branches/zip: The freed_page_clock for both buf_pool and block starts ticking

only after the buffer_cache has warmed up. During the initial warmup
phase these values are set to zero. We should recommend to move the
block to the start of the LRU in this case.

Also added some comments to buf0buf.c about the LRU logic.

reviewed by: Sunny

buf/buf0buf.c

@@ -135,16 +135,35 @@ The LRU-list contains all the blocks holding a file page
 except those for which the bufferfix count is non-zero.
 The pages are in the LRU list roughly in the order of the last
 access to the page, so that the oldest pages are at the end of the
-list. We also keep a pointer to near the end of the LRU list,
-which we can use when we want to artificially age a page in the
-buf_pool. This is used if we know that some page is not needed
-again for some time: we insert the block right after the pointer,
-causing it to be replaced sooner than would noramlly be the case.
-Currently this aging mechanism is used for read-ahead mechanism
-of pages, and it can also be used when there is a scan of a full
-table which cannot fit in the memory. Putting the pages near the
-of the LRU list, we make sure that most of the buf_pool stays in the
-main memory, undisturbed.
+list. The LRU logic has two optimizations implemented in it.
+
+Artificial aging of pages:
+--------------------------
+This is achieved by maintaining a pointer at 3/8 from the tail of
+the LRU. All pages that are read in are inserted at this location.
+The pages are moved to the front of LRU when they are *accessed*.
+How this helps? A regular read request will result in immediately
+accessing the page and the block will be moved to the front of the
+LRU. The reads done as part of read-ahead mechanism (random read-ahead,
+linear read-ahead etc.) will leave these pages in the lower 3/8 of the
+LRU. If these pages are not accessed they will be eventually evicted
+without polluting the whole buffer cache.
+
+Reduced LRU manipulation:
+-------------------------
+In a classic LRU each access to a page should result in putting it
+to the front of LRU. LRU manipulation, however, can lead to mutex
+contention. To avoid this, we only make a block "young" if we see
+that it has slipped past the top 1/4th of the LRU.
+We implement this heuristic by maintaining a freed_page_clock at the
+buffer pool level and in each block as well. By comparing these two
+values, without holding any locks, we can determine how far a block is
+from the front of LRU.
+Both these clocks start ticking only after victim eviction has started
+in earnest (i.e.: no blocks in free list). Hence we cannot determine
+location of a block during the initial warm up phase and, therefore, we
+disable this optimization during that period. Look at
+buf_block_peek_if_too_old() in include/buf0buf.ic
 
 The chain of modified blocks contains the blocks
 holding file pages that have been modified in the memory

include/buf0buf.ic

@@ -38,7 +38,11 @@ buf_block_get_freed_page_clock(
 /************************************************************************
 Recommends a move of a block to the start of the LRU list if there is danger
 of dropping from the buffer pool. NOTE: does not reserve the buffer pool
-mutex. */
+mutex.
+Note that freed_page_clock for both buf_pool and block starts ticking
+only after the buffer_cache has warmed up. During the initial warmup
+phase these values are set to zero. We always recommend to to move the
+block to the start of the LRU in this case. */
 UNIV_INLINE
 ibool
 buf_page_peek_if_too_old(
@@ -47,9 +51,10 @@ buf_page_peek_if_too_old(
 					younger */
 	const buf_page_t*	bpage)	/* in: block to make younger */
 {
-	return(buf_pool->freed_page_clock
+	return(buf_pool->freed_page_clock == 0
+	       || (buf_pool->freed_page_clock
 		   >= buf_page_get_freed_page_clock(bpage)
-	       + 1 + (buf_pool->curr_size / 4));
+		   + 1 + (buf_pool->curr_size / 4)));
 }
 
 /*************************************************************************