Improve prometheus remote_write docs (#17384)

metricbeat/module/prometheus/remote_write/_meta/docs.asciidoc

@@ -7,6 +7,23 @@ remote_write:
   - url: "http://localhost:9201/write"
 ------------------------------------------------------------------------------
 
+
+TIP: In order to assure the health of the whole queue, the following two configuration
+ https://prometheus.io/docs/practices/remote_write/#parameters[parameters] should be considered:
+
+- `max_shards`: Sets the maximum number of paralelism with which Prometheus will try to send samples to Metricbeat.
+It is recommended that this setting should be equal to the number of cores of the machine where Metricbeat runs.
+Metricbeat can handle connections in parallel and hence setting `max_shards` to the number of parallelism that
+Metricbeat can actually achieve is the optimal queue configuration.
+- `max_samples_per_send`: Sets the number of samples to batch together for each send. Recommended values are
+between 100 (default) and 1000. Having a bigger batch can lead to improved throughput and in more efficient
+storage since Metricbeat groups metrics with the same labels into same event documents.
+However this will increase the memory usage of Metricbeat.
+- `capacity`: It is recommended to set capacity to 3-5 times `max_samples_per_send`.
+Capacity sets the number of samples that are queued in memory per shard, and hence capacity should be high enough so as to
+be able to cover `max_samples_per_send`.
+
+
 Metrics sent to the http endpoint will be put by default under the `prometheus.metrics` prefix with their labels under `prometheus.labels`.
 A basic configuration would look like: