 |
 |
Well, that was my exact point. We need to understand whether or not there *is* temporal locality (at least within a 64MB page) to figure out where this is best solved. Without that knowledge it's basically a whack a mole exercise. On Mon, 10 Aug 2020, Brian P Bockelman wrote: >> As posted in my previous comment, delaying the read will likely push the problem elsewhere and won't get rid of it > > This is likely true unless there's temporal locality. That is, if LHCb happens to revisit the same 64MB block repeatedly, it might be beneficial to slow down and cache it at the WN. This was the CMS case because CMS scans a single metadata branch scattered throughout the file then reads linearly through the file. > > Given how LHCb is nicely using ReadV, I suspect we are not in this exception... > > -- > You are receiving this because you were mentioned. > Reply to this email directly or view it on GitHub: > https://github.com/xrootd/xrootd/issues/1259#issuecomment-671544584 -- You are receiving this because you were mentioned. Reply to this email directly or view it on GitHub: https://github.com/xrootd/xrootd/issues/1259#issuecomment-671549782 ######################################################################## Use REPLY-ALL to reply to list To unsubscribe from the XROOTD-DEV list, click the following link: https://listserv.slac.stanford.edu/cgi-bin/wa?SUBED1=XROOTD-DEV&A=1