we are just about to put together a new server for our applications using ObjectDB (v 2.6). We have a dilemma about the storage holding the ObjectDB files:
if to invest in a mirrored RAID of SSDs which is clearly the ultimate solution but it quite pricey, or to choose a RAID5/6 of 7200 RPM HDDs.
The reason we are asking the question in this forum is that we would need to know how ObjectDB is reading/writing the files. The tests are showing that the HW RAID5/6 of HDDs deliver comparable (but still lower) speeds of data transfer to a single SSD, however only when tested on large media files, so in fact it means sequential reads/writes. If it is tested on small files the speed is getting considerably worse.
Our DB files range from 700MB to 5GB.
So the question which arises: does ObjectDB read/write the DB files by large chunks and then it parses it within the RAM (so it would mean reading/writing of large data fewer times) or the DB file is only a big container in which every DB request is searching for the correct place to read/write right on the disk (it would mean then r/w of small data more often, so the seek-time is the most crucial and not the transfer-rate) ?