Learn how to use Skylight to make your app faster.
This is the App Dashboard, the first thing you’ll see when you log into Skylight. The dashboard gives you a high-level view of how your app is performing and acts as a starting point for digging into the details.
At the top of the App Dashboard, you will see a graph with the "typical" and "problem" response times and requests per minute (RPM) for your app during the currently selected time range. Hover over the graph to see the specific numbers at a given time.
The Response Timeline is useful for keeping an eye on your application and making sure the response times don’t shoot up suddenly, and for detecting spikes in traffic.
Skylight’s deploy tracking beta feature allows you to zero in on performance improvements or regressions caused by a deployment. With the deploy tracking feature enabled, each of your deployments will be marked on the Response Timeline. (If deploys occurred very close to one another, they will be grouped under a single rocket with an indicator of the number of the deploys represented by that rocket.) Learn more here.
Below the timeline is the Endpoints List. This list shows you all of the endpoints in your application that have been used in the currently selected time range; that is, all of the controllers and their actions.
In addition to each endpoint’s name, we display the Response type, the Typical Response and Problem Response times, Popularity, and Agony, as well as "heads up" icons to alert you to potential code mistakes that could cause slowdowns in your app.
Once you’ve figured out where you’d like to focus your performance-tuning efforts, just click on the endpoint name and you’ll be taken to a wonderland of performance information.
You may have noticed that by default, we order endpoints by our patent-pending Agony-Detection Algorithm™. (Just kidding about the patent-pending bit.) We calculate how much agony your endpoint is causing customers by looking at response times and endpoint popularity. Using a combination of these factors, we determine which endpoint is having the most adverse affect on your users.
Of course, we also allow you to sort the list by any of the other columns. Just click on the column name to re-sort. If you ever forget what a column signifies, hover over the column header. We recommend you sort by Agony, though, and start at the top and work your way down.
Skylight separates html, json, and errors into their own endpoints so you get a clear picture of each. Learn more about response types here.
Skylight notifies you of potential code mistakes that may slow down your app.
Repeated SQL Queries: In general, you will get better performance out of your database if you group similar queries together. Skylight identifies endpoints and events that repeatedly make similar SQL queries so that you can group them.
Allocation Hogs: We call out endpoints with abnormally high allocations that could be causing issues for your application. You can then drill into the individual endpoint and see exactly where the allocations occur so that you can optimize them.
With Endpoint Search, you can search for a particular endpoint or group of endpoints by typing all or part of an endpoint name into the field; a type of fuzzy search.
Insider users can also enable Advanced Search (currently in beta), which allows the user to search by filtering based on the Endpoints List column headings.
Our new Grades beta feature gives you perspective on the performance of your endpoints, relative to other Skylight apps, so you can focus on meeting your customers’ expectations. Learn more about Grades here.
The Endpoint View is the heart and soul of Skylight. More than just looking pretty, this page is the end result of distilling thousands of data points into actionable information that you can use to speed up your app.
At the top of the Endpoint View is the Response Time Distribution, showing you the distribution of the response times for this particular endpoint during the currently selected time range. The Endpoint View, by default, shows all of the requests in the selected time range. Often, it’s helpful to focus on slower requests to see exactly why they are so slow. You can filter the endpoint data by selecting just part of the histogram to zero in on problematic requests.
Once you’ve selected a target time range and grouping of response times, check out the Selection Summary. Here, you can toggle between Time or Allocations modes. You can also condense or uncondense the trace shown in the Event Sequence.
The Summary gives you an overview of the currently selected requests, including:
In addition, the Summary shows the types of events that make up the selected requests, broken down in terms of percentage of average response time or percentage of average allocations.
The heart of the Endpoint View is the Event Sequence—also referred to as the “trace”—which shows you exactly where your app is spending time or allocating objects when servicing an endpoint. Each row represents a different event, and they’re color coded:
When looking at an endpoint with the high object allocations heads up, be sure to switch to Allocations Mode. In this mode, the same trace will be re-scaled based on the number of allocations during each event, allowing you to quickly identify potentially problematic events (i.e. the largest bars in your traces), helping you to pinpoint your allocation hogs.
It’s important to understand that the Event Sequence does not represent a single request. It represents many (potentially thousands) of requests all merged into one. Showing single requests can send you on a wild goose chase, because that request may not be representative. Because we aggregate all requests together, if something looks like it’s taking a lot of time in the Event Sequence, that means it was taking enough time in your production environment to be statistically significant.
You can get more information about a particular event by clicking on it to get the Detail Card. In database events, for example, we show the SQL that was executed, so it’s easy to track down exactly what query was slow. Note that the SQL query is aggregated. Any variables are stripped from SQL queries before they are sent to Skylight, allowing us to aggregate similar queries and give you a more holistic view of the performance of the query (not to mention the information security benefits).
Wondering what the light and darker segments mean? If you see a dark segment, that’s “self-time”—time that was spent for that particular event. The total self-time for a row is shown at the end of the row.
Light colored segments represent child events. For example, if your controller’s Ruby code calls out to the database and then does something with that data, the time spent calling out to the database would be represented as a lighter shade of blue. You’ll see that the lighter shaded segments always line up with a child segment that appears below.
In order to minimize the amount of noise in the Event Sequence, Skylight shows it as “condensed” by default. Events with no children that have a very brief total duration are either condensed into a single line or hidden from view. You can uncondense the Event Sequence by unchecking “condense trace” in the Summary.
The Endpoint Hotspots beta feature highlights events that take up more than 10% of the total response time or allocations on an endpoint. If your endpoint is slow, take a look at its hotspots to identify opportunities for performance improvements.
The time explorer lives at the bottom of your app views. For both the App Dashboard and the Endpoint View the data shown is for the range selected in the explorer. Displayed in the graph is the problem response time for either the entire application or for the specific endpoint being viewed.
You can drag the selected range, use the arrows buttons, or choose a predefined range of time. When you update this range, you’ll immediately see the current data in the rest of the page update accordingly.
Our weekly Trends Email distills your performance data into a beautiful, actionable report that helps you catch slowdowns before your customers do.
Track "typical" and "problem" response times with lists of your most improved endpoints your biggest slowdowns. Endpoints that have slowly declined over a 6 week period will be called out as “Frog Boils”.
Sign up for Trends Emails on your account settings page.