gnunet-svn
[Top][All Lists]
Advanced

[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

[GNUnet-SVN] r35429 - gnunet/src/fs


From: gnunet
Subject: [GNUnet-SVN] r35429 - gnunet/src/fs
Date: Mon, 23 Mar 2015 18:37:07 +0100

Author: grothoff
Date: 2015-03-23 18:37:07 +0100 (Mon, 23 Mar 2015)
New Revision: 35429

Modified:
   gnunet/src/fs/gnunet-service-fs_cp.h
Log:
-re benchmarking and adjusting MAX_QUEUE_PER_PEER

Modified: gnunet/src/fs/gnunet-service-fs_cp.h
===================================================================
--- gnunet/src/fs/gnunet-service-fs_cp.h        2015-03-23 17:21:14 UTC (rev 
35428)
+++ gnunet/src/fs/gnunet-service-fs_cp.h        2015-03-23 17:37:07 UTC (rev 
35429)
@@ -33,19 +33,16 @@
 /**
  * Maximum number of outgoing messages we queue per peer.
  *
- * Performance measurements for 2 peer setup for 50 MB file
- * (with MAX_DATASTORE_QUEUE = 1 and RETRY_PROBABILITY_INV = 1):
+ * Performance measurements for 2 peer setup for 10 MB file
+ * (using perf_gnunet_service_fs_p2p):
  *
- *   2: 1700 kb/s, 1372 kb/s
- *   8: 2117 kb/s, 1284 kb/s, 1112 kb/s
- *  16: 3500 kb/s, 3200 kb/s, 3388 kb/s
- *  32: 3441 kb/s, 3163 kb/s, 3277 kb/s
- * 128: 1700 kb/s; 2010 kb/s, 3383 kb/s, 1156 kb/s
+ *  24: 2-3 MB/s # ~ 24 MB RAM
+ * 256:   8 MB/s # ~256 MB RAM
  *
- * Conclusion: 16 seems to be a pretty good value (stable
- * and high performance, no excessive memory use).
+ * Conclusion: 24 should suffice (reasonable
+ * performance, no excessive memory use).
  */
-#define MAX_QUEUE_PER_PEER 16
+#define MAX_QUEUE_PER_PEER 24
 
 /**
  * Length of the P2P success tracker.  Note that having a very long




reply via email to

[Prev in Thread] Current Thread [Next in Thread]