Safety Guarantee of continuous join queries over punctuated data streams

Continuous join queries (CJQ) are needed for correlating data from multiple streams. One fundamental problem for processing such queries is that since the data streams are infinite, this would require the join operator to store infinite states and eventually run out of space. Punctuation semantics has been proposed to specifically address this problem. In particular, punctuations explicitly mark the end of a subset of data and, hence, enable purging of the stored data which will not contribute to any new query results. Given a set of available punctuation schemes, if one can identify that a CJQ still requires unbounded storage, then this query can be flagged as unsafe and can be prevented from running. Unfortunately, while Punctuation semantics is clearly useful, the mechanisms to identify if and how a particular CJQ could benefit from a given set of punctuation schemes are not yet known. In this paper, we provide sufficient and necessary conditions for checking whether a CJQ can be safely executed under a given set of punctuation schemes or not. In Particular, we introduce a novel punctuation graph to aid the analysis of the safety for a given query. We show that the safety checking Problem can be done in polynomial time based on this punctuation graph construct. In addition, various issues and challenges related to the safety checking of CJQs are highlighted.