David - that’s a correct interpretation, although we’re interested in having a measurement for every tuple, rather than just our aggregate. For our use case we actually require the ability to track any individual tuple end-end through the entire system. We are aware that it’s possible to measure latency on a window basis but this does not provide sufficient granularity. The main purpose of this is to be able to conclusively identify bottlenecks and limitations within the Apex architecture.
On 11/17/15, 4:07 PM, "David Yan" <[email protected]> wrote: >Hi Ilya, > >Just wanted to make sure I understand correctly. When you said latencies >for individual tuples, do you mean how many milliseconds on average a >tuple >stays in the queue in the upstream operator before >DefaultInputPort.process() is called in the current Operator? > >We currently have latency per operator based on the endWindow tuple >dequeueing time. Although this is not the same, it gives you an idea how >much latency an operator imposes on the stream on a per-window basis. >Example: > >$ curl >http://node0.morado.com:9099/ws/v2/applications/application_1447661686850_ >17918/physicalPlan/operators/11 >| python -mjson.tool >{ > ... > "latencyMA": "287", > ... >} > >David > > >On Tue, Nov 17, 2015 at 3:48 PM, Ganelin, Ilya ><[email protected]> >wrote: > >> Specifically, I¹m interested in being able to track latencies for >> individual tuples. For example, as they move through the BufferServer or >> the CircularBuffer. >> >> On 11/17/15, 3:12 PM, "David Yan" <[email protected]> wrote: >> >> >Hi Ilya, >> > >> >The AutoMetric interface is mostly designed for application-specific >> >metrics within an operator -- things like how many frauds are >>detected, or >> >the dollar amount processed. >> > >> >We do have BufferServer bytes and queue tuple count (a.k.a. queue >>size) as >> >system metrics. If you have specific metrics that pertain to >>BufferServer >> >and CircularBuffer that you're interested in, please state them. We >> >should >> >probably add them to system metrics since it's probably useful in >>general. >> > >> >David >> > >> >On Tue, Nov 17, 2015 at 2:58 PM, Pramod Immaneni >><[email protected]> >> >wrote: >> > >> >> Some of them are available in system stats like port queue size and >>are >> >> exposed via REST and to a stats listener that you can register via >>the >> >>API. >> >> Some of the others you mentioned can be added. >> >> >> >> On Tue, Nov 17, 2015 at 2:52 PM, Ganelin, Ilya < >> >> [email protected]> >> >> wrote: >> >> >> >> > Hi, all presently Apex supports surfacing metrics via the >> >>AutoMetrics >> >> > interface, documented here: >> >> > http://docs.datatorrent.com/autometrics/ >> >> > >> >> > However, this only provides utility to monitor operators. I would >> >>like to >> >> > provide a more general latency suite - fFor example, I would like >>to >> >> > instrument metrics around the BufferServer or the CircularQueue. >>Not >> >> > everything in Apex is an operator, and there are other components >>of >> >>the >> >> > system around which we may want explicit measurements or metrics. >>Is >> >>this >> >> > something that could exist? I know other projects like Spark have >> >>similar >> >> > efforts, although they are equally underdeveloped :) >> >> > ________________________________________________________ >> >> > >> >> > The information contained in this e-mail is confidential and/or >> >> > proprietary to Capital One and/or its affiliates and may only be >>used >> >> > solely in performance of work or services for Capital One. The >> >> information >> >> > transmitted herewith is intended only for use by the individual or >> >>entity >> >> > to which it is addressed. If the reader of this message is not the >> >> intended >> >> > recipient, you are hereby notified that any review, retransmission, >> >> > dissemination, distribution, copying or other use of, or taking of >>any >> >> > action in reliance upon this information is strictly prohibited. If >> >>you >> >> > have received this communication in error, please contact the >>sender >> >>and >> >> > delete the material from your computer. >> >> > >> >> >> >> ________________________________________________________ >> >> The information contained in this e-mail is confidential and/or >> proprietary to Capital One and/or its affiliates and may only be used >> solely in performance of work or services for Capital One. The >>information >> transmitted herewith is intended only for use by the individual or >>entity >> to which it is addressed. If the reader of this message is not the >>intended >> recipient, you are hereby notified that any review, retransmission, >> dissemination, distribution, copying or other use of, or taking of any >> action in reliance upon this information is strictly prohibited. If you >> have received this communication in error, please contact the sender and >> delete the material from your computer. >> >> ________________________________________________________ The information contained in this e-mail is confidential and/or proprietary to Capital One and/or its affiliates and may only be used solely in performance of work or services for Capital One. The information transmitted herewith is intended only for use by the individual or entity to which it is addressed. If the reader of this message is not the intended recipient, you are hereby notified that any review, retransmission, dissemination, distribution, copying or other use of, or taking of any action in reliance upon this information is strictly prohibited. If you have received this communication in error, please contact the sender and delete the material from your computer.
