Testing with realistic network conditions | Hacker News Reader

Testing with realistic network conditions

calendar.perfplanet.com

95 pointsinian7 comments

This is an incredibly real problem. As someone who sells high performance networking to mobile apps[1], it's unbelievable how many people will say their load times are "fast enough," but when you dig into how they reached that conclusion it's in fact only on their near-flawless enterprise WiFi with 0% packet loss. They just don't put themselves into the actual shoes of their users.

More, I have yet to find an organization that understands -- or even measures -- their business metrics with respect to network type. There is decent understanding of the notion of performance, clearly, but abhorrent understanding of performance as a _parameter_. The vast majority of people just don't ask basic questions like "how do my users on 3G engage differently than on LTE?".

If anyone has any ideas about how we can educate people about their performance problems I'm all ears!

[1] http://www.caffei.net/ (happy to answer any questions or rant offline, just drop me a line: james at our site address)

I concur 100%, it is important to test under realistic conditions.

I used Xcode's link conditioner tool to test a custom protocol under packet loss and latency.

With Wireshark you can see TCP retransmissions. In my case, TCP introduced too much latency since the guarantees it offers are costly under certain circumstances.

So I moved to UDP and reworked the protocol to have fire-and-forget messages.

Any apps on the store using this? Would be interested to understand more

Does anyone know of a Windows tool that test various proxy setups for apps, like PAC/authenticated proxy/etc?

What kind of flaws does this kind of testing typically uncover?

The testing conditions that are used when developing websites aren't the same as what users experience - especially when they are on a mobile device which has lower bandwidth, power, etc. It is important to ensure that the site is usable from the user's perspective. This kind of testing is usually required for you to see how your site performs when there is higher latency, packet loss, single points of failure (when a domain is unreachable) and so on..

Though these are great tools to help us test, I still strongly recommend using real devices with real network conditions to do tests like these.

Sounds like every media website: Forbes, Vanity Fair, NY Times, on and on. Mobile LTE and their sites are slow, very laggy, like they only care to test with desktop systems and low latency high bandwidth networks. Annoying.

Reply on news.ycombinator.com