Optimise your network

The network is rarely the first consideration of businesses when identifying the impact of new cloud computing services: the attraction of increased flexibility and efficiency and reduced cost usually hits the top of the agenda.

However, the impact of cloud computing on the network can be substantial, and it is unwise to overlook the consequences. Without a reliable, high-performance network linking the cloud datacentre to users’ locations, cloud services will operate poorly.

There are many considerations CIOs must bear in mind with regard to that network. Performance is vital: high availability is essential and latency must be less than the required threshold. Furthermore, businesses must require of the network service provider SLAs that are appropriate to the cloud services in question.

Application performance

Wherever applications are located, their performance from an end-user perspective is paramount. Availability and latency are two of the most important network metrics. Compared with on-site hosting (all other things remaining equal), availability and latency are likely to suffer with the move to cloud computing.

Latency issues can be addressed using WAN optimisation, which streamlines the delivery of applications between sites, for certain applications: it is generally excellent at accelerating CIFS- and MAPI-based traffic, but ineffective with voice.

WAN optimisation can be used between user sites and the datacentre, but will also become increasingly useful for workers needing to access cloud resources on the move. For public cloud services, WAN acceleration is more difficult because of the lesser control offered to the business concerned.

Location and topology

Latency can be substantially affected by the location of the cloud datacentre. Where a third-party datacentre is used, CIOs should clarify with the third-party provider exactly where their data will be stored. Aside from ascertaining the effect on latency, understanding this is critical for compliance.

Latency is heavily affected by network topology. Where traffic between the cloud datacentre and users is forced through one of a limited number of network nodes – such as a centralised internet gateway – as opposed to direct links to the location of the user, latency is increased.

Should that delay be unacceptable, new, more direct links may be required, with security controls being moved from the central location into the cloud as well. A lot of latency issues are introduced by the parallel use by businesses of internet-based IP VPNs and separate MPLS networks, which make a great deal of sense from a cost and security perspective, but do not aid application performance.

According to the analyst Gartner, more than three in five businesses will experience slow or inconsistent application performance from hosted applications over the next three years because of their network topology.

One tool that can help reduce latency for some applications is a content delivery network, which can be provided by specialists such as Akamai, or by telecoms service providers.

High availability

Availability is clearly crucial, too. Uptime can be safeguarded by a number of factors. Businesses should consider the installation of redundant links for their major sites in case of failure on their primary link. Service-level agreements with carriers should meet the requirements of the new cloud services. Businesses should ensure their continuity strategy is updated to take account of cloud services, and they should quiz their cloud provider on their continuity strategy.

Network costs

Calculating network traffic costs is also important and can affect the business case for cloud computing. However, the pay-as-you-go payment structures involved can be complex, based on data transit, storage and CPU usage, among many other variables.

Firms may therefore find their network costs increase unexpectedly, and some will prefer a more traditional, flat-rate hosted package from a telecoms service provider.

The cloud in practice

One business that has been grappling with extensive cloud considerations is newspaper publisher Telegraph Media Group. Facing a considerable decline in group revenue, its chief technology officer Toby Wright has implemented a transformation of the company’s IT from on-premises to cloud, with Microsoft Office and Exchange swapped for Google Apps and Gmail. Delivery of content is now also from the cloud, with the aid of a content delivery network.

Wright says that, having modelled traffic load, he needed to make no major changes to his network. SLAs have changed: he now has one for the WAN and one for the cloud service, with the latter being common to all customers, simple in scope and not end to end. Message archiving and retrieval is now much simpler, he says.

Transitioning to cloud computing has allowed Wright’s network team to be downsized, with some of the retained server and infrastructure staff being retrained to work on supporting cloud services. The CTO says he is happy with the level of network support. “Potentially, I have lost control of my infrastructure, but if my on-premises solution broke, does Microsoft send someone in 10 minutes? When Google breaks, they fix it for everyone,” he says.

Wright’s confidence in cloud computing is striking, underlining the belief in cost savings that can be achieved. But businesses must look beyond the return on investment argument and consider the network too. With poorly performing applications, there can be no ROI from the cloud.