One of the last projects I worked on had a database that would take 8 days to "restore" / "re-import". But by doing some crazy hardware and software things we changed this to just over two hours.
I was going to post the same question. (lost newbie here) There was an interesting thread a while ago that touched on storing data. http://www.elitetrader.com/vb/showthread.php?s=&threadid=25212&highlight=firebird+sql In this thread, it was suggested that if you "serialize" the data (using c#) and store this in a database, it is fast and efficient. What does this mean? I believe "serialize" means to store an object, but what are you defining as an object? A chunk of data or each piece of data?
Don't know if you directed the question at me. I am still a newbie as far as C# goes. I would expect (from other experience) that this would be a "chunk" of data rather than individual data. Further to nitro: I dug into some of my archives. One way of making it "power safe" is to have a "solid state" hard disk and have to the solid state hard disk a physical hard disk attached with a UPS. Power fails - data from solid state hard disk gets swapped out to hard disk and then everything gets shuts down cleanly. If you have two solid state hard disks then the whole setup becomes redundant (this was the setup when I audited an exchange for their hard and software architecture and what got me interested in trading a long time ago) Hope this helps to clarify a bit. Marinus
Memory-based DBs typically operate like write-through caches for those parts that need powerfail resistance. Adds and updates are updated in memory while simultaneously flushed to a disk-based transcript/log file. Upon restart, the DB manager code rebuilds the DB from the disk log and then proceeds with normal operation.