hahahahahaha now *that* is another topic altogether On Fri, Jan 30, 2015 at 7:50 PM, Konstantin Boudnik <[email protected]> wrote:
> With Mesos - who needs damn YARN? > > On Fri, Jan 30, 2015 at 07:06PM, jay vyas wrote: > > here mesos enters the picture :) deploying mesos as a deployment > platform > > will allow bigtop to support > > > > 1) YARN > > > > 2) non-YARN systems equally, and also (great soln for spark, hbase) > > > > 3) container based systems > > > > On Fri, Jan 30, 2015 at 6:36 PM, Konstantin Boudnik <[email protected]> > > wrote: > > > > Good points! I think if we find a good way to express the cluster > > topology and > > orchestrate it in such a way that we are only running a single > Hadoop > > service > > per a container, then we might get away with covering both.... > sorta. > > > > Still leaves the question about HDFS storage though. > > A Cos > > On Fri, Jan 30, 2015 at 05:43PM, jay vyas wrote: > > > honestly, i think you could go either way.A certainly people > will be > > > running hadoop services in docker containers in the very near > future, > > > and at definetely at scale. > > > > > > at the same time, some folks will still want /usr/lib and /etc > and VM > > style > > > multitenant solutions. > > > > > > my point is : We cant really support both :) so we probably would > be > > better > > > off picking one :) ... > > > > > > both are valid paths to go down , IMO, just different audiences, > and > > > different trade offs > > > > -- > > jay vyas > -- jay vyas
