Re: Write speed during Local Checkpoints

[Jörg Nowak]

Hi Anton,

I analysed the startup procedure a little more and found out that the cluster writes a local checkpoint to the disks before it is available. Read data from disk and load into the memory + synchronize with other nodes takes less than 50 % of this whole procedure. 13 min of 23 minutes it needs to write the local checkpoint.
Of course the cluster is designed for high availibity. But what do you want to do in case of a cluster shutdown (for instance: hardware problem, power failure, fire, water) ? In this case we want to start up a new cluster in a different location with the same data (disks are mirrored to a remote location). But no important business can wait more than 20 min for a cluster start up.


I hope you understand my problem.

Joerg Nowak