> 25.09.2014 21:27, Leyne, Sean wrote: > > Having a SQL/FB running in VM on a HA VM cluster connected to HA disk > storage cluster provides a high level of resiliency, with multiple failures > necessary in order for functionality to be lost. > > I wonder how they manage synchronization between nodes in storage > cluster... AFAIU from your links it is either asynchronous with high > performance or synchronous with performance loss. Async mode is no go for > DBMS.
All changes are written to disk first, there is no in memory write-back cache. So there is never a situation where pending changes are in memory/being sync'd before being synchronized. Further, all that is required for a write operation is to send a message which invalidates the cache object when the write happens. The propogation of the changed data can be done later, if the other nodes get a request of the block in the meantime, they can read the data from disk, Win2012R2 added support for the definition of a disk/SSD which can act as a write cache, there is a read cache (recommended to not exceed 10G -- to reduce the inter-node sync overhead). It is also likely that the SMB protocol includes a 'sequence' number for each packet so that if a node sees a unexpected request (as the result of the failure of the original servicing node) it read the disk change log for the changes that it might have missed. Sean ------------------------------------------------------------------------------ Meet PCI DSS 3.0 Compliance Requirements with EventLog Analyzer Achieve PCI DSS 3.0 Compliant Status with Out-of-the-box PCI DSS Reports Are you Audit-Ready for PCI DSS 3.0 Compliance? Download White paper Comply to PCI DSS 3.0 Requirement 10 and 11.5 with EventLog Analyzer http://pubads.g.doubleclick.net/gampad/clk?id=154622311&iu=/4140/ostg.clktrk Firebird-Devel mailing list, web interface at https://lists.sourceforge.net/lists/listinfo/firebird-devel
