It's an obvious point but worth stating: the performance of any customer web port and associated hosted applications is crucial to retaining and expanding the customer base. However, most providers don’t actually know the current performance being experienced by their customers. The service is typically distributed across multiple data centers and numerous physical and virtual servers. As well, there are front-end servers, middle layer processes and back-end services such as databases and storage systems. Designing a system to manage and report customer experience for these environments can seem quite daunting. Typically people investigate multiple solutions that provide instrumentation for each tier of the service, agents for every server and process, and potentially performance reporting for the database and storage layers.
But there is a very simple initial approach to performance monitoring for these environments that most people tend to overlook. By instrumenting just the network access links at the front door of each data center, you can obtain a ton of visibility into the performance of the service from the customer’s perspective. And this approach offers three immediate advantages:
Isolate performance issues between external factors and internal data center issues: Instrumenting at the front door of the data center enables reporting of performance across the client access network verses performance within your data center infrastructure. This is a crucial first step when clients are complaining about performance. If you can isolate the problem to the client access layer, you can focus your resources on resolving that issue rather than spending time looking for problems that don’t exist within your data center.
Report user experience: A passive capture of all user data on the network coming into your data center is invaluable. This data enables real-time reporting of transaction response times for every user request into the service and every corresponding response. This is not simulated traffic. It is actual user production traffic, so the transaction response measurement is reporting real user experience. See our previous blog post "The Problem with APM – Part 1" for how instrumenting the network can report user experience.
Simple and Cost Effective: This approach is much simpler than trying to instrument different servers and services within the data center. That type of instrumentation can involve multiple different types of agents distributed across potentially hundreds or thousands of instances. In contrast, instrumenting the access links typically only requires a single appliance. A single appliance in each data center can report performance for all clients and all servers, and provide one central view of performance across all data centers.
So in summary, simply instrumenting the network access links at the front door of your data center is a simple yet powerful first step in managing and reporting the performance of your web hosted application. In subsequent posts, we will examine the benefits of using this approach to probe deeper into the data center tiers, revealing the performance of storage, database, and authorization services.
Click here to learn more about the Corvil solution for monitoring web application performance.