sql server replication:
We have the below requirement: We need high performing read/write capacity for users across globe (concurrency appx 20K users). Users can work offline and later synchronize data. Data synchronization frequency is medium to high (every hour or so). Data size that has to be synchronized is also high (1 - 2 MB) We also need failover. The data cannot be horizontally partitioned based on region across servers(what i mean here is user from AsiaPacific and NorthAmerica can work on the same piece of data. User from any region will be equally interested in reading/writing data of any other region). There is no other good candidate for horizontal partitioning to suit our business needs. This way DPV(s) do not fit this requirement as routing request to the SQL Server that can provide most data is not logically viable in the application layer. DPV(s) also make the solution more complex. We cannot consider peer-to-peer transactional replication as we will need out of box conflict resolution support. So the option available is to have one server for each region and use Merge replication (Users can receive data, make changes offline, and later synchronize changes with the Publisher and other Subscribers.). This way users can connect to the server which is geographically closest. But my concern is replication will also have a negative impact on performance when percentage of write operation, data conflicts increases. This can be mitigated to some extent by scaling up the individual regional servers. Other than this is there any other viable option to suit this requirement that i missed out? Or is there anything that I have to take care to give high performance. Thanks much, Lalitha
Don't see what you're looking for? Try a search.
|