End-user performance monitoring in the cloud and multi-cloud era

Online shoppers weren’t the only ones encountering dog images on this year’s Amazon Prime Day. Amazon Web Services (AWS) -- which supports thousands of websites globally -- experienced its own problems, as businesses attempting to access their management consoles also got puppy images.

These issues may have slipped under the radar, but they were significant -- reinforcing, once again, that even the most reputable service providers can and do experience outages (sometimes at the most inopportune times). The companies using them need to take proactive, diligent steps to insulate themselves.

End-user performance management is even more acute for cloud users. They must ensure their cloud-hosted websites and apps hosted are fast, reachable and reliable for geographically dispersed end users. These modern end users have a "right here, right now" mentality. Lightning-fast experiences are critical to engagement, satisfaction and company reputation, whether these end users are site visitors, online shoppers or employees accessing and using productivity applications like Office365 or Salesforce.

Google recently found that increasing webpage load time from 1 to 3 seconds increases bounce rate to 32 percent; up to 5 seconds, this figure balloons to 90 percent. But how can cloud users be expected to manage and control the end-user experience when they lack visibility into cloud infrastructure? Reliance on the public cloud ups the ante and changes the landscape of performance monitoring in several key ways:

Measuring from the Cloud is Helpful, But Not Sufficient

Measuring the speed and reliability of cloud-based websites and applications from a co-located, cloud-based performance monitoring solution may seem like a logical approach. This supports getting as "close" to the website or app as possible to obtain the cleanest view of how it -- and the supporting cloud -- are performing.

This type of monitoring is available through SaaS-based vendors as well as through shared cloud-based measurement nodes (when a vendor deploys nodes specifically on a cloud infrastructure, to provide measurements for numerous cloud users). This monitoring is also available through private enterprise nodes -- when a vendor places private nodes in the cloud to measure applications or workloads running within dedicated cloud instances.

Cloud-based measurements can be very helpful in accurately gauging the speed and responsiveness of a cloud infrastructure. This helps identify when a cloud service provider is, in fact, the root cause of a slow or unreliable website or application. However, these measurements only offer a "first-mile" view and do not reflect end users’ true, ultimate "last mile" experiences. End users do not live in the cloud after all!

Many other factors impact the end-user experience including content delivery networks (CDNs), domain name services (DNS), regional and local ISPs and even end users’ own browsers. End-user performance monitoring in the cloud era, therefore, requires a more comprehensive approach that may originate in the cloud, but must extend further -- across all the performance-impacting elements standing between the cloud and the end user.

Monitoring exclusively from the cloud can be problematic for another reason: if a cloud service provider goes down, so do performance monitoring applications—a scenario that most cloud users cannot afford.

 Multi-Cloud and Hybrid Environments Require Special Attention

The trend towards multi-cloud implementations -- where a company leverages multiple cloud regions within one provider or multiple cloud service providers -- is growing. This is especially true for organizations serving end-users in far-flung geographies, because rarely can a single provider deliver such comprehensive global support.

Today it is not uncommon for a single company to use up to five clouds. Multi-cloud implementations are also common for distributed applications, which run on multiple systems simultaneously to complete a single task or job. 

Adopting a multi-cloud approach has many benefits, including the ability to failover workloads from one cloud service provider or region to another in the event of a primary provider experiencing problems. But it can present unique performance monitoring challenges.

Not only do cloud users need to measure individual performance for various cloud service providers and regions, but they must also measure the speed and reliability of interactions between providers and regions. This is because end-user performance ultimately depends on end-to-end reliability and speed.

The same holds true for hybrid environments -- when a cloud user certain components of a distributed service in the cloud, and other components in-house. Hybrid environments are popular among enterprises that may have heightened needs for speed/reduced latency, reliability and/or security, for certain application components.

A leading bank offering a mobile check deposit application is an example of this. The application front-end may reside in the cloud, while the transaction-processing component may reside in an on-premise mainframe. Ultimately, an end user’s overall strong performance perception requires ultra-fast, seamless connectivity between the cloud and on-premise environment. 

Advanced Diagnostics are a Must

Reliance on the cloud can be a double-edged sword. Sure, the cloud offers many benefits, particularly flexibility and cost-savings -- which is why by the end of this year, more than half of global enterprises will rely on at least one public cloud service for digital transformation, according to Forrester.

But substantial risk often coincides with these benefits, in terms of increasing the number and likelihood of performance-impacting potholes. When multi-clouds and hybrid environments are introduced, the surface area for performance problems multiplies. This further swells the ranks of all external third-parties in the line of service delivery that can wreak havoc on website and application usability.

The existence of so many variables means that a website or application experiencing performance problems is no longer a matter of 'if,' but 'when.' This is why businesses -- especially cloud users -- would do well to shift their focus away from the unrealistic goal of achieving performance perfection, to speeding mean-time-to-repair (MTTR). This means the ability to accurately and precisely identify and fix the source of performance issues, wherever they may be. Increasing reliance on the cloud only intensifies this need.

Conclusion

The key takeaway for businesses focused on maintaining exceptional end-user experiences is this -- do not mistake more bandwidth achieved through the cloud as a magic bullet for end-user performance. Cloud-based monitoring provides a very useful data baseline, but it does not paint the whole picture of end-user performance.

Reliance on the cloud -- whether a simple cloud deployment, or a multi-cloud or hybrid infrastructure -- requires a diverse monitoring solution. Various vantage points (including, but not limited to the cloud) must be included, as well as the numerous cloud-to-cloud interlinks ultimately shaping the end user’s performance perception.

Mehdi Daoudi is the CEO and co-founder of Catchpoint. Mehdi’s experience in IT leadership inspired him to build the digital experience platform he envisioned as a user. He spent more than ten years at Google and DoubleClick, where he was responsible for quality of services, buying, building, deploying, and using internal and external monitoring solutions to keep an eye on the DART infrastructure delivering billions of transactions a day.