batch queries to maximize throughput

query-side streams for first level of callsites

Signed-off-by: Jason Volk <jason@zemos.net>

conduwuit-example.toml

@@ -1457,6 +1457,19 @@
 #
 #stream_width_scale = 1.0
 
+# Sets the initial amplification factor. This controls batch sizes of
+# requests made by each pool worker, multiplying the throughput of each
+# stream. This value is somewhat abstract from specific hardware
+# characteristics and can be significantly larger than any thread count or
+# queue size. This is because each database query may require several
+# index lookups, thus many database queries in a batch may make progress
+# independently while also sharing index and data blocks which may or may
+# not be cached. It is worthwhile to submit huge batches to reduce
+# complexity. The maximum value is 32768, though sufficient hardware is
+# still advised for that.
+#
+#stream_amplification = 1024
+
 # Number of sender task workers; determines sender parallelism. Default is
 # '0' which means the value is determined internally, likely matching the
 # number of tokio worker-threads or number of cores, etc. Override by