The move of processing to hyperscale Infrastructure as a Service (IaaS) and the move of on-premise apps to Software as a Service (SaaS) looks set to continue. Scale economies for the providers mean lower unit costs and flexible consumption models where you “pay as you grow” shift expensive capex purchase to flexible opex charges that can scale up and down with your business.
But getting the right performance from cloud-based services remains complex. Yes, there is ubiquitous access to all cloud services by just connecting to your local internet – but this does not always give you the quality connectivity you need to ensure a great user experience every time. In fact, surveys have suggested that less than 50% of enterprises rely on basic internet connections to access their IaaS providers.
Options to connect to the cloud
At one end, for ease, ubiquity but offering the lowest and most variable performance you have local internet services from you ISP. Here your traffic will trace whatever path is set by the routing policies of your ISPs to reach the Region or Zone where your Cloud Service Provider is hosting your service. You can also use the internet to link your cloud platform into your secure SD-WAN network by deploying a software image of your choice of SD-WAN within your compute area, however you will still not control the route traffic takes between source and destination.
At the other end you have dedicated interconnect. These are direct connections where you hand your traffic into the cloud service provider at locations termed “on-ramps”. There are many on-ramps, Google Cloud has the most at around 120, AWS with around 100, Azure with around 85, Alibaba with around 60 and the likes of Oracle and IBM Cloud with around 40. However, you need to aggregate your traffic, get it to the right on-ramp and have sufficient traffic at that location to connect at speeds of 1Gb. This can be complex for the multi-cloud environment that most organisations use and would seem to re-impose the very hub and spoke architectures that “cloud” and “internet” enabled them to leave behind.
The middle ground aggregators
In the middle ground you will find many companies that will aggregate your traffic onto their own private global backbones. The traditional Network Service Providers each use their own global private networks (typically MPLS) and have linked these into various on-ramp locations. This is great if you already use their services, but otherwise you need to get to their PoPs.
Of perhaps more interest are a range of “new” entrants who have built their private networks specifically to interconnect between a range of third-party data centres and a variety of cloud on-ramps. Companies such as BSO, Packetfabric and Megaport offer access to a range of on-ramps for AWS, Azure, Google Cloud, Alibaba Cloud, IBM Cloud, Oracle Cloud as well as some SaaS providers such as SAP and Salesforce. Their offer is simple – one connection to their network and you can set up your cloud connectivity from a portal with connections delivered via automation in a matter of hours. However, you still need to get onto their network – you typically need to cross-connect at one of their many access locations. BSO offer 240 such access locations, Packet Fabric 295 and Megaport 700 locations.
A novel approach to the problem
Megaport have realised that for distributed or branch locations, all of these carriers have just substituted the problem of “link to an on-ramp” with “link to our PoP”, hence have just launched an interesting new service which they are calling Megaport Virtual Edge. Here they are deploying NFV hosting servers at some of their network locations – so far 13 in the Americas, 7 in Europe and 8 across APAC. They will then enable their customer to deploy a software image of the SD-WAN they use on this device, whilst hosted on Megaport hardware, is controlled as an integral part of their customer’s SD-WAN deployment. Internet access is also provided into these NFV devices meaning that traffic can now be routed over a short internet hop to a nearby Megaport Virtual Edge node, then via the Megaport private network around the world to access any of the 200 or so cloud on-ramps that they connect to. The service has initially been built enabling deployment of Cisco SD-WAN with other vendors to follow.
The best of both worlds?
With a solution such as this, enterprise networks can be built using internet branch connectivity, bringing benefits of ubiquity, local breakout and of course cost, yet mitigate the performance issues arising from long distances and unknown transit hops to target cloud destinations. They can enjoy the automated multi-cloud connectivity of the new range of backbone providers not just from their data centres, but from all distributed locations. This is just another example of how new technologies and new services are continually being introduced to make internet transport even more competitive against traditional MPLS or fixed line Ethernet.
Have a question? Get in touch!
Here at Brodynt, we specialise in procuring and managing the optimal internet services to support your network, whatever SD-WAN vendors you choose and however you chose your cloud connectivity.
You are always welcome to drop us an e-mail to firstname.lastname@example.org and one of our Account Managers will be happy to assist you! Alternatively, engage in a discussion by leaving a comment below and don’t forget to share the article with your friends and colleagues!