summary |
shortlog |
log |
commit | commitdiff |
tree
raw |
patch |
inline | side by side (from parent 1:
20b3c42)
RAS_INCREASE_STEP is by pages instead of by bytes.
However, I found it caused performance loss by setting it to be 4MB so
I set it back to 1MB. After 4MB RPC is enabled by default, more work
should be done to pick up a right value of max_read_ahead_mb to
maximize performance gain of 4MB RPC.
Signed-off-by: Jinshan Xiong <jinshan.xiong@intel.com>
Change-Id: Ic000f266cfefc827e112e03f76cf467c73ba88ad
Reviewed-on: http://review.whamcloud.com/5691
Tested-by: Hudson
Reviewed-by: Andreas Dilger <andreas.dilger@intel.com>
Tested-by: Maloo <whamcloud.maloo@gmail.com>
Reviewed-by: Oleg Drokin <oleg.drokin@intel.com>
* know what the actual RPC size is. If this needs to change, it makes more
* sense to tune the i_blkbits value for the file based on the OSTs it is
* striped over, rather than having a constant value for all files here. */
* know what the actual RPC size is. If this needs to change, it makes more
* sense to tune the i_blkbits value for the file based on the OSTs it is
* striped over, rather than having a constant value for all files here. */
-#define RAS_INCREASE_STEP(inode) (1UL << inode->i_blkbits)
+
+/* RAS_INCREASE_STEP should be (1UL << (inode->i_blkbits - CFS_PAGE_SHIFT)).
+ * Temprarily set RAS_INCREASE_STEP to 1MB. After 4MB RPC is enabled
+ * by default, this should be adjusted corresponding with max_read_ahead_mb
+ * and max_read_ahead_per_file_mb otherwise the readahead budget can be used
+ * up quickly which will affect read performance siginificantly. See LU-2816 */
+#define RAS_INCREASE_STEP(inode) (ONE_MB_BRW_SIZE >> CFS_PAGE_SHIFT)
static inline int stride_io_mode(struct ll_readahead_state *ras)
{
static inline int stride_io_mode(struct ll_readahead_state *ras)
{