The impact of cloud computing on networking is potentially critical to any organisation, with staff accessing databases, applications and other materials over the Internet from a corporate network or from other alternative remote locations every day.
Whilst the cloud can be used to store company data and deliver applications without the overheads of dedicated in-house servers and other hardware, the third party hosting the data or applications is not usually responsible for the network links from the customer to its facilities.
Companies therefore have to plan how they connect their corporate networks to third party cloud providers to ensure they have enough affordable and reliable bandwidth to serve their users' needs.
They must consider cost, service level agreements, security, network back up/failover to the cloud provider, network management, and any specialist 'network to the cloud' service offerings on the market.
Cloud data explosion
A recent report forecasts just how dramatically clouds are transforming business IT and consumer services. The study predicts that global cloud traffic will soon account for nearly two-thirds of total global datacentre traffic.
Globally, cloud traffic will grow from 39 per cent of total datacentre traffic in 2011 to 64 per cent (almost two-thirds) of total datacentre traffic by 2016.
Global cloud traffic will grow faster than overall global datacentre traffic too. The report says global datacentre traffic will grow four-fold (a 31 per cent compound annual growth rate) from 2011 to 2016, while global cloud traffic will grow six-fold (a 44 per cent CAGR) over the same period.
Dr Graham Oakes is a consultant and author of 'Project Reviews, Assurance and Governance'. Over the last 20 years he has worked on IT and networking projects for the likes of Sony Computer Entertainment, Vodafone, the Open University, Oxfam, and the Council of Europe.
Oakes says of cloud migrations, "Moving applications into the cloud clearly has an impact on network utilisation. But before buying more bandwidth, you have to do a couple of other things."
Oakes believes organisations first have to understand the traffic patterns associated with each app. Some apps are "surprisingly frugal" on the network, he said, while some are "surprisingly chatty." Making decisions based on data, not assumptions is the key, he continued.
Also remember that data latency is as important as bandwidth. Often it's not the amount of data you're transferring that matters, but the amount of time you have to wait while apps build web pages or are fully loaded onto desktops. High latency means data packets are having to rely on multiple round trip journeys across many network portions that make up your cloud operating jigsaw.
What Oakes is talking about is illustrated in research about the effects of certain apps on the network, which shows that corporate networks across Europe are experiencing significant performance problems due to network 'blind spots' around the applications being run across them.
Networked app performance
Network management firm Ipanema Technologies and network provider Easynet questioned a mixture of CIOs, IT directors and network managers for its 'Killer Apps' survey of 550 respondents at large enterprises.
The survey found that performance problems are highly prevalent across networks, with 74 per cent citing enterprise critical applications: line of business (26 per cent); enterprise resource planning (ERP) or customer relationship management (CRM) (26 per cent); and video-based applications (22 per cent) are those most at risk.
User complaints were the second most common source of monitoring issues across networks, admitted respondents, indicating that many network monitoring systems were failing, or that companies weren't using them properly.
The survey revealed that nearly one in three respondents didn't even know the number of apps running on their corporate network, and that 69 per cent did not have visibility of the bandwidth requirements each network application required.
Thierry Grenot, CTO at Ipanema Technologies, said enterprises need to be "more agile" to fully benefit from the expected advantages usually delivered by cloud computing. He said firms have have to understand their networks more thoroughly in order to target bandwidth to the applications their businesses rely on, while aiming to reduce unnecessary investment by not throwing bandwidth at less critical applications.
Oakes said that good WAN (wide area network) optimisation between the cloud facility and the corporate network is key to cloud deployments, citing efficient caching and protocol optimisation as established ways to improve both latency and bandwidth performance.
Tweaking your network for the cloud
Whether its timelines, security, cost, bandwith or network protocols and performance to consider, Steve Salmon, principal advisor within KPMG's CIO Advisory practice, said network administrators have to "tweak" their networks to prepare for the cloud.
"Once data reaches the customer network from the cloud, there is work to be done by the network administrators to ensure that the information reaches the users in time to provide a rich or enhanced experience," he said
According to Salmon, reducing any latency in the network is key, and can be done by revisiting the network design to reduce bottlenecks and removing unnecessary switching or firewall infrastructures. He also said that the uptake of cloud has seen the rise of WAN acceleration services, where traffic can be prioritised, encrypted and sped up through branch networks of the customer's distributed IT environments.
Linking to cloud providers
James Wright of the infrastructure consulting group at Accenture echoes Oakes' views about getting the corporate network right before thinking about branching out to the cloud.
Wright said, "Until you have flexibility at the infrastructure level, including LANs (local area networks) and WAN networks, you will struggle to realise the benefits of shifting parts of your operations to the cloud.
"Put simply, it can very quickly become impractical, expensive and complicated, not to mention inoperable, to proceed if you don't get these areas of the cloud project right."
Wright said that large organisations should consider an MPLS (multiprotocol label switching) network connection to link their corporate network to an external cloud provider. He said MPLS can go a long way to making sure the right people get access to the right applications at the right time, effectively addressing security, bandwidth performance, latency and availability requirements.
"This is why implementing voice over MPLS is a great entry point for the corporate cloud. Most organisations give all their people a desk phone which sits alongside their workplace apps," said Wright. "Therefore, once the voice is in the cloud the CIO has confidence that whatever else he or she wants to put in the cloud then it will work, from a network access perspective."
Image credit (middle): Flickr (akakumo)
For more information on HP cloud services, visit http://www.hp.com/uk/cloudinaday