Beyond best effort: why some apps need a better cloud performance

When tuning cloud applications for performance, lag, latency and location are all related.

Cloud-based applications can deliver cost and convenience benefits. Critically, they release corporate IT teams from the demands of maintaining internal infrastructure, enabling them to focus on core digital transformation programs. However, if implemented without proper planning, cloud application performance can be unpredictable and difficult to manage.

 

Global enterprises face daily challenges connecting global users to online applications that they depend on for their work. In many cases, these applications are hosted in other countries, or even other continents.

This can lead to network latency problems that impact the user experience. Existing local internet connections are not good enough for most cloud usage scenarios. Performance varies due to the level of contention on the lines during periods of high local usage.

Performance is paramount, together with security, when moving away from “on-premise” hosting to a cloud deployment, according to Paul Pizzey, Business Partner at Orange Business. “When customers speak to us, one of their major concerns is whether performance will suffer. They need assurance of a good end-user experience when applications are hosted on a server miles away.”

Microsoft’s Office 365, one of the world’s most popular cloud computing productivity applications, suffers from this problem. A 2016 Gartner study revealed that one in five enterprise users reported performance issues when using it.

Location, location, location

Today’s cloud applications require a radically different approach to connectivity. Traditionally, when running applications on-site, enterprises consolidated workloads into a handful of regional data centers serving local users.

In the move to a software as a service (SaaS) model, workloads tend to be run as a single global instance from one data center and traffic is transported over the internet. This centralized cloud application deployment model is often combined with a traditional hub-and-spoke network set-up for securing the application Internet traffic via WAN appliances at a headquarters office. For a multinational company, this means that employees who are based on the other side of the world from their application software and data can face latency issues.

This problem is growing, said the analyst firm. It predicts more than half of all Office 365 deployments will suffer from network performance problems by 2019.

When it comes to remote application access, a few milliseconds can make all the difference. Microsoft recommends round-trip latency of below 275ms for Office 365 deployments, and less than 50ms for Exchange Online traffic. SharePoint Online is even more demanding – round trips exceeding 30ms can make the product virtually unusable, says Gartner.

WonderNetwork is a network testing company that maintains servers around the world for performance monitoring services. Its tests show that Internet latency varies widely between cities. At the time of writing, a packet travelling from Toronto to Amsterdam and back again was taking 115ms. The round trip from Tokyo to Barcelona was 272ms. Add to this a requirement to “hair-pin” traffic through a WAN appliance at a headquarters and employees will suffer from applications that take far too long to load data or save updates.

Three approaches for improved performance

Enterprises have three options when addressing these problems. The first is distributed Internet access, which provides IT teams with local Internet breakout points where the traffic is secured. To benefit most from this approach, enterprises need to prioritize business-critical apps, like Office 365, from a bandwidth perspective over less important, bandwidth-hungry services, such as YouTube.

However, internet-based connectivity is still a “best effort” service without guarantees of performance, and leaves the enterprise vulnerable to outages and disruption caused by the variable performance of the internet or DDoS attacks.

The second option is to enhance the internet backbone itself by controlling the paths taken across the two locations and caching the traffic. This can help avoid contention on the connections to deliver more consistent cloud application performance.

The final option is to use direct WAN connectivity, which extends an enterprise’s WAN router to a carrier network hub that is directly connected into the SaaS provider – avoiding the use of the Internet for traffic routing entirely. This type of cloud interconnect service enables end-to-end SLAs and QoS in addition to improved and stabilized performance, and reduces security and privacy risks.

Microsoft points to performance of up to 10Gbit/sec using this approach, depending on the enterprise WAN. It is particularly valuable in parts of the world where the internet may suffer consistently poor performance due to government filtering of content. Optimization software can be used to de-duplicate and compress data sent over WAN links, minimizing congestion by reducing the amount of data that is transferred.

In addition, by dealing directly with a single carrier for Internet backbone and enterprise WAN, enterprises can get network latency guarantees from desktop to the application server. This will help them overcome routing and performance issues, while minimizing costs.

Ultimately enterprises need to choose the best connectivity options depending on the distribution of their employee base and the criticality of applications being used.

Whichever route they take, enterprises need to address the cloud performance issue proactively, concludes Pizzey. “Delivering a great user experience is a key measure of success for any cloud project. So, more than ever, businesses need to plan for and measure actual user experience a crucial activity for the IT department,” he warns. “Don't find out too late that user experience is suffering.”

Download our cloud ebook ‘Create a cloud experience your business can depend on to find out more about application performance best practices in cloud and overcoming the most common cloud challenges.

 

Danny Bradbury

Danny Bradbury has been writing about technology since 1989. He covers consumer and enterprise technology subjects for a variety of publications including the Guardian, the Financial Times and Canada's National Post.