PaaS: A performance monitoring headache

PaaS is at the heart of any cloud-native application design, and the use of PaaS is growing much faster than the now traditional IaaS (Infrastructure as a Service). 

PaaS frees up your team’s time from spending time managing and monitoring systems: you only focus on data and code! 

While IaaS spending is expected to still be larger in volume from 2020, PaaS spending is growing much faster than IaaS – from 2020-2026, PaaS is expected to grow 300% while IaaS spending will increase by 50%.

Let’s explore the reasons why PaaS is expanding much faster than IaaS: 

• Faster: PaaS helps you deliver customer usable features way faster and shortens the Time to Market for digital services. 
• Simpler: it is all taken care of! With PaaS you can take away all the complexity, not only of managing networks, data centers, but also systems, monitoring, applications, etc… 
• Scale: it scales to nearly any extent and runs on a global scale in minutes. 

Is it all monitored? Some monitoring is performed by the CSPs and some monitoring tools are provided out of the box: Azure Monitor and AWS X-Ray. 

PaaS is accompanied by a series of performance management challenges

While PaaS brings a lot of simplification for developers, it also brings its own new performance management challenges.

Below we highlight some off the challenges:

• Monitoring may be simpler: you are able to get some monitoring capabilities from your PaaS provider out of the box. Nevertheless, it is only as good as the coverage it offers, and you need to have practices that align with those capabilities (for example, implementing code tracing for X-Ray).
• PaaS may be used for some of your application chain, but not all of it: in some cases, you may run all of your applications in a given PaaS environment, but in most cases, your application chain will go beyond different PaaS platforms, CDNs, SaaS workloads, links to your legacy systems, or 3rd party applications. Monitoring your end-to-end applications requires the ability to cover not only a given PaaS environment but the overall chain as a whole and the end user experience it provides.
• Old performance methods simply do not work: all the performance measurement methods that require access to the system (e.g. APM) or the network (e.g. NPMD) may be obsolete in these environments. Many performance monitoring and management methods are also not suitable for dynamic environments (e.g. synthetic testing with manual configuration).

What’s the performance management recipe for modern applications?

Depending on what matters for your team and the overall design of your application, you will need to combine monitoring capabilities with different performance monitoring methods like:  

• Synthetic testing: you need to figure out what to test, from where, and how in order to maintain the scenarios you are testing
• Network traffic analysis: for this type of monitoring, you need to have access to the place you want to capture traffic from under your immediate control, as well as a way to work around encryption
• Endpoint instrumentation: either on the server or the client, depending on what works best for your environment

Of course it is easy to pop things in a list, but we assume that your equation will be harder: instrumentation has to be designed around how you use PaaS and will have to either be partial or combined with multiple performance data sources. 

Thinking about migrating your apps to new PaaS offerings? Check out the 3 M’s for cloud migration performance.