Adding additional comments

src/server/mapper/block_allocator.js

@@ -93,9 +93,12 @@ function refresh_pool_alloc(pool) {
  * allocate_node
  *
  * @param avoid_nodes array of node ids to avoid
+ * @param content_tiering_params - in case of content tiering, the additional
+ * replicas will be saved in nodes that have the best disk read latency, but only
+ * from the chunk of nodes that we've received in pools.
  *
  */
-function allocate_node(pools, avoid_nodes, params) {
+function allocate_node(pools, avoid_nodes, content_tiering_params) {
     let pool_set = _.map(pools, pool => pool._id.toString()).sort().join(',');
     let alloc_group =
         alloc_group_by_pool_set[pool_set] =
@@ -106,9 +109,13 @@ function allocate_node(pools, avoid_nodes, params) {
             })))
         };
 
-    // Used in map_builder, for content tiering
-    if (params && params.special_replica) {
+    // If we are allocating a node for content tiering special replicas,
+    // we should run an additional sort, in order to get the best read latency nodes
+    if (content_tiering_params && content_tiering_params.special_replica) {
         alloc_group.nodes = _.sortBy(alloc_group.nodes, node =>
+            // In order to sort the nodes by the best read latency values.
+            // We need to get the average of all the latency disk read values,
+            // and sort the nodes by the average that we've calculated.
             _.sum(node.latency_of_disk_read) / node.latency_of_disk_read.length
         );
     }

src/server/mapper/map_builder.js

@@ -110,6 +110,8 @@ class MapBuilder {
                     'digest_type',
                     'digest_b64');
                 block._id = md_store.make_md_id();
+                // We send an additional flag in order to allocate
+                // replicas of content tiering feature on the best read latency nodes
                 let node = block_allocator.allocate_node(alloc.pools, avoid_nodes, {
                     special_replica: true
                 });