Tests to evaluate the importance of various BlueSky features, so
different features are turned off for the benchmark runs.

20110305-west-fullsegments-4G
Test to S3-US-West, 4 GB disk cache, but all fetches from S3 are full
segments instead of byte ranges, and the complete segments are stored to
the disk cache.  So there are multiple effects: first, for a random
workload there is more garbage data taking up cache space; second,
fetches from S3 take longer which increases read latencies.

    --- a/bluesky/log.c
    +++ b/bluesky/log.c
    @@ -651,7 +651,7 @@
         BlueSkyRCStr *str = NULL;
         int location = 0;
         size_t file_offset = 0, file_size = 0;
    -    gboolean range_request = TRUE;
    +    gboolean range_request = FALSE;

         if (page_size == 0) {
             page_size = getpagesize();

    Sat Mar 05 17:20:01 PST 2011
     Executing run 1 of 25 ...  done
    Sat Mar 05 17:32:59 PST 2011
     Executing run 2 of 25 ...  done
    Sat Mar 05 17:47:27 PST 2011
     Executing run 3 of 25 ...  done
    Sat Mar 05 18:01:20 PST 2011
     Executing run 4 of 25 ...  done
    Sat Mar 05 18:14:38 PST 2011
     Executing run 5 of 25 ...  done
    Sat Mar 05 18:28:56 PST 2011
     Executing run 6 of 25 ...  done
    Sat Mar 05 18:43:20 PST 2011
     Executing run 7 of 25 ...  done
    Sat Mar 05 18:56:59 PST 2011
     Executing run 8 of 25 ...  done
    Sat Mar 05 19:11:18 PST 2011
     Executing run 9 of 25 ...  done
    Sat Mar 05 19:25:20 PST 2011
     Executing run 10 of 25 ...  done
    Sat Mar 05 19:39:26 PST 2011
     SfsManager shutting down ...

    Cloudlog cache: 0 dirty, 0 writeback, 0 journal, 1734715 cloud
    All segments have been flushed, journal < 12924 is clean
    Ending cache size: 4191588 kB

    Proxy statistics:
    Store[s3:mvrable-bluesky-west]: GETS: count=21781 sum=88260767105
    Store[s3:mvrable-bluesky-west]: PUTS: count=13788 sum=55866496509
    NFS RPC Messages In: count=2741556 sum=52779236892
    NFS RPC Messages Out: count=2741556 sum=1931572500

20110305-west-nogrouping-4G
All objects written out to separate log segments, to simulate storage
that is not log-structured.

  (runs of 10..50/10 ops)
    Cloudlog cache: 0 dirty, 0 writeback, 0 journal, 296986 cloud
    All segments have been flushed, journal < 2254 is clean
    Ending cache size: 4190572 kB

    Proxy statistics:
    Store[s3:mvrable-bluesky-west]: GETS: count=745 sum=17229569
    Store[s3:mvrable-bluesky-west]: PUTS: count=343683 sum=9649158666
    NFS RPC Messages In: count=492087 sum=9174696132
    NFS RPC Messages Out: count=492087 sum=577322960


Repeat run at 50 ops:
    All segments have been flushed, journal < 1430 is clean
    Ending cache size: 4190572 kB

    Proxy statistics:
    Store[s3:mvrable-bluesky-west]: GETS: count=1124 sum=24190660
    Store[s3:mvrable-bluesky-west]: PUTS: count=205132 sum=6011691408
    NFS RPC Messages In: count=253370 sum=5915763308
    NFS RPC Messages Out: count=253370 sum=202442048