In the big machine, disk striping is likely a nested graph model, right? That is how most of these linked memory mangers work. But that is exactly what we want to manipulate with the graph convolution syntax. So we cpondier the query optimization process as a mobile web bot. This thing can cruise through the Cray XMT semantic nest looking for high click thru counts, passing and collecting the keywords that get them, along with their predicates. The result is a better key word index, groupign the highest click tru counts together, maybe.
Then you run the second web bot, it looks over the list and breaks it out into sub graphs based upon required striping patters. So, striping patterns become named graphs, subject to morphing. The result is high eficiency bandwidth matching, click thru counts, matching disk io, matching nested store decomoposition optimized in shared memory. You get flow with minimal number of processing steps, but flow matched to client habits.
Threads, javabyte codes, dom trees cvs files, mallocs; all of them can be modeled as a graph of sets and descents, incorporated into an abstract graph layer with exposed node pointers and predicates.
Related, what is named graph transparency?
It means that at any level, when the client wants to search slightly beyond his norm, then the client may be slightly annoyed by named graph syntax.
No comments:
Post a Comment