Last week the remote satellite of the Data Warehouse in Japan crashed as I wrote about in the previous post.. It stubbornly refused to start up again. It was dead, it was no more, it had expired and gone to meet his maker, end of story.
What do you do when a server crash? When it is beyond any hope for a quick resurrection. Either you buy a new server and do a bare-metal-restore which is almost always an agonising business.If you never done a bare metal restore, believe me you do not want to do that. Or you switch to the failover backup if you have one.
The satellite in Japan was a partial replica of the central Data Warehouse, identified by it's DNS address, all reports, processes etc. picked up the data from the Data Warehouse via the DNS address, so it was just a matter of changing the DNS address and the Data Warehouse was back in business again, everything worked flawlessly a seamless transition.
Today failover is common but very expensive. The Data Warehouse failover solution costed about 1000$. Take any of the big guys Business Intelligence solutions, you will spend way more than 10 times that amount and you will probably not get a thenth of the performance improvements of a DW satellite. The very reason for a DW satellite is performance, LAN response times even for a faraway workshop and more secure connections, if there are WAN disruptions the local Satellite BI data is still there. The BI satellite concept is nothing less than brilliant.