[Assorted-commits] SF.net SVN: assorted:[1190] ydb/trunk/README

[<yan...@us...>]

Revision: 1190
          http://assorted.svn.sourceforge.net/assorted/?rev=1190&view=rev
Author:   yangzhang
Date:     2009-02-17 22:39:20 +0000 (Tue, 17 Feb 2009)

Log Message:
-----------
added some more notes post-meeting

Modified Paths:
--------------
    ydb/trunk/README

Modified: ydb/trunk/README
===================================================================
--- ydb/trunk/README	2009-02-17 22:39:12 UTC (rev 1189)
+++ ydb/trunk/README	2009-02-17 22:39:20 UTC (rev 1190)
@@ -333,7 +333,7 @@
 - DONE make readmsg perform fewer syscalls (buffer opportunistically)
   - like magic: now can sustain 90 Ktps all the way up through 3 xacts!
 
-Period 2/12-2/19
+Period 2/12-2/17
 
 - DONE p2 prototype
   - some interesting performance bugs
@@ -373,13 +373,33 @@
   - 1: 220 -> <240
   - 3: 220 -> <240
 - DONE try adding fake-execution
-  - made a huge difference
+  - made a huge difference, but not in the direction we want
   - -1: 680K
   - 0: 2M
   - 1: 730K
   - 2: 600K
   - 3: 450K
-- TODO commit
+- DONE try reverting to map
+  - still not in the direction we want
+  - -1: 122K
+  - 0: 140K
+  - 1: 122K
+  - 2: 122K/107K
+  - 3: 122K/122K/97K
+- DONE commit
+- DONE add zero-copy structs/(de-)serialization
+
+Period 2/17-
+
+- TODO dynamic switch between pb and zero-copy
+- TODO async (threaded) wal
+- TODO 0-node 0-copy (need to use threads)
+- TODO google dense hash map
+
+- TODO show aries-write
+- TODO checkpointing + replaying log from replicas (not from disk)
+- TODO scale-up on multicore
+
 - TODO remove extraneous copies; use custom buffer-backed data structures
   designed for serialization/deserialization
 - TODO flushing


This was sent by the SourceForge.net collaborative development platform, the world's largest Open Source development site.