That looks like it will work fine. When you add a new trade type (limit, watch, short, fail, etc), you'll need to update your subscribers for each type. On Apr 26, 2013 8:58 AM, "Nishant Mittal" <[email protected]> wrote:
> Eric, just had my coffee and have a better solution... > trades go with +AAPL > quotes go with -AAPL > > then > > a) everything: subscribe to "+" and "-" > b) everything for a symbol (AAPL): subscribe to "+AAPL" and "-AAPL" > c) all trades for all symbols: subscribe to "+" > d) all quotes for all symbols: subscribe to "-" > e) trades only for a symbol AAPL: subscribe to "+AAPL" > > right? > > thanks > Nishant > > > On Fri, Apr 26, 2013 at 9:23 AM, Nishant Mittal <[email protected]> wrote: > >> understood.. I appreciate your help with this. >> >> thanks >> Nishant >> >> >> On Fri, Apr 26, 2013 at 9:21 AM, Eric Hill <[email protected]> wrote: >> >>> Sorry, you are correct about a). I missed the asterisk. >>> >>> While the publisher can publish 2 copies, I like to separate concerns of >>> the code. In my opinion, the publisher should only publish what actually >>> happened. The duplication for the purposes of receiving the messages is a >>> different feature and should not taint the publisher code. Hence the >>> broker. You can have the broker live on the same machine if you'd like. >>> It's just a way to keep the code organized. >>> >>> >>> >>> On Fri, Apr 26, 2013 at 8:18 AM, Nishant Mittal <[email protected]>wrote: >>> >>>> Eric, then for a) i will have to subscribe to "*" otherwise i'll get >>>> all msgs twice. also why do I need a broker in the middle.. the publisher >>>> itself can publish 2 copies.. one with T (or Q) and the other with *..rt? >>>> >>>> thanks >>>> Nishant >>>> >>>> >>>> On Fri, Apr 26, 2013 at 9:10 AM, Eric Hill <[email protected]> wrote: >>>> >>>>> Place a broker between your publisher and subscribers. Have your >>>>> broker subscribe to every transaction your publishers send. Publishers >>>>> send every trade prefixed with T (TAAPL500.00) and every quote prefixed >>>>> with Q (QAAPL). Your broker re-publishes those, and also publishes the >>>>> message replacing the type (first character) with an asterisk. >>>>> TAAPL500.00 >>>>> -> *AAPL500.00 and QAAPL -> *AAPL. >>>>> >>>>> Then: >>>>> >>>>> a) everything - Subscribe to "" >>>>> b) everything for a symbol (AAPL) - Subscribe to "*AAPL" >>>>> c) all trades for all symbols - Subscribe to "T" >>>>> d) all quotes for all symbols - Subscribe to "Q" >>>>> e) trades only for a symbol AAPL - Subscribe to "TAAPL" >>>>> >>>>> >>>>> >>>>> On Fri, Apr 26, 2013 at 7:45 AM, Nishant Mittal <[email protected]>wrote: >>>>> >>>>>> Pieter thanks for the answers. on #3.. >>>>>> I am publishing market data (trades & quotes) for all symbols. the >>>>>> requirement is that a subscriber can subscribe to say >>>>>> a) everything >>>>>> b) everything for a symbol (AAPL) >>>>>> c) all trades for all symbols >>>>>> d) all quotes for all symbols >>>>>> e) trades only for a symbol AAPL >>>>>> >>>>>> the way I was hoping to "solve" this was by adding 2 extra frames to >>>>>> my payload frame. trades will be marked as "+", quotes with a "-".. so >>>>>> for >>>>>> example when publishing a trade for AAPL will send out... >>>>>> frame 1: +AAPL >>>>>> frame 2: AAPL >>>>>> frame 3: <pay load> >>>>>> >>>>>> similarily when publishing a quote ... >>>>>> frame 1: -AAPL >>>>>> frame 2: AAPL >>>>>> frame 3: <pay load> >>>>>> >>>>>> now for the above use cases the subscriber could subscribe to... >>>>>> a) "" >>>>>> b) "AAPL" >>>>>> c) "+" >>>>>> d) "-" >>>>>> e) "+AAPL" >>>>>> >>>>>> but if the filtering only looks at first frame this clearly wont >>>>>> work.. how do you suggest I achieve this? >>>>>> >>>>>> thanks >>>>>> Nishant >>>>>> >>>>>> >>>>>> On Fri, Apr 26, 2013 at 8:34 AM, Pieter Hintjens <[email protected]>wrote: >>>>>> >>>>>>> On Fri, Apr 26, 2013 at 2:19 PM, Nishant Mittal <[email protected]> >>>>>>> wrote: >>>>>>> > In a pub sub scenario in 3.2.2 >>>>>>> > 1) does the filtering happen on the Publisher (i think yes) or the >>>>>>> > Subscriber? >>>>>>> >>>>>>> Yes, the publisher filters. >>>>>>> >>>>>>> > 2) does number of subscribers effect performance of the publishers >>>>>>> (I think >>>>>>> > yes, as the pub has to filter/send msgs to all subscribers) >>>>>>> >>>>>>> Not really, your network will saturate before the publisher is >>>>>>> affected. >>>>>>> >>>>>>> > 3) I send out 3 part msgs.. 1st 2 parts are really for filtering >>>>>>> and the 3rd >>>>>>> > part is a binary object.. when I subscribe.. would zeromq try to >>>>>>> match with >>>>>>> > all 3 parts?.. >>>>>>> >>>>>>> No. the match stops at the end of the first frame in the worst case >>>>>>> (and stops when it mismatches a single octet anyhow). >>>>>>> >>>>>>> -Pieter >>>>>>> _______________________________________________ >>>>>>> zeromq-dev mailing list >>>>>>> [email protected] >>>>>>> http://lists.zeromq.org/mailman/listinfo/zeromq-dev >>>>>>> >>>>>> >>>>>> >>>>>> >>>>>> -- >>>>>> *Nishant Mittal* >>>>>> Director, Product Development >>>>>> *Rosenblatt Securities Inc*. >>>>>> 20 Broad Street >>>>>> New York, NY 10005 >>>>>> >>>>>> Direct: 212-607-3159 >>>>>> Mobile: 646-504-2629 >>>>>> >>>>>> _______________________________________________ >>>>>> zeromq-dev mailing list >>>>>> [email protected] >>>>>> http://lists.zeromq.org/mailman/listinfo/zeromq-dev >>>>>> >>>>>> >>>>> >>>>> _______________________________________________ >>>>> zeromq-dev mailing list >>>>> [email protected] >>>>> http://lists.zeromq.org/mailman/listinfo/zeromq-dev >>>>> >>>>> >>>> >>>> >>>> -- >>>> *Nishant Mittal* >>>> Director, Product Development >>>> *Rosenblatt Securities Inc*. >>>> 20 Broad Street >>>> New York, NY 10005 >>>> >>>> Direct: 212-607-3159 >>>> Mobile: 646-504-2629 >>>> >>>> _______________________________________________ >>>> zeromq-dev mailing list >>>> [email protected] >>>> http://lists.zeromq.org/mailman/listinfo/zeromq-dev >>>> >>>> >>> >>> _______________________________________________ >>> zeromq-dev mailing list >>> [email protected] >>> http://lists.zeromq.org/mailman/listinfo/zeromq-dev >>> >>> >> > > > _______________________________________________ > zeromq-dev mailing list > [email protected] > http://lists.zeromq.org/mailman/listinfo/zeromq-dev > >
_______________________________________________ zeromq-dev mailing list [email protected] http://lists.zeromq.org/mailman/listinfo/zeromq-dev
