Reader/Writer Lock Pattern

Reader/Writer Lock Pattern

A reader-writer lock is a lock which will allow multiple concurrent readers but only one writer. A reader-writer lock can be significantly more efficient than a standard mutex if reads on your shared memory far outnumber writes.

Reader-writer locks naturally fit together with caches, as caches are only effective if reads far outnumber writes.

Here is a general pattern for using a reader-writer lock with a cache:

  1. Acquire a reader lock.
  2. Check the cache for the value. If it exists, save the value and go to step 8.
  3. Upgrade the reader lock to a writer lock.
  4. Check the cache for the value. If it exists, save the value and go to step 7.
  5. Calculate the value (expensive, otherwise we wouldn’t cache it)
  6. Insert the value into the cache.
  7. Release the writer lock.
  8. Release the reader lock.
  9. Return the value.

The reason why we have to check the cache for the value again in step (4) is because of the following possibility (assume step 4 doesn’t exist):

Thread 1 Thread 2
Acquire reader lock Acquire reader lock
Check cache for value with key A (not found) Check cache for value with key A (not found)
Upgrade to writer lock (block)
Calculate value (expensive)
Insert value into cache
Release writer lock
Release reader lock
Upgrade to writer lock
Calculate value (expensive) (We are paying this cost twice)
Insert value into cache (We are inserting two values with the same key, which may be fatal)

With step 4 it becomes:

Thread 1 Thread 2
Acquire reader lock Acquire reader lock
Check cache for value with key A (not found) Check cache for value with key A (not found)
Upgrade to writer lock (block)
Calculate value (expensive)
Insert value into cache
Release writer lock
Release reader lock
Upgrade to writer lock
Upgrade to writer lock
Check cache for value with key A (found)
Release writer lock
Return value Return value