New storage mechanism with better concurrency characteristics

[GoogleCodeExporter]

The current storage relies on One Big Lock around the entire cache. This 
strategy puts a cap on concurrent accesses to the cache, making one write 
block all readers.

There are better data structures that could be used; for example the 
ConcurrentLinkedHashMap project here on google code.   However, this piece 
of code would require extensive modification to support the locked-delayed-
deletes functionality of memcache.

This task is quite far reaching; it really means refactoring the call chain 
from 
the protocol handler down to move the locking/cas/delayed-delete support 
further down.

Original issue reported on code.google.com by ryan.daum on 11 Sep 2009 at 3:46

[GoogleCodeExporter]

Original comment by ryan.daum on 11 Sep 2009 at 8:12

• Changed state: Started

[GoogleCodeExporter]

I made use of an early version of the ConcurrentLinkedHashMap project, modified 
to 
support sizing of elements.

This is still being tested, but early results are promising.

Original comment by ryan.daum on 8 Nov 2009 at 6:54

• Changed state: Fixed