Stable buildings need solid foundations and enterprise-grade services are no different. If the underlying transport is too erratic no application service will look too pretty. Add in the limited number of routes and long distances between Internet regions and global, Internet-based enterprise services become predictably erratic. So if you’re not going use the Internet as your basis for a global SD-WAN, what are your options? The traditional answer, of course, has been MPLS. But several technological improvements are converging now to offer another choice, what we call the UberNet.
The Case for MPLS
As a privately managed backbone with built-in Quality of Service (QoS), MPLS suffers none of the Internet’s erraticness. Yes, that’s old news, but the strengths and pains of MPLS bear reiteration to understand the value of the UberNet. MPLS services deliver the predictability the Internet is lacking. Whatever contention exists for its backbone is managed by the MPLS provider. Packet loss and latency statistics are more consistent and much lower than those of the Internet. And to back up that point, MPLS services come with guarantees around availability (99.99% per year uptime), packet loss (1% is typical) and latency on a route-by-route basis. Just as important, MPLS services are mature services built for the enterprise. Aside from the SLAs, they come with integrated invoicing, end-to-end delivery and management.
The cost of MPLS bandwidth impacts more than the bottom line. IT managers must economize bandwidth spend to meet budgets. As such, branch offices get sized with just large-enough connections. These narrow connections are increasingly incompatible with today’s larger data flows.
The problem is only made more acute traffic shifts to the Internet and in the cloud. Providing remote offices with direct access to the Internet necessitates securing that connection with a full stack of advanced security services. To avoid those costs, many MPLS-based enterprises centralize Internet access. But centralizing Internet access requires Internet- and cloud- bound traffic to be back hauled to the centralized Internet portal. Precious MPLS capacity is consumed and Internet and cloud performance may degrade due to the well known trombone effect.
The UberNet: An Alternative SLA-backed Backbone
A combination of industry developments including massive global IP transit capacity deployments, accelerated packet processing platforms, and cloud- based software services are enabling a new kind of high-quality, SLA-backed backbone. We call this new backbone: the UberNet.
The UberNet is a global, predictable, and secure network with MPLS-like performance and low costs. The UberNet is built from IP transit services across global tier 1 IP backbones. Internet providers access the greater Internet in one of two ways. If they’re large enough, then other Internet providers will want to access their networks and the two providers will peer with one another, swapping traffic.
If they cannot attract that level of attention then Internet providers will purchase access to a backbone in what are called Internet transit services. Internet transit, the private Internet, avoids erraticness that largely comes from provider peering. Transit services generally keep packets on one backbone. They’re typically backed with guarantees of “5 9’s” availability and 1 percent loss. And transit services cost a fraction of MPLS.
To select between backbones, the points of presence (PoPs) comprising an UberNet monitor the tier 1 backbones for latency and packet loss. The PoPs build an encrypted mesh of tunnels and direct traffic to the optimum tunnel using application-based routing protocols.
Availability means more than multiple backbones. Each PoP is also built from multiple, redundant computing units. Should one computing unit fail, another one automatically takes its place. Here’s the third innovation. Whereas PoPs were built from proprietary hardware and appliance, the UberNet leverages improvements in software architecture and COTS hardware. All core functions in the PoP are implemented in distributed software. No proprietary hardware or appliances are used for core functions. As fully distributed software, the UberNet can be made incredibly resilient at comparatively low cost. And in the unlikely event that an entire PoP should fail or become unreachable, the distributed architectures allows traffic flows to fail over seamlessly to the next closest PoP. There is no direct bind between a customer location or users and a particular provider resource.
MPLS or UberNet?
Every major disruption starts with a displacement of the “tried true.” The cloud displaced virtualization who disrupted the server industry who changed the mini computer market.
Backbones are no different. With the UberNet providing the same level of uptime as MPLS and bringing built-in advanced security, ubiquitous coverage, and support for the cloud, SaaS, and mobility, why would anyone pay 10x more for MPLS?
The Cato Cloud is the first UberNet service. The Cato Cloud converges networking and security into one seamless resource. A single set of security and networking policies govern all sites, SaaS applications, cloud resources, and mobile users. With the Cato Cloud, networking and security becomes simple again.