SteelCentral Web Analyzer Review: What We Found Can Make Your Site Faster
We are the editorial team for the Riverbed Connections newsletter. When we heard that Riverbed Technology was offering a 30-day free trial for monitoring web application performance, we thought: “What do we really know about our website’s performance?” And, within minutes of installing Riverbed SteelCentral™ Web Analyzer, we were poring over performance data on our way to an answer.
SteelCentral Web Analyzer lets you monitor web application behavior; helps you understand performance variations by browser, device, and platform; and gives you end-to-end insight into potential problems before they pop up.
Over the next two weeks we gathered enough information to take significant action. What we discovered can help speed up your site too.
10 minutes to set up the 30-day trial
Riverbed offers a 30-day free trial of SteelCentral Web Analyzer, which is ample time to gather and analyze actionable performance intelligence. Setup took all of 10 minutes. In a set-up process similar to that of Google Analytics, we signed up for the trial, copied the code snippet, and pasted it into the page header before the end of the HTML <head> tag. There was zero footprint to deploy, and data immediately started populating the dashboard.
Response times sliced and diced
The sheer depth of useful information and the ability to analyze that information in varying ways is what makes SteelCentral Web Analyzer so valuable. For us, the key factor was response time, which measures how long it takes a webpage to load. On our site the times varied widely from .02 seconds to a disappointing 85 seconds for one outlier. Our summary of delays split the total response time into 27% for Time to First Byte and 73% for Download and Render. The fat mean was 7.9 seconds — which is OK, but certainly in need of improvement.
Response time is subdivided into Time to First Byte and Download and Render, both of which are broken down further.
Time to First Byte is broken down into four components:
- Redirect: The time it takes for any HTTP redirects (301 or 302) to complete.
- DNS: The time it takes to complete the domain name lookup of the server where the request will be sent.
- Connection Setup: The time it takes to establish a TCP connection between the browser and the server.
- First Request to First Byte: The time it takes the server to process the request. This time also includes the round-trip network latency.
Download and Render is broken down into two components:
- Initial Layout: The time from the first byte reaching the browser until the page is made visible to the user. It includes the time it takes to download enough data to lay out the page. (This is often referred to as DOM Visible or DOMContentLoaded.)
The geographic puzzle
In our analysis, we noticed that a few visitors from India and Indonesia experienced slow response times — the aforementioned 85 seconds was the worst case. The map view quickly offered more insight, showing much of Asia, Canada, and Mexico in need of attention. Europe was also slower than expected.
Other views give more detail to complete the picture:
- By page: Is the page consistent or varying by user and location?
- By visitor: Do individual visitors’ experiences vary or are they consistent? We noticed that category and home pages were slower than article pages.
Next, we got a good, comprehensive sense of performance overall courtesy of the scatter plot graph. This showed that a few slow response times were outliers. The big mass of visitors saw response times that were less than 8 seconds.
Analysis and actions
Our analysis showed results in three areas:
- Initial layout needed improvement
- Redirects from our email newsletter took around three seconds
- Geographic disparities need addressing
The information we gathered in SteelCentral Web Analyzer led to further investigation with other tools, including Google Developers PageSpeed Insights, which produced some sound advice. We also worked in Pingdom.com, which gave us a waterfall view of individual file load times and let us check DNS health.
Armed with inside information from SteelCentral Web Analyzer, we then approached our hosting provider for their ideas. Working together we came up with a plan that included adjustments to the caching setup, setting up a content delivery network (CDN), and other tweaks.
These actions immediately shaved 1.8 seconds off the average response time (from 7.9 to 6.1 seconds). And the mix of response times responded too, with Download and Render now at 62% of the total, which is down from 73% previously. That’s all thanks to the detailed information we received from SteelCentral Web Analyzer.