Novel algorithmic ideas for big data have not been accompa- nied by advances in the way central memory is allocated to concurrently running programs. Commonly, RAM is poorly managed since the pro- grams' trade o s between speed of execution and RAM consumption are ignored. This trade o is, however, well known to the programmers. We adopt mechanism design tools to truthfully elicit this (multidimensional) information with the aim of designing more clever RAM allocation al- gorithms. We introduce a novel paradigm wherein programs are bound to overbidding declarations of their running times. We show the limi- tations of this paradigm in the absence of transfers and prove how to leverage waiting times, as a currency, to obtain optimal money burning mechanisms for the makespan.
|Publication status||Published - 9 Dec 2015|
|Event||11th Conference on Web and Internet Economics - CWI Amsterdam, Amsterdam, Netherlands|
Duration: 9 Dec 2015 → 12 Dec 2015
|Conference||11th Conference on Web and Internet Economics|
|Abbreviated title||WINE 2015|
|Period||9/12/15 → 12/12/15|