degrade application performance.
The additional bandwidth required for these extra network hops also increases the cost of already pricey MPLS connections. Although prices vary based on the carrier, MPLS circuits typically cost about $600 per Mbps per month — about 100 times more than standard broadband Internet.
That cost may have been reasonable 20 years ago when branch offices had minimal Internet usage, but it is becoming a burden now that most companies are transitioning from on-premises apps to cloud solutions. With the shift to so-called “lean IT” operations, some organizations have hundreds of cloud applications running across thousands of branch office and home office connections.
The cost and performance penalties of moving cloud traffic are among the chief reasons organizations are looking to enhance or even replace MPLS WANs with software-defined WANs (SD-WAN). With any-to-any connectivity between branch offices, remote users, data centers and cloud resources, SD-WAN dramatically improves application performance by eliminating the need for backhauling.
SD-WAN can also reduce connectivity expenses by optimizing or even eliminating the use of MPLS circuits. Software-based intelligence evaluates application and network characteristics to route traffic over a variety of lower-cost transport types such as broadband Internet and cellular.
MPLS will continue to be a valuable networking element by providing secure, point-to-point connections for mission-critical and sensitive workloads. However, SD-WAN’s ability to create a hybrid environment that intelligently chooses optimal connections is a game-changer for branch and remote operations.