evaluation.tex

Use non-determinism in timing of distributed executions and targeted node delays to converge on stable set of relationships

\subsection{Non-determinism in multiple runs}
TODO: Plots of the reduction in number of hypotheses v/s number of runs

TODO: How does number of hypotheses reduce in more general case? (table)

\subsection{Delaying targeted nodes}
TODO: Graph of 500 hypotheses -- Graph obtained from reducing dependencies from multiple runs -- Graph from reducing dependencies using targeted delays vis a vis True graph

TODO: How does number of hypotheses reduce in more general case? (table)

\subsection{Scaling and convergence}
TODO: Runs for 5 node and 7 node cluster for a single insert query and determine how many runs it take to converge. Primarily to show that the number of runs does not scale linearly with increase in replication factor