On 4/21/2016 12:50 AM, Mike Small wrote: > "Sadly it seems that we now need to either wait for Linux or Windows to > catch up with the 1980s state of the art in distributed systems (think > Locus or AFS). What went wrong? Products like DataSynapse’s FabricServer > look like an interesting attempt to address the problem, at least for > the Java world, but it feels to me that mainstream operating systems > designers seem to have lost the plot somewhere along the way."
What went "wrong" is that the author has the expectation that distributed systems are the be-all, end-all of computing. They're not. As with any highly specialized tool, distributed systems are great at some kinds of tasks and mediocre to terrible for everything else. On 4/21/2016 8:11 AM, Dan Ritter wrote: > No, because you need to deal with parallelism issues on a single > server. Pocket computers now have four simultaneously working cores. > It got really hard for CPUs to get faster -- how long has the state > of the art hovered around 4GHz -- so the process improvements lead to > more cores, instead. Eight cores in some of the flagships, now. My Tab S2 has two quad-core processors in it, one 1.9GHz and one 1.3GHz. But yeah, the single system image model is a dead end. The idea is to scale vertically but the complexity and performance issues with non-uniform memory and processing architectures scale worse than linearly. Horizontal scaling typically scales better. The performance hits are heavily front loaded in the batch/queue mechanisms. Once past these the job is free to run on all nodes in the pool or cluster independent of the other nodes' resources. -- Rich P. _______________________________________________ Discuss mailing list [email protected] http://lists.blu.org/mailman/listinfo/discuss
