My client is wondering if we can tell how much of the 'Network Time' is spent on the client side of the AMD versus the server side. My first thought is that we could infer from the Server and Client RTTs or bandwidths the transmission speed on each side and split up the Network Time that way. Am I off base or missing something? I plan to make a custom metric out of it.
Thanks for your thoughts!
Solved! Go to Solution.
I might be somewhat off the mark but I believe that the RTT is based on the actual handshake which usually takes place at the beginning of a session (there are occasions where this is not true).
So statistically the base for calculation is rather limited.
There is the ACK RTT though that from what I understand is based on continous RTT during the length of the binode conversation.
In strict sense, I think that the "true" network time portions vary independently and is hard to tell in detail. So I'd go for an average. If you need some input or ideas, check Gary Kasiers bloggin and also the manual for DNA around the ALD.
I guess what Ulf means is Dynatrace Network Analyzer documentation on Active Latency Detection. In short DNA is able to detect network delays of captured nodes and apply measurements to sent/receive times. This way you can get pretty accurate times with one-side capture as if it was merge of both client and server captures. ALD periodically sends connection requests to presumably closed port and measures how long does it take a server to refuse the connection (usually immediate TCP RST). In case when ALD is not able to give accurate results (as it may be considered as port scanning), DNA also offers latency detection based on handshake analysis.
I missed that there was a follow up question.
Further to what Tomasz is saying, the ALD lets you plot the actual latency with very high granularity so you don't just get an average xxx ms (well you do that too) but a nice graph that will give most people a surprise as the latency during a session can be very spikey.
I did not work much with NV. However I can say that, although DNA is a component independent from NV, it may be one of the closest to it among all Dynatrace products. Maybe @ulf thornander has some thoughts on this. I am sure @Mike Hicks, known as both DNA and NV enthusiast, can tell you what NV use cases and to what degree can be handled by DNA.