Abstract: Ranking systems are described. In an embodiment a large scale data center has peta bytes of items and a query engine is provided to find the top k most frequently occurring items. In embodiments, samples are taken from the data center at least until a specified number of samplings is met, or until a stopping rule is met. In examples, the samples form a sample sketch which is used to find the top k most frequently occurring items without the need to examine every item in the data center. In other examples, the number of samplings or stopping rule is varied to provide ranks or frequencies. In other embodiments the ranking system operates on items having values to find separators which divide the items into bins such that the proportion of the items in each bin is different. For example, a data set may be apportioned to different types of processor.

Abstract: A method of ranking and suggesting objects, such as tags, is described in which a group of objects are suggested to a user in relation to a particular information object, such as a web page or file. A user either selects one of the suggestions or enters an alternative object and on receipt of data identifying the particular object selected by the user, the group of suggestions is updated based on whether the object selected by the user was one of the suggestions or not. The updating of the group may also be based on other factors, such as ranking parameters associated with each of the objects.