Address load-aware active-slot review comments

Author: sunchao

common/src/main/scala/org/apache/celeborn/common/CelebornConf.scala

@@ -3158,7 +3158,6 @@ object CelebornConf extends Logging {
 
   val MASTER_SLOT_ASSIGN_LOADAWARE_ACTIVE_SLOTS_WEIGHT: ConfigEntry[Double] =
     buildConf("celeborn.master.slot.assign.loadAware.activeSlotsWeight")
-      .withAlternative("celeborn.slots.assign.loadAware.activeSlotsWeight")
       .categories("master")
       .doc(
         "Weight of active slots when calculating ordering in load-aware assignment strategy")

docs/configuration/master.md

@@ -77,7 +77,7 @@ license: |
 | celeborn.master.slot.assign.extraSlots | 2 | false | Extra slots number when master assign slots. Provided enough workers are available. | 0.3.0 | celeborn.slots.assign.extraSlots | 
 | celeborn.master.slot.assign.interruptionAware | false | false | If this is set to true, Celeborn master will prioritize partition placement on workers that are not in scope for maintenance soon. | 0.7.0 |  | 
 | celeborn.master.slot.assign.interruptionAware.threshold | 50 | false | This controls what percentage of hosts would be selected for slot selection in the first iteration of creating partitions. Default is 50%. | 0.7.0 |  | 
-| celeborn.master.slot.assign.loadAware.activeSlotsWeight | 0.0 | false | Weight of active slots when calculating ordering in load-aware assignment strategy | 0.7.0 | celeborn.slots.assign.loadAware.activeSlotsWeight | 
+| celeborn.master.slot.assign.loadAware.activeSlotsWeight | 0.0 | false | Weight of active slots when calculating ordering in load-aware assignment strategy | 0.7.0 |  |
 | celeborn.master.slot.assign.loadAware.diskGroupGradient | 0.1 | false | This value means how many more workload will be placed into a faster disk group than a slower group. | 0.3.0 | celeborn.slots.assign.loadAware.diskGroupGradient | 
 | celeborn.master.slot.assign.loadAware.fetchTimeWeight | 1.0 | false | Weight of average fetch time when calculating ordering in load-aware assignment strategy | 0.3.0 | celeborn.slots.assign.loadAware.fetchTimeWeight | 
 | celeborn.master.slot.assign.loadAware.flushTimeWeight | 0.0 | false | Weight of average flush time when calculating ordering in load-aware assignment strategy | 0.3.0 | celeborn.slots.assign.loadAware.flushTimeWeight | 

docs/developers/slotsallocation.md

@@ -46,6 +46,11 @@ Load-aware slots allocation will take following elements into consideration.
 
 Slots allocator will find out all worker involved in this allocation and sort their disks by 
 `disk's average flushtime * flush time weight + disk's average fetch time * fetch time weight + disk's active slots * active slots weight`.
+The average flush/fetch times are measured in nanoseconds, while active slots is a slot count, so
+`activeSlotsWeight` is effectively a nanoseconds-per-slot conversion factor. For example, if the
+average fetch time is around `100 ms` (`10^8` ns) and a disk has about `1000` active slots,
+`activeSlotsWeight=10^5` makes the active-slot term contribute about `10^8`, comparable to the
+fetch-time term.
 After getting the sorted disks list, Celeborn will split the disks into
 `celeborn.master.slot.assign.loadAware.numDiskGroups` groups. The slots number to be placed into a disk group 
 is controlled by the `celeborn.master.slot.assign.loadAware.diskGroupGradient` which means that a group's