Tuning the network
Another key area to focus on is the network. A poorly performing network will cause high latencies and possible disconnections. One easy method to determine whether your GoldenGate implementation is suffering from network delays is to check that the process write checkpoints are increasing at the same rate. For example, if your primary Extract process is check-pointing frequently when compared to the Data Pump process, this indicates that the Data Pump process cannot write to the remote trail quickly enough. So what can be done about it?
We can look at increasing data throughput by adjusting OS TCP socket buffers, TCP packet sizes, transmission queue lengths, and so on. Let's take a look at some of the common network tuning parameters and tools.
Linux TCP tuning
Like most operating systems, the default maximum kernel parameter settings for networking are way too small. For example, Linux kernel 2.6 has a maximum TCP buffer size of 256KB for both send and receive. During high...
