The non-buffered writer is now deprecated. Also the amount of compression achieved and the CPU overhead is much better when using the buffered writer.
In our testing the amount of data sent by the NSQ library with Snappy compression enabled didn't improve much at all without this fix - the data was perhaps 4/5s of the original size. With the fix it's closer to 50%
The non-buffered writer is now deprecated. Also the amount of compression achieved and the CPU overhead is much better when using the buffered writer.
In our testing the amount of data sent by the NSQ library with Snappy compression enabled didn't improve much at all without this fix - the data was perhaps 4/5s of the original size. With the fix it's closer to 50%