Windows Azure is becoming a de facto platform to distribute the applications and services using Microsoft stack. With the platform gaining momentum with tons of new features getting added every 3 weeks, one dimension cannot be missed which is maintaining these apps and services. For IT folks – it’s a total paradigm shift. Laying down infrastructure topology, deployment, formulating disaster recovery strategy etc. are few examples. Here we are going to discuss only monitoring aspect.
Application monitoring can be defined as processing and use of related IT tools to detect, diagnose, remediate and report an application’s performance to ensure that it meets or exceeds end-users’ and businesses’ expectations.
There are two basic methods by which application performance can be assessed.
- Measuring the resources used by the application
- Measuring the response time of applications
The most common aspect from operations perspective is monitoring of all components and services and the ability to troubleshoot if things go wrong. This typically involves monitoring various performance counters, monitoring log files for alerts, monitoring availability of the application and the ability to configure automated alerts for all these monitoring activities.
With Azure deployed applications, applications/services need to be enabled in certain way so that those can be monitored by operations team. For such applications/services health monitoring, operations team can follow following strategies:
Inside out monitoring
If the application and services has logging, tracing and performance counters embedded directly into the code then we can use this strategy. Code reports all the health and performance counters, which Windows Azure Diagnostics monitor can then collect and report.
This also can be achieved using a health dashboard page which is authorized to be viewed only by the operations team. This page or component makes dummy service calls to all the services and components based upon the scheduled frequency and displays the success/failure for the same. It can ping the database (SQL Azure), external endpoints (E.g.: endpoints exposed on Service Bus, Storage etc.) and publish the response on the page.
This means that application itself does not take care of reporting the health and performance and it’s the duty of the external component. In this technique – you need a piece of software (an agent) that gets deployed along with the package which works in a low priority demon and reports the health and performance to a remote server. The agent looks at the deployed code and “reflects” the status dynamically.
My vote always will be with outside-in strategy. This gives you more flexibility and achieves what’s popularly known as separation of concerns. Separating the actual product code from monitoring code. E.g. If Ops team want to change let’s say some perf. counter they do not need to touch the actual code for that.
I came across this awesome tool/service in Windows Azure Store called NewRelic. NewRelic can be used to monitor Web Sites, Web/Worker Roles and VMs. The kind of statistics it produces with a large degree of simplicity is just breathtaking. New Relic is the all-in-one web application performance tool that lets you see performance from the end user experience, through servers, and down to the line of app code. As far as I know, New Relic is the only SaaS app monitoring tool that monitors the entire app using a single product and a single UI. It also gives you this beautiful real-time performance dashboard that can be accessed anytime, anywhere.
As you can see New Relic has multiple plans to choose from. The paid plans offer more detailed diagnostics and data. Visit their web site for more details regarding plans. For the tutorial sake – I am sticking to Slandered (Free) which is far more than sufficient. Go ahead, and provide a name to your New Relic endpoint, which creates the New Relic endpoint.
In Visual Studio go to your Azure Web Site or Web role. Open Nuget Package Manager window and install NewRelicWindowsAzure package. You will prompted for authorization, provide the API key which we got from the above step. After the setup is done, deploy the application/website.
If things are deployed correctly and successfully you see various beautiful graphs like as follows: