Hi On Sat, Nov 8, 2014 at 7:57 PM, Lahiru Sandaruwan <[email protected]> wrote:
> Hi all, > > I'm working on $subject. When we predict the instance count from > statistics and dependency based ratio in application, we will get numbers > with fraction. E.g. 1.2, 3.9, 4.5. Therefore we have to round the instance > count for creating instances in IaaS. > > > I have thought of three types to choose from. > > > 1. Floor > 2. Ceiling > 3. Custom percentage > > #3 means, user can give a limit of fraction to decide whether to round to > floor or ceiling. > > e.g. User fives fraction as 0.2, > > Instance count 3.1 Rounded value 3 > Instance count 3.21: Rounded value 4 > Instance count 1.19: Rounded value 1 > > Feel free to give feedback. > +1 to use #3 as it covers all the scenarios. But how are we going to handle the ratio between dependent cartridges? For e.g.: If C1:C2 = 2:3 where C1 and C2 are dependent clusters, then we got the instance count to be scaled up for C1 as 1.19 by the statistics of C1 which will scale C1 by 1. Then the relative instance count to be scaled for C2 can be either decided by (1/2)*3 or against with the existing count of C1 and C2. Are we going to use the same fraction to keep the ratio as well, since even the ratio calculation won't be round figures all the time? Thanks, Reka > > Thanks. > > > > > -- > -- > Lahiru Sandaruwan > Committer and PMC member, Apache Stratos, > Senior Software Engineer, > WSO2 Inc., http://wso2.com > lean.enterprise.middleware > > email: [email protected] blog: http://lahiruwrites.blogspot.com/ > linked-in: http://lk.linkedin.com/pub/lahiru-sandaruwan/16/153/146 > > -- Reka Thirunavukkarasu Senior Software Engineer, WSO2, Inc.:http://wso2.com, Mobile: +94776442007
