|
From: Ghislain A. <ghi...@gm...> - 2017-01-23 17:09:36
|
Igor, Ummmm... I think your syntax does not work, at least on Windows. See the error here https://gist.github.com/gatemezing/4c7da5fb727672a1c8de2d0a8eca441e However, the one I sent you actually works fine (up to now) . +1 to the suggestion of integrating 'warm-up' phase. I will definitively include one of the options suggested, at least the first option. Thanks again. Ghislain El lun., 23 ene. 2017 a las 17:17, Igor Kim (<igo...@ms...>) escribió: > Ghislain, > > AFAIK, --data-urlencode could not read data from file in the form > you've written. Try following syntax > > curl -X POST http://localhost:9999/blazegraph/namespace/<mykb>/sparql > --data-urlencode "query=`cat myQuery.sparql`" --data-urlencode > 'format=json' -o myQuery.json > > Testing pattern usually should follow expected workload. For standard > application, server does not get restarted often, so it is 'warmed up' > most of the time. > During warm-up phase query execution times might differ from main phase, > so you should probably add 'warm-up cycle', which is basically execution > of queries mix which follow same pattern as main application issues. > Running 100-200 queries (or applying workload for 5-10 minutes) would > warm-up server. After that main query mix might be executed. Also note, > that queries and their order, which were executed during warm-up and main > stage is important and might reflect in different results. So there are 2 > options: first is to establish fixed mixture of queries for performance > testing, and the second option is to apply a workload for a longer time to > get results averaged over time, so particular fluctuations would not impact > performance measurement. But for the second option you need to make > queries varying, to reduce impact from caching. > > > On Mon, Jan 23, 2017 at 4:32 PM, Ghislain ATEMEZING < > ghi...@gm...> wrote: > > Igor, thanks again for the explanations. > Since I have a lot of queries to do, I'll prefer the cURL option instead. > I will make an average time of 5 queries, using json header in the POST > query. > > Here is a sample cURL query I am sending: > curl -i -X POST http://localhost:9999/blazegraph/namespace/<mykb>/sparql > --data-urlencode 'qu...@my...' --data-urlencode analytic=true > -H 'Accept: application/json' > myQuery.json > > Would that makes sense? > > Best, > > El lun., 23 ene. 2017 a las 10:24, Igor Kim (<igo...@ms...>) > escribió: > > Ghislain, > > Basically CUrl and Workbench do almost the same thing - issue POST query > to the server, which should analyze and run it then stream the result back > to client. > But upon receiving, curl just stores a stream into a file or logs to > console, but Workbench has to parse JSON results and show it in a web page, > which takes additional time. > Note, that JSON is native format for JavaScript code of the Workbench > (much easier to parse than XML for example), so using other formats would > not improve execution time. > > The difference will be more obvious for large responses (many rows and > large literals). > > You could use developers console in your browser to see execution times > (including transport times, but CUrl will count that time as well). > > See attached screenshot, the query has taken 52ms while waiting for a > response start (prepare a query and start its evaluation), > additional 78ms to get whole result (iterate through the results on server > and stream it through network using JSON format), > but the Workbench have taken additional 322ms to parse and prepare results > for the results table on the web page. > > Also, after execution a query on the Workbench, server's file system > caches index data for a time, so repeating requests to the same query might > be much faster for subsequent calls using CUrl. > > As a workaround you might apply 'limit 50' to evaluate less results (only > first page), that will reduce overhead time for the Workbench. > > > [image: Inline image 1] > > On Mon, Jan 23, 2017 at 1:15 PM, Ghislain ATEMEZING < > ghi...@gm...> wrote: > > Hi Igor, > Thanks for your answer and sorry if O was not clear enough in my previous > post. Basically I thought I can get also the time execution of a query with > the ?explain parameter. > When comparing the response time of a CUrl in localhost w.r.t the answer > on the workbench, I notice it's sometimes 10x higher. Is there a way to get > almost the same result somehow? > > Best, > Ghislain > > Sent from my iPhone, may include typos > > Le 23 janv. 2017 à 04:51, Igor Kim <igo...@ms...> a écrit : > > Ghislain, > > ?explain parameter is a rough equivalent of SQL analyze, so there is no > time of response returned, just a query execution plan explanation. > Could you expand a little bit on what are you going to use that response > time? There is a workaround with using UNION or BIND in the query, which > could provide a timestamp for the results: > > for the whole result: > > select * from { > { > ?s ....... # statement pattern and filters > } > UNION > { > BIND (now() as ?timestamp) . > } > } > > or for each of the rows: > > select * from { > ?s ....... # statement pattern and filters > BIND (now() as ?timestamp) . > } > > > > > On Fri, Jan 20, 2017 at 10:07 PM, Ghislain ATEMEZING < > ghi...@gm...> wrote: > > Many thanks Brad! > I was looking at the Wiki to see if I can use a parameter in a cUrl > request to retrieve the time response of a query. I've seen the ?explain > parameter, but not sure to find the that value in the .html output. > Is there a way to retrieve that value in a REST-ish manner? > > TIA, > Ghislain > > El vie., 20 ene. 2017 a las 17:14, Brad Bebee (<be...@sy...>) > escribió: > > Ghislain, > > Yes, you may use the list for those topics. In general, before posting, we > recommend checking out the Wiki. > > https://wiki.blazegraph.com/wiki/index.php/Main_Page > > Thanks, --Brad > > On Fri, Jan 20, 2017 at 1:38 AM, Ghislain ATEMEZING < > ghi...@gm...> wrote: > > Hi, > [Sorry if I am in the wrong list.] > I am new to blazegraph and I wanted to know if this list also helps for > issues related to the use of the product, setting up issues and so on. > > TIA > Best, > Ghislain > -- > ------- > "Love all, trust a few, do wrong to none" (W. Shakespeare) > Web: http://atemezing.org > > > ------------------------------------------------------------------------------ > Check out the vibrant tech community on one of the world's most > engaging tech sites, SlashDot.org! http://sdm.link/slashdot > _______________________________________________ > Bigdata-developers mailing list > Big...@li... > https://lists.sourceforge.net/lists/listinfo/bigdata-developers > > > > > -- > _______________ > Brad Bebee > CEO, Managing Partner > SYSTAP, LLC > e: be...@sy... > m: 202.642.7961 > f: 571.367.5000 > > CONFIDENTIALITY NOTICE: This email and its contents and attachments are > for the sole use of the intended recipient(s) and are confidential or > proprietary to SYSTAP, LLC. Any unauthorized review, use, disclosure, > dissemination or copying of this email or its contents or attachments is > prohibited. If you have received this communication in error, please notify > the sender by reply email and permanently delete all copies of the email > and its contents and attachments. > > -- > ------- > "Love all, trust a few, do wrong to none" (W. Shakespeare) > Web: http://atemezing.org > > > ------------------------------------------------------------------------------ > Check out the vibrant tech community on one of the world's most > engaging tech sites, SlashDot.org! http://sdm.link/slashdot > _______________________________________________ > Bigdata-developers mailing list > Big...@li... > https://lists.sourceforge.net/lists/listinfo/bigdata-developers > > > > > -- > *Igor Kim* | Team Leader / Backend | Maginfo, Ltd > Mobile: +7-912-402-4622 > > > > > -- > *Igor Kim* | Team Leader / Backend | Maginfo, Ltd > Mobile: +7-912-402-4622 > > -- > ------- > "Love all, trust a few, do wrong to none" (W. Shakespeare) > Web: http://atemezing.org > > > > > -- > *Igor Kim* | Team Leader / Backend | Maginfo, Ltd > Mobile: +7-912-402-4622 > -- ------- "Love all, trust a few, do wrong to none" (W. Shakespeare) Web: http://atemezing.org |