I would be suspicious of the results in that paper. The same with a similar study published by IBM comparing their streaming offering to storm.
It's not that I think the studies were invalid. It's that I have no way to tell. Without publishing the code and configuration used in the study, there is no way the results can be independently verified. Storm come with a lot of knobs for tuning, and it is very easy to cripple performance with the wrong settings. So without the code and config, there's no way to tell if it's anything other than marketing. If anyone knows where the code AND config for those studies can be found I would be very interested. I couldn't find anything. -Taylor > On May 11, 2014, at 3:19 PM, Weide Zhang <[email protected]> wrote: > > Hi Cody, > > Thanks for your reply. Do you know if there is any performance test result on > Trident on throughput versus spark streaming ? According to the spark > streaming paper, it claims to achieve better throughput than storm but I'm > not sure if there is some insights why from storm community ? > > >> On Sun, May 11, 2014 at 11:13 AM, Cody A. Ray <[email protected]> wrote: >> Have you read these? >> >> https://github.com/nathanmarz/storm/wiki/Trident-tutorial >> >> https://github.com/nathanmarz/storm/wiki/Trident-state >> >> -Cody >> >> >>> On Mon, May 5, 2014 at 6:38 PM, Sid Kiblawi <[email protected]> wrote: >>> I'm also interested in seeing some documentation on how trident implements >>> this. Any help would be appreciated. >>> >>> -Sid >>> >>> On 05/04/14, Weide Zhang wrote: >>> > Hi , >>> > >>> > Do you know if there is a good documentation on how trident implement >>> > exact once semantic ? Is there any performance trade off on using trident >>> > on achieving exact once semantic. How does trident performance compared >>> > with spark streaming ? >>> > >>> > >>> > Thanks a lot, >>> > >>> > >>> > Weide >> >> >> >> -- >> Cody A. Ray, LEED AP >> [email protected] >> 215.501.7891 >
