Hi group,
Unless I’m mistaken, the procedure to create a zip archive reads like a recursive collection of all relevant objects, and then writing them out sequentially, in a single thread.
Is this assessment correct?
I was wondering if a highly concurrent fetching phase could be optionally added…
Background is the proliferation of cold archive / high latency storage – traversing a tree in a recursive manner, single threaded, will accumulate the maximum amount of wallclock time;
A first improvement could be to run all the recursive traversals in a parallel (multi threaded) way – with dispatched threads to issue prefetch read calls (to warm up the various caches). And in a second phase traverse the (now warmed up) tree single-threaded, and append the objects to the archive…
Subsequently, if sufficient memory can be allocated, the asynchronously dispatched reads could be fetched in memory, and as soon as the reads for the next in-order object is completed (these reads individually may not complete in-order), append the objects to the archive…
Best regards,
Richard Scheffenegger