Hi, +1 for using Common math library, as we already using it.
- Currently we have calculating numbers per minute, so it is not per specified number of events, but for a specified time. - IMO we can find the polynomial in each minute, considering 10 minutes of past data, we can change this if required. - Normally, health stat events are received in each 10 seconds. - Therefore we will have 60 ( = 6 * 10) events per 10 minutes. On Wed, Mar 25, 2015 at 9:11 AM, Pranavan Theivendiram < [email protected]> wrote: > Hi Lahiru and Raj, > > I have gone through the curve fitting materials that you have provided. I > have understood the regression part and prediction of degree polynomials. > We can use apache common math libraries to do this. I have a question below. > > How often we are going to find the polynomials? > > For example, > > If we have *3 samples* events during *10 minutes interval*, then we can > produce a second degree polynomial. (We will get constants *a*,*b*, and > *c*). If we are going to use this second degree polynomial for next *60 > minutes *(Suppose in this 60 minutes we are getting *200 samples *events). > As we can see our *extrapolation > <http://en.wikipedia.org/wiki/Extrapolation>* becomes a severe problem in > this particular example. > What do you mean by " our *extrapolation <http://en.wikipedia.org/wiki/Extrapolation>* becomes a severe problem"? Complexity of the calculation of predicted value? Thanks. > I do not know how earlier versions handle this scenario. > > Can you tell that are we going produce new polynomials for *a specified > time* or *a specified number of events*? > > Please clear this issue. > > *T. Pranavan* > *BSc Eng Undergraduate| Department of Computer Science & Engineering > ,University of Moratuwa* > *Intern Software Engineer**| WSO2 Lanka (Pvt) Ltd.* > *Mobile| *0775136836 > > On 24 March 2015 at 09:02, Lahiru Sandaruwan <[email protected]> wrote: > >> Thanks Pranavan, >> >> I'll have a look. >> >> >> On Tue, Mar 24, 2015 at 8:52 AM, Pranavan Theivendiram < >> [email protected]> wrote: >> >>> Hi Lahiru and Raj, >>> >>> This is my draft proposal which has been done up to 50%. I am going to >>> add things about curve fitting as well. Still many things need to be added. >>> Please comment on my initial draft. >>> I am expecting pros and cons regarding my proposal. Please find the link >>> for the initial draft below. >>> >>> >>> https://docs.google.com/a/cse.mrt.ac.lk/document/d/1TzHYI2o9bIdZWj4qAffNmH5FTgeTOV4TuSsxlISlU8c/edit?usp=sharing >>> >>> *T. Pranavan* >>> *BSc Eng Undergraduate| Department of Computer Science & Engineering >>> ,University of Moratuwa* >>> *Intern Software Engineer**| WSO2 Lanka (Pvt) Ltd.* >>> *Mobile| *0775136836 >>> >>> On 22 March 2015 at 23:36, Pranavan Theivendiram < >>> [email protected]> wrote: >>> >>>> Thanks Lahiru for the prompt reply. I have already started working on >>>> the proposal. I will send it before tomorrow. >>>> >>>> *T. Pranavan* >>>> *BSc Eng Undergraduate| Department of Computer Science & Engineering >>>> ,University of Moratuwa* >>>> *Intern Software Engineer**| WSO2 Lanka (Pvt) Ltd.* >>>> *Mobile| *0775136836 >>>> >>>> On 22 March 2015 at 23:31, Lahiru Sandaruwan <[email protected]> wrote: >>>> >>>>> Hi, >>>>> >>>>> Yes, You have found the information correctly and on right track. It >>>>> would be great if we can incorporate CEP 4.0.0 changes as it would >>>>> minimize >>>>> our customized function. >>>>> >>>>> I would like to suggest you starting the drafting of the proposal as >>>>> you have some background now. >>>>> >>>>> Thanks. >>>>> >>>>> On Sun, Mar 22, 2015 at 9:48 PM, Pranavan Theivendiram < >>>>> [email protected]> wrote: >>>>> >>>>>> Hi all, >>>>>> >>>>>> With the few days of research on the project "Introducing “curve >>>>>> fitting” for statistics prediction algorithm of Autoscaler". I have found >>>>>> followings regarding the project >>>>>> >>>>>> 1. ClusterMonitor class passes the calculated values from the >>>>>> CEP. This value passing happens after >>>>>> org.apache.stratos.messaging.event >>>>>> happens. >>>>>> 2. RuleTasksDelegator class calculates the prediction values >>>>>> using the values(Ex gradient) from the CEP >>>>>> 3. The classes in the package org.apache.stratos.cep.extension >>>>>> deals with the analysis of data in the real time and publish the >>>>>> relevant >>>>>> summarized health stats, which are needed for the autoscaling. >>>>>> >>>>>> So according to my understanding, I need to modify above mentioned >>>>>> classes, in order to implement the curve fitting for statistics >>>>>> prediction. >>>>>> >>>>>> Furthermore,for this project, we can use the regression >>>>>> implementation of CEP 4.0, which is expected to be released on May 15th >>>>>> according to WSO2 public jira. I am planning to do a feasibility study >>>>>> for >>>>>> the first two weeks of the project, where we can explore other better >>>>>> options as well. >>>>>> >>>>>> Am I on a right track? Please share your thoughts on this. >>>>>> >>>>>> [1] https://wso2.org/jira/browse/CEP >>>>>> >>>>>> Thanks >>>>>> >>>>>> Regards >>>>>> *T. Pranavan* >>>>>> *BSc Eng Undergraduate| Department of Computer Science & Engineering >>>>>> ,University of Moratuwa* >>>>>> *Intern Software Engineer**| WSO2 Lanka (Pvt) Ltd.* >>>>>> *Mobile| *0775136836 >>>>>> >>>>> >>>>> >>>>> >>>>> -- >>>>> -- >>>>> Lahiru Sandaruwan >>>>> Committer and PMC member, Apache Stratos, >>>>> Senior Software Engineer, >>>>> WSO2 Inc., http://wso2.com >>>>> lean.enterprise.middleware >>>>> >>>>> phone: +94773325954 >>>>> email: [email protected] blog: http://lahiruwrites.blogspot.com/ >>>>> linked-in: http://lk.linkedin.com/pub/lahiru-sandaruwan/16/153/146 >>>>> >>>>> >>>> >>> >> >> >> -- >> -- >> Lahiru Sandaruwan >> Committer and PMC member, Apache Stratos, >> Senior Software Engineer, >> WSO2 Inc., http://wso2.com >> lean.enterprise.middleware >> >> phone: +94773325954 >> email: [email protected] blog: http://lahiruwrites.blogspot.com/ >> linked-in: http://lk.linkedin.com/pub/lahiru-sandaruwan/16/153/146 >> >> > -- -- Lahiru Sandaruwan Committer and PMC member, Apache Stratos, Senior Software Engineer, WSO2 Inc., http://wso2.com lean.enterprise.middleware phone: +94773325954 email: [email protected] blog: http://lahiruwrites.blogspot.com/ linked-in: http://lk.linkedin.com/pub/lahiru-sandaruwan/16/153/146
