Difference between "Response Time over Time" and "Latency" over time

[lux.lu...@gmail.com]

With all load testing terminology, certain words mean certain things to certain people. However in the case of JMeter, I'm having trouble making the distinction between these two plugins:

• Response Times Over Time (ttp://www.jmeter-plugins.org/wiki/ResponseTimesOverTime/)
  
• Latencies Over Time (http://www.jmeter-plugins.org/wiki/LatenciesOverTime/)
  

I always think of latency as necessarily including network time, and therefore response time (in my world) is always strictly how long it took the server to process a single request/transaction. Latency, in my world, would therefore include the response time, but also how long it took the network to transfer the data to the browser, so in most cases, latency will be slightly (or largely) above the server's response time.

However, JMeter defines latency as "duration between the end of the request and the beginning of the server response," which to me sounds like response time. 

Could someone be kind enough to lay this up for me in plain English; I wish the JMeter plugin documentation was a bit more robust.

[Andrey Pohilko]

Hi,

Unfortunately, you haven't provided any reference to your opinion.

Here's some:

http://en.wikipedia.org/wiki/Talk:Latency_(engineering) - Martin Fowler sounds impressive.

http://javidjamae.com/2005/04/07/response-time-vs-latency/ - anothe interesting point

http://nirajrules.wordpress.com/2009/09/17/measuring-performance-response-vs-latency-vs-throughput-vs-load-vs-scalability-vs-stress-vs-robustness/ - one more

http://stackoverflow.com/questions/18510846/jmeter-latency-vs-load-timesample-time - just stating JMeter core logic

What do you think after reading this?

пятница, 4 апреля 2014 г., 18:03:30 UTC+4 пользователь lux.lu...@gmail.com написал:

[lux.lu...@gmail.com]

Thanks, and I'm in total agreement with the definitions being provided in the articles you noted (thanks for these):

Latency:

1. The delay incurred in communicating a message (the time the message spends “on the wire”)
2. "We refer to latency as the time a request spends "on the wire" before getting to the software system" -Fowler

However, my question was more around why JMeter appears to alter that definition by stating the following:

Latency

1. The duration between the end of the request and the beginning of the server response.
   

Maybe I'm just interpreting JMeter's definition too literally, but when I read this, specifically the italicized part above, it seems to include server processing time, since the server can't begin to respond until after its performed some processing. Is that putting to fine a point on it?  The stack overflow article seems to acknowledge this perhaps:

Latency:

1. The difference between time when request was sent and time when response has started to be received
   

This seems to include server processing time as well, where as other definitions only define it as time on the wire! Please do not hesitate to tell me if I'm being super dense here, but there does seem to be some conflicting definitions surrounding what the industry perceives latency to be, and how JMeter implements/calculates it.

So for all intents and purposes, in most circles, latency is simply the time the request (or response) sat on the wire from Host A to Host B, but in JMeter world, latency is the duration between the end of the client request to the beginning of the server's response, which will include "classic latency" (i.e. network time) but also server processing time.

If I am totally off base here, please don't hesitate to call me super dense, but there does seem to be a discrepancy is definitions here.

[Andrey Pohilko]

I agree that "time spent on wire" definitions takes its place in the references. My next three points for you:

1. There is some historical tradition for such RT/latency usage in JMeter, most of people are OK with it.
2. JMeter-Plugins.org have to follow core JMeter glossary. Just to be sane in user's eyes.
3. Core JMeter mailing list is the best place to raise this question. Like this one: http://jmeter.512774.n5.nabble.com/Analyzing-results-Load-Time-vs-Latency-td533733.html

Good luck!

--

Andrey

пятница, 4 апреля 2014 г., 20:54:14 UTC+4 пользователь lux.lu...@gmail.com написал:

[lux.lu...@gmail.com]

Thanks very much, Andrey - much appreciated!