Holiday Rush - Are your servers ready?
Is your Magento site ready to take the load surge during The Holiday Rush?
The holiday season is upon us - the marketing plans are all in place . ad campaigns, discounts, product mix, etc. Now, you expect a surge in traffic to your eCommerce / Magento site. Are you prepared?
You do not even know what your customers are experiencing!
As a business, you are concerned about the orders or the rate of orders you get. On a well advertised sale, you will get a your top-line number, but would you have an idea of the site speed experience users got on your site? If you think yes, think again.
As we show you not all users see the same response from the site. Do you know that it is possible your site does not scale linearly? Do you know if your hosting was designed to take a surge in traffic? For example, on your magento site, if the hits and visitors increase, the site will keep slowing down.
Slow site = lost sales ... or a lost customer?You are aware that slow site means lost sales. But, did you know that
- The measurements you make on a normal or low traffic time of speeds is different from the results users see when they actually try to purchase.
- Magento will slow down as hits increase . the site will not immediately crash. Visitors will see internal server errors and buttons will not respond..
- Each user sees a different response . with a huge gap between fastest and slowest.
- You will not even know why sales are falling . GA will report increased visitors to the site, just reduced conversions..
What does a slowdown look like?Slowdown is not the same for all visitors. So, it takes a study to understand slowdown. Some examples :
Example 1 : Study of slowdown when simultaneous users increase
With 300 simultaneous users trying to add 1, 2, 3 or 4 products to a cart, here is how the performance degraded in a private study.
- Each user saw a different response
- Variation of 3x from fastest to slowest - so variance is more important not average
- On heavy load, speed degrades even with small actions . like number of products added to cart
- You do not have these stats from any source - you are blind-sighted to what is happening
Example 2 : The big Flipkart problem : Some users find buttons like address save do not work on checkoutA typical site uses "Ajax" for UI . the technology that avoids refresh of screen. Typically used in the checkout process to add address, select shipping options, etc.
- Ajax "internal server errors" are not reported to the visitor - since your developers never imagined it would happen.
- Visitor simply sees a click did nothing
- You will not even know how often it happens - since you are not even watching
- Leads to bad user experience and frustation
- A user takes the screenshot and posts it on social media causing the PR cost to go up
Our service offeringWe have been working on optimizing speed of eCommerce sites for more than a year. Our service include test, measure, reconfigure and monitor.
- Study your preparedness to expected high volume of traffic and give you a report to mitigate the shortcomings
- Execute on our suggested plan and setup live monitoring for your site
- Report on actual performance on a daily / hourly / minute basis
How do we test?We can test how your site would behave in a slowdown. Let us take the mystery of how this test is done.
- Decide on a flow to test
- Decide on the rate of hits . the traffic pattern to simulate
- Simulate unique users from a test generation VM on the cloud
- Test app server to be as close as possible to the production server
- Change rate of test generation to either simulate traffic patterns or determine the effect of growing traffic on the application server
How do we measure?Measurement is extremely important - you cannot control what you do not measure. The measurements on the test site should be similar to what you can do on the live site.
- Server side change in logging parameters to record time for each access
- Use server side monitoring using munin
- Analyze log files using our tools to give you information about each type of hit and rate of response slowdown
- Correlate with munin to analyze the slow down
How do we reconfigure?
- Bottleneck analysis drives what we can do
- We know not all paths are of equal priority . ajax calls for example should be given higher priority using our URL based distribution philosophy
- Appropriate caching mechanisms
- Mysql caching reconfiguration
- Study preparedness : ₹ 80,000 + ₹ 5,000 per server
- Execution of our suggested plan, setup and report live monitoring : ₹ 10,000 per server per month.
- Checkout our monthly plans