Preventing Loss of Data – failover

[Jim]

Given that KDB stores data/tables in memory, is there a best practice to periodically commit the data to disk, or rather work from disk?

I fear massive amounts of data could be lost if let’s say there was a server hardware failure/crash or maybe a kernel panic

[Ryan Hamilton]

This is a massive topic….too big to answer in one forum post.

Kdb does provide mechanisms for replication, logging and check-pointing. There’s also a standardized tick architecture for capturing market data while preventing data loss. Not to mention how to cover fail-overs amongst machines. In our advanced training course this area is spread over 3-4 modules at-least.

I recommend checking out the -r -l command line parameters as a starting point.

[Author]

Posts