A fisherman depends on his net to live. A modern company depends upon its IT networks. As businesses’ IT usage and requirements develop, so too will the need to make them more efficient and able to adapt to the company’s transformations. We have witnessed major growth in network traffic for businesses of all sizes, and this has led many to adjust their networking strategies to cope. Ipanema’s 2012 ‘KillerApps’ study found that 69% of those surveyed didn’t understand how much bandwidth each application on their network consumed, so ignorance about network capabilities is also a clear problem.
As businesses are increasingly thinking about “everything” virtualisation as a method of streamlining and increasing efficiency, so what does it involve on a network perspective, and how are companies set to benefit from it?
The evolution of enterprise private networks
Over the past three decades, we’ve seen enterprise networks develop furiously. The 80s were a celebration of the ‘point-to-point’ protocol infrastructure where there was a direct connection between two sites, for example with private physical circuit, TDM private lines or Frame Relay circuits. Dial-up modem with its now unthinkably slow load-time and techno aural dialling tone, was a common example.
The late 90s saw a major improvement with the burst of MPLS (multiprotocol label switching) bringing fundamental improvements to enterprise networking. MPLS remains the de-facto standard for carrier and service provider IP/VPN services. There were two fundamental concepts behind MPLS:
· The virtualization of private networks: enterprise can make use of a shared physical infrastructure (the carrier transport and switching network) as if it is its own network, with isolation of the information that circulate on that network between each enterprise using the same MPLS infrastructure;
· The automation of the connectivity management: IT managers do not need any more to care about the network topology as MPLS can automatically handle any-to-any traffic.
Moreover MPLS translated into its Class of Service (CoS) concept older mechanisms allowing to differentiate the traffic within a same IP/VPN. Rustic and coarse CoS traffic management used to do a good job to manage simple application traffic mix, like protecting voice over IP from the rest of the data flows.
MPLS brought clear advantages to the market: it’s secure; it is reasonably flexible; and the use of telco’s mutual infrastructure makes MPLS services very stable and reliable. On the other hand, MPLS is not open (you rely on a single service provider) and can be pricey. A true business infrastructure, isn’t it?.
What going on with the public Internet?
At the same time as the well managed MPLS spread within enterprises’ IT, the public Internet became pervasive within people private life and quickly dominated the communication world: browsing, shopping, listening music, looking to video or TV… Internet is everywhere thanks to smartphones and tablets. We are using it at each moment of our life: every minute, the Internet conveys 700,000 GB (!) of data across the planet. Next year, the same minute will see more than 1 petabyte (1,000,000 GB) of information flowing across the fuzzy Internet…
Many characteristics of the Internet are fantastic: it is ubiquitous, prices are low and bandwidth can be huge… But some are scaring for enterprises: it is a fully open public network without security or privacy, with no real governance, lacking any kind of availability or performance guarantee… www: the wild wide web…
Enterprises always tried to use the Internet for their business network. Whilst IPsec (Internet Protocol Security) and TLS (Transport Layer Security) brought a solid shield around their private data, they did not solve the other aspects of using public Internet (management complexity, lack of availability and performance guarantee) and remained so far limited to connecting small and non-critical sites to the corporate network.
Network Virtualisation
Beyond the hype of everything-virtualisation, there is a true similarity between what network virtualisation is looking after and IT virtualisation: the desire to unlock the dependency of the service from the underlying physical infrastructure. Like a virtualized datacentre decouples operating systems and application software from the underlying HP/Dell/IBM/Google… server hardware, a virtual network should be able to run over whatever (reasonable) network infrastructure.
A fully virtualized network has to provide enterprises with the best of both worlds: the security, reliability and predictability of an MPLS network with the ubiquity, flexibility and affordability of the public Internet.
Whilst recent initiatives like SDN (Software-Defined Networking) and NFV (Network Functions Virtualization) are making attempts in that direction, both for LAN (within datacentres) and WAN networks, the reality is that there is currently neither a true nor de facto standard - like VMware for server virtualization - for network virtualization. This is not going to happen soon and enterprises that want to move along that way have to build their own implementation, making use of the different technologies proposed by their vendors of choice.
WAN optimisation
WAN optimisation (taken in its broader meaning) encompasses a large variety of technics that aim to mitigate the barriers of wide area networks: long delays, loss of data, scarce bandwidth, poor control, lack of visibility and so on.
During the last few years, WAN Optimization has enabled major IT transformations like server consolidation in just a few private datacentres thanks to the ability to compress data and accelerate data using protocol acceleration. Next generation WAN Optimization solutions now go beyond these traditional benefits and provide advanced features like application visibility, traffic QoS and Control that are necessary to handle the next set of new It objectives such as implementing unified communications and collaboration, social media, video, professional and personal device mobility (the list is large). The result is a smoother experience for the end user and manageable network governance for IT teams and the business.
Dynamic WAN Selection
‘Dynamic WAN Selection’ (or ‘WAN Path Selection’) enables businesses to combine multiple networks (e.g. MPLS + Internet, dual MPLS, dual Internet, etc.), unifying the management and control of hybrid networks. It automatically chooses the best network for each application flow by taking into account the characteristics of each network path. These include bandwidth assurances and the tendency towards delay, loss or jitter. Using this method, an application which would suffer from large jitter or delay (such as voice) would automatically be sent down the more stable route (e.g. MPLS) while a less sensitive flow (such as file transfer or screen sharing) can use the larger and cheaper route (e.g. Internet), guaranteeing high performance to each application.
Dynamic WAN Selection has quite a spectacular impact on corporate networking, allowing a true combination of MPLS, Internet VPN and local breakout for public cloud applications. As much as 80% of enterprise traffic is usually travelling via Internet access whilst MPLS today carries around 20%. Dynamic WAN selection allows companies to cope with huge traffic increase (think about MS Lync and desktop video) while reaching “five nines” availability level.
Looking forwards
Business efficiency requires guaranteeing application performance across corporate networks. At the same time as business as usual IT departments must grapple with new applications coming online. The situation is becoming more complex as IT teams seek to combine different network to achieve greater value. Future deployments of network virtualization will likely increase the complexity still further and so the need to control and optimise application traffic so that employees can really get the performance they require. Only the future will tell us exactly how the complimentary relationship between WAN virtualisation and WAN optimisation will develop as a result.