From charlieyounghusband at gmail.com Thu Oct 13 21:41:25 2016 From: charlieyounghusband at gmail.com (Charlie Younghusband) Date: Thu, 13 Oct 2016 17:41:25 -0400 Subject: RPTM object types Message-ID: Hi all, I was interested in doing a test which focused on the response time per object. Following a test, I run the reporter tool but on the rptm.html page, it doesn't report the min/mean/max response time (msec) per object. It just says'n/a' with a footnote, and the footnote page just states that no measurement was collected or stored at run-time. The sizes part is reported. Does anyone know what I need to do in order to have this information collected and reported? There is a note that "Some statistics may not be available because either no objects of the corresponding class were seen during the test or no facilities to collect the stats exist for the class." I'm hoping that the answer is that polygraph is just not capable of storing the metadata associated with this. All of the other parts of the table are filled in, including the byte latency and server byte latency, times for hits and misses etc. FYI the DUT prioritizes (at the TCP/IP QoS level) data flows based on MIME type, we have established that that can improve user experience. I'm using polygraph as a check to see what happens when we put the system under duress. Thanks! Charlie Here is a simplified config .pg: #include "contents-httpqos.pg" Phase phPlat = { name = "plateau"; goal.duration = 2min; }; Server S = { kind = "PolyServerA"; contents = [ cntImage: 33%, cntHTML: 33%, cntOther ]; http_versions = [ "1.1" ]; // always use a 1.1 capable server addresses = ['127.0.0.1:801' ]; // where to create these server agents }; Robot R = { kind = "PolyClientB"; pop_model = { pop_distr = popUnif(); }; req_rate = 1000/sec; origins = S.addresses; // where the origin servers are addresses = ['10.1.0.9' ]; // where these robot agents will be created }; schedule(phPlat); use(S, R); From rousskov at measurement-factory.com Thu Oct 13 23:51:25 2016 From: rousskov at measurement-factory.com (Alex Rousskov) Date: Thu, 13 Oct 2016 17:51:25 -0600 Subject: RPTM object types In-Reply-To: References: Message-ID: On 10/13/2016 03:41 PM, Charlie Younghusband wrote: > I was interested in doing a test which focused on the response time per > object. By "per object", you mean "per content type", right? Response times for content types should be available since v4.5.0. I assume you are using something newer. > Following a test, I run the reporter tool but on the rptm.html > page, it doesn't report the min/mean/max response time (msec) per > object. It just says 'n/a' with a footnote, and the footnote page just > states that no measurement was collected or stored at run-time. The > sizes part is reported. Does anyone know what I need to do in order to > have this information collected and reported? Perhaps Polygraph was not able to determine the content type of the response or could not match the content type back to PGL? Are you filtering any HTTP response headers out? > I'm hoping > that the answer is that polygraph is just not capable of storing the > metadata associated with this. I just checked a random v4.9.0 report, and I do see min/mean/max response time measurements in the Reply object table (the first table on the rptm.html (response times) page. If nobody has a better idea, I may have to ask you for console and binary logs to investigate. Alex. From charlieyounghusband at gmail.com Fri Oct 14 21:26:16 2016 From: charlieyounghusband at gmail.com (Charlie Younghusband) Date: Fri, 14 Oct 2016 17:26:16 -0400 Subject: RPTM object types In-Reply-To: References: Message-ID: I am somewhat embarrassed to admit that the test system I was using polygraph 4.3.2, I realize now. Running on a 4.9 works great. Thanks for the tip on the 4.5 or this might have taken me longer to figure out! Charlie On 2016-10-13 7:51 PM, Alex Rousskov wrote: > On 10/13/2016 03:41 PM, Charlie Younghusband wrote: > >> I was interested in doing a test which focused on the response time per >> object. > By "per object", you mean "per content type", right? Response times for > content types should be available since v4.5.0. I assume you are using > something newer. > > >> Following a test, I run the reporter tool but on the rptm.html >> page, it doesn't report the min/mean/max response time (msec) per >> object. It just says 'n/a' with a footnote, and the footnote page just >> states that no measurement was collected or stored at run-time. The >> sizes part is reported. Does anyone know what I need to do in order to >> have this information collected and reported? > Perhaps Polygraph was not able to determine the content type of the > response or could not match the content type back to PGL? Are you > filtering any HTTP response headers out? > > >> I'm hoping >> that the answer is that polygraph is just not capable of storing the >> metadata associated with this. > I just checked a random v4.9.0 report, and I do see min/mean/max > response time measurements in the Reply object table (the first table on > the rptm.html (response times) page. > > If nobody has a better idea, I may have to ask you for console and > binary logs to investigate. > > Alex. >