Managing Performance for SaaS-Based Applications

Software as a service (SaaS), one of the earliest and most successful cloud services, has reached mainstream status. According to Cisco, by 2019 more than four-fifths (83 percent) of all data center traffic will be based in the cloud, up from 65 percent today. The majority of this traffic will be applications.

Businesses of all sizes are adopting a variety of SaaS-based services - everything from collaboration tools to mission-critical commerce-oriented applications. The rise in SaaS usage has many positive benefits, but one drawback is that as demand grows, SaaS providers are having a harder time ensuring a high-performing (fast, reliable) end-user experience - at the same time performance expectations are growing higher than ever.

SaaS Providers Struggle to Meet Modern Day User Performance Expectations
Websites and applications must be fast. People want to read an article as soon as it's clicked; shoppers will abandon their cart on an ecommerce site if the website is too slow; customers expect to be able to quickly make transactions - and the list goes on. According to Google's latest research (February 2017), as page load time increases from 1 second to 7 seconds, the likelihood of a visitor abandoning the page increases 113 percent. For SaaS companies to be successful, they need to adopt the same laser focus on the end-user experience that has fueled the rise of ecommerce and B2B applications.

Meeting these performance demands is proving to be a significant challenge. One recent survey of SaaS providers found more than half reported difficulties in delivering sufficient performance levels, while more than a quarter said their organizations had incurred financial penalties as a result of unmet service-level agreements. Twenty-one percent of respondents whose organizations had an unplanned service interruption said it resulted in the loss of a customer relationship. Among those SaaS providers suffering financial penalties, the average cost incurred was $359,000 - not including the time and effort spent finding and fixing the source of performance declines.

Why Is Ensuring Superior Performance So Difficult?
Ensuring strong performance levels is proving difficult for several reasons:

Infrastructure Build-Outs - As many SaaS providers expand into new geographies and support more businesses, they are adding infrastructure as well as partitioning more existing systems. According to the SaaS provider survey mentioned above, almost half of respondents noted they have at least doubled the number of systems supporting their workloads and applications over the past two years. While this helps them support more customers, there is a nasty side effect - greater complexity, which makes it harder to find and fix the source of performance problems when they do occur.

Users' Geographic Expansion - Heightening these performance challenges is the growth of end users in new geographies. Application performance tends to deteriorate based on distance from the SaaS provider's data center. This means as end users move further away, there are more variables standing between them and the data center which can degrade performance - networks, ISPs, even browsers. This makes it impossible for SaaS providers to gauge worldwide performance levels for a specific SaaS user, based solely on the experiences of end users in just one or two geographies.

To expand and improve geographic reach, many SaaS providers build out their infrastructure, which in turn adds more complexity. The net result is a vicious cycle that ultimately results in decreased visibility into infrastructure health and end-user application performance.

The Capricious Nature of the Internet - Full SaaS provider outages are rare, but when they do happen, they can be disastrous. The Amazon S3 outage on February 28 served as the latest example, causing widespread availability issues for thousands of websites, apps, and IoT devices.

There have been other instances recently - in February 2016, Office 365 went dark for wide swaths of Europe. In August 2015, a Google data center in Belgium was struck by successive lightning strikes, causing problems for numerous Google cloud infrastructure service users. Not even tech giants like Amazon, Microsoft, or Google are immune to acts of God or nature. Inevitable outages are not the fault of these providers - rather, the "ripple effect" is the result of the tendency for business users to concentrate too much work in the hands of a single provider.

More common than full-blown outages are instances of SaaS applications just not performing well. Even though many SaaS providers have monitoring systems in place, 54 percent report the primary way they find out about performance problems is through customer complaints.

Managing Performance Must Be a Shared Endeavor
SaaS providers need better tools to manage performance, enabling them to get ahead of problems before their customers' end users are affected. This is a fundamental shift from traditional application performance management (APM) to what Gartner refers to as digital experience management (DEM). DEM treats the end-user experience as the ultimate metric, and identifies how the myriad of underlying services, systems, and components influence it. This approach is consistent with a recent EMA survey in which more than three quarters of respondents prioritized the ability to troubleshoot and analyze root causes of application performance problems, down to the platform level.

As the number of performance-impacting variables increases, SaaS providers may find themselves drowning in data, but searching for insights. They need advanced analytics to make data more actionable by identifying the cause of performance issues, swiftly and accurately. Sometimes performance issues are within the SaaS provider's control (such as when a particular region or data center requires more capacity). Other times, the issue may not be directly within the IT team's control - for example, a slow ISP or CDN. Even with these external factors, the information is still useful because time spent unnecessarily "war-rooming" can be avoided.

Business users of SaaS services should also be contributing to the performance management effort. They should consistently measure their SaaS providers' response levels, as well as monitor and measure real end- user performance at the closest points of geographic proximity. This is the key to gaining the most realistic view of end-user performance levels around the world. Having this information can help SaaS users uphold provider SLAs, but perhaps, more important, identify performance problems in advance and quickly determine if the problem is with the cloud service provider, or another component or internal infrastructure element.

Finally, as the recent Amazon S3 example demonstrated, SaaS users should always have a redundancy plan in place for their critical applications. In essence, they need to architect for failure. This may require an investment of time and effort, but can make all the difference in keeping SaaS-based businesses running smoothly when faced with a completely unpredictable event.

Conclusion
The growth in SaaS reliance is not going away, and SaaS providers cannot afford setbacks. When the SaaS platform that employees spend 80% of their day on goes down, productivity, and therefore a company's bottom line, takes a hit.

SaaS providers will continue to struggle with rapid growth and increasing complexity as business users port more mission-critical applications to the cloud and end users demand increasingly higher levels of performance and productivity.

This situation has the potential to escalate unless SaaS providers address the issue head-on with new approaches that evolve APM to DEM. Business users must also do their part. Ultimately, this can translate to more proactive, productive performance management, with less finger pointing and war rooming and more accurate, decisive issue resolution.