Synchronous Flash Storage for Extended Distance Clusters
Extended distance clusters provide a local disaster recovery solution across two locations with high availability characteristics of a local cluster. Unlike replication techniques, in extended distance clusters the data is mirrored by ASM synchronously across the two locations. At any given time all locations have exact full copy of the data.
Extended distance ASM clusters can be configured with RAC, RAC One Node, or single-instance databases.
The added latency on direct long-distance connections is approximately 0.005 ms/km. For a 100 km distance this produces 0.5 ms latency, which is still 10X lower than a typical latency of a traditional HDD based storage. Distances up to 100 km are acceptable for most OLTP loads. More careful testing of the target application is recommended for distances over 100 km.
The smallest storage configuration can consist of one NVMe SSD in each of the database nodes. Additional SSDs can increase capacity, performance, and also improve HA.
Dedicated 10 GbE connectivity with no routing is recommended between the sites. RDMA enabled network can be used for high-throughput applications to minimize CPU utilization. Long-haul InfiniBand switches can be used for distances up to 80 km.