Several our customers recently had the requirement to record and compare the end user performance of Oracle Fusion Cloud in different scenarios. For most use cases the customers were successful with using mtr - the command line based network latency approach outlined here: https://www.ateam-oracle.com/a-simple-guide-to-verify-fusion-saas-network-latency. However, for some customers this approach didn’t deliver the results they were looking for as they were focusing on the end user experience - specifically when trying to compare performance between FastConnect, VPN and the Public Internet.
This post will outline a way to record and compare real end user access times to the Oracle Fusion Cloud using HAR files based on Google Chrome Developer Tools and how to get these into excel.
In order to achieve meaningful results, you should validate how you are connected to the Oracle Fusion Cloud. The most common connection options are shown in the diagram below.
Keep in mind that Akamai can have a significant influence on the caching behavior and therefore on the performance results. If you are not sure if you have Akamai enabled in your environment – it is enabled by default – you can have a look at the DNS record that is being used to direct your access as show below. If akamaiedge.com is showing up as shown in the screenshot, you will most likely have Akamai CDN enabled. Please log a Service Request if you are unsure.
Akamai CDN can be disabled by logging a Service Request. It can be disabled for a single environment, so for example only your xxxx-test environment. Please note if you are comparing different environments, it’s strongly recommended to validate with Support that the sizing for these are identical to provide best results.
There are several pitfalls to avoid when comparing performance results. First stop should be to review the Support Note “Best Practices for Browser Settings and Performance on Fusion Applications (Doc ID 1385107.1)” https://support.oracle.com/knowledge/Oracle%20Cloud/1385107_1.html It’s strongly suggested to follow all recommendations, but I would like to highlight a few things as they have shown up as issues for me in the past.
It is always recommended to make sure that the workstation you are testing from is in line with the system requirements that can be accessed here: https://www.oracle.com/system-requirements/
All other tabs and windows should be closed, and you should consider incognito mode to make sure no old sessions are influencing what you are trying to measure.
Be aware of the First Time Access Issue – this applies on two levels – the first level is your browser cache. Clearing your cache will increase the load time for the first time the page is retrieved but it is necessary to have data for a proper comparison when running these tests at different times and different environments. The second level is the first-time access on the server side – this can happen especially after a maintenance on the weekend or similar. Things might just take a little longer the first time they get requested after maintenance. Keep this in mind when testing on the weekend or first thing Monday morning.
We frequently see a multitude of browser plugins checking the content that is being access – which might be necessary and safe but can impact the load times significantly. It would be best to disable all the browser plugins where possible.
Also make sure that there is no scheduled maintenance on your workstation like backups, scheduled virus scans or updates – these can have a heavy impact on your test results.
The actual execution of the performance test is relatively trivial. Follow the best practices above and open a new Incognito Tab. Hit Control, Shift and I (Ctrl+Shift+i) at the same time to launch the Chrome Developer Tools. Make sure to navigate to the network tab and ensure that a new HAR file is being recorded (Ctrl+e) and all previous data is cleared. Then simply navigate to the page you want to measure and note the Load time at the bottom of the page. Check the arrows below for orientation.
Keep in mind you will need to reauthenticate if you are using a fresh session. So, you might want to exclude the initial navigation from your comparison. I suggest repeating the steps several times to get a good average of the load times that are meaningful.
Next switch to the next connection method, for example connect the VPN and simply repeat the steps to retrieve the Load time for that connection. Repeat as required across a few days to get more meaningful data.
In most cases the Load time alone will be enough to satisfy the performance measuring requirements – note them down on different times of the day and across different environments and graph them out in your preferred spreadsheet. If you would like to compare and visualize the results in more detail you will need to load the full performance data into your spreadsheet. To achieve that you can export the data in HTTP Archive Files, which are simple JSON files. Hit the button highlighted by the red arrow below to save the HAR file.
You can convert these HAR files to spreadsheet friendly CSV files using the command line as shown here: https://community.akamai.com/customers/s/article/HOWTO-generate-a-simple-Excel-report-based-on-your-web-page-HAR?language=en_US or simply look for an online converter from HAR to CSV like https://hintdesk.github.io/networkhartocsv/input
You can then open the files and select the data you want to analyse. For example, if a particular part of the page is taking a long time to render on a particular day.
Chrome Developer Tools and the equivalents on other browsers (e.g. Firefox Developer Tools) offer a simple solution to measure performance from an end user point of view. The procedure outlined in the blog will give you a solid tool for your performance analysis.
A Simple Guide to Verify Fusion SaaS Network Latency