<slide>
<title>Sharding</title>

<list>
	<bullet>Split the pipeline at first %$group% or %$sort%
		<list>
			<bullet>Shards execute pipeline up to that point</bullet>
			<bullet>mongos merges results and continues</bullet>
		</list>
	</bullet>
    <bullet>Early %$match% may excuse shards</bullet>
    <bullet>CPU and memory implications for mongos</bullet>
</list>
</slide>
