clusterperf August 13, 2013
- Collin Lee
Owned by Collin Lee
As of a0f6889
basic.read100 5.2 us read single 100B object with 30B key
basic.readBw100 18.4 MB/s bandwidth reading 100B object with 30B key
basic.read1K 6.8 us read single 1KB object with 30B key
basic.readBw1K 140.5 MB/s bandwidth reading 1KB object with 30B key
basic.read10K 10.2 us read single 10KB object with 30B key
basic.readBw10K 938.5 MB/s bandwidth reading 10KB object with 30B key
basic.read100K 46.4 us read single 100KB object with 30B key
basic.readBw100K 2.0 GB/s bandwidth reading 100KB object with 30B key
basic.read1M 419.4 us read single 1MB object with 30B key
basic.readBw1M 2.2 GB/s bandwidth reading 1MB object with 30B key
basic.write100 15.5 us write single 100B object with 30B key
basic.writeBw100 6.2 MB/s bandwidth writing 100B object with 30B key
basic.write1K 19.2 us write single 1KB object with 30B key
basic.writeBw1K 49.7 MB/s bandwidth writing 1KB object with 30B key
basic.write10K 35.3 us write single 10KB object with 30B key
basic.writeBw10K 269.9 MB/s bandwidth writing 10KB object with 30B key
basic.write100K 226.7 us write single 100KB object with 30B key
basic.writeBw100K 420.8 MB/s bandwidth writing 100KB object with 30B key
basic.write1M 2.2 ms write single 1MB object with 30B key
basic.writeBw1M 431.8 MB/s bandwidth writing 1MB object with 30B key
broadcast 194.7 us broadcast message to 9 slaves
netBandwidth 572.3 MB/s many clients reading from different servers
netBandwidth.max 17.1 MB/s fastest client
netBandwidth.min 14.2 MB/s slowest client
readNotFound 13.6 us read object that doesn't exist
# Gauges impact of asynchronous writes on synchronous writes.
# Write two values. The size of the first varies over trials
# (its size is given as 'firstObjectSize'). The first write is
# either synchronous (if firstWriteIsSync is 1) or asynchronous
# (if firstWriteIsSync is 0). The response time of the first
# write is given by 'firstWriteLatency'. The second write is
# a 100 B object which is always written synchronously (its
# response time is given by 'syncWriteLatency'
# Both writes use a 30 B key.
# Generated by 'clusterperf.py writeAsyncSync'
#
# firstWriteIsSync firstObjectSize firstWriteLatency(us) syncWriteLatency(us)
#----------------------------------------------------------------------------
0 100 16.4 15.9
0 1000 20.5 18.6
0 10000 38.8 17.7
0 100000 230.4 21.4
0 1000000 2187.8 24.2
1 100 16.8 18.5
1 1000 19.8 16.7
1 10000 39.3 19.6
1 100000 237.2 19.5
1 1000000 2173.8 24.1
# RAMCloud multiWrite performance for 100 B objects with 30 byte keys
# located on a single master.
# Generated by 'clusterperf.py multiWrite_oneMaster'
#
# Num Objs Num Masters Objs/Master Latency (us) Latency/Obj (us)
#----------------------------------------------------------------------------
1 1 1 16.0 16.0
2 1 2 19.7 9.8
3 1 3 21.8 7.3
4 1 4 24.3 6.1
5 1 5 28.0 5.6
6 1 6 28.8 4.8
7 1 7 29.7 4.2
8 1 8 33.7 4.2
9 1 9 33.7 3.7
10 1 10 35.0 3.5
20 1 20 54.4 2.7
30 1 30 71.6 2.4
40 1 40 91.8 2.3
50 1 50 107.8 2.2
60 1 60 127.3 2.1
70 1 70 145.2 2.1
80 1 80 177.3 2.2
90 1 90 202.6 2.3
100 1 100 207.1 2.1
200 1 200 345.0 1.7
300 1 300 495.6 1.7
400 1 400 657.6 1.6
500 1 500 807.5 1.6
600 1 600 981.0 1.6
700 1 700 1187.9 1.7
800 1 800 1310.9 1.6
900 1 900 1489.5 1.7
1000 1 1000 1645.2 1.6
2000 1 2000 3223.9 1.6
3000 1 3000 4796.4 1.6
4000 1 4000 6428.6 1.6
5000 1 5000 8028.6 1.6
# RAMCloud multiRead performance for 100 B objects with 30 byte keys
# located on a single master.
# Generated by 'clusterperf.py multiRead_oneMaster'
#
# Num Objs Num Masters Objs/Master Latency (us) Latency/Obj (us)
#----------------------------------------------------------------------------
1 1 1 5.6 5.6
2 1 2 7.4 3.7
3 1 3 8.6 2.9
4 1 4 10.1 2.5
5 1 5 10.9 2.2
6 1 6 12.6 2.1
7 1 7 13.3 1.9
8 1 8 13.8 1.7
9 1 9 15.1 1.7
10 1 10 16.1 1.6
20 1 20 24.2 1.2
30 1 30 33.4 1.1
40 1 40 45.9 1.1
50 1 50 50.5 1.0
60 1 60 58.9 1.0
70 1 70 66.3 0.9
80 1 80 73.1 0.9
90 1 90 75.6 0.8
100 1 100 81.6 0.8
200 1 200 126.8 0.6
300 1 300 179.3 0.6
400 1 400 230.0 0.6
500 1 500 286.2 0.6
600 1 600 336.5 0.6
700 1 700 393.3 0.6
800 1 800 427.4 0.5
900 1 900 482.8 0.5
1000 1 1000 532.9 0.5
2000 1 2000 1035.1 0.5
3000 1 3000 1538.6 0.5
4000 1 4000 2033.0 0.5
5000 1 5000 2572.6 0.5
# RAMCloud multiRead performance for 100 B objects with 30 byte keys
# with one object located on each master.
# Generated by 'clusterperf.py multiRead_oneObjectPerMaster'
#
# Num Objs Num Masters Objs/Master Latency (us) Latency/Obj (us)
#----------------------------------------------------------------------------
1 1 1 5.6 5.6
2 2 1 7.9 3.9
3 3 1 8.7 2.9
4 4 1 9.8 2.5
5 5 1 10.4 2.1
6 6 1 11.4 1.9
7 7 1 13.5 1.9
8 8 1 13.7 1.7
9 9 1 16.4 1.8
10 10 1 17.4 1.7
11 11 1 22.0 2.0
12 12 1 25.3 2.1
13 13 1 26.9 2.1
14 14 1 26.4 1.9
15 15 1 33.0 2.2
16 16 1 35.2 2.2
17 17 1 34.1 2.0
18 18 1 38.2 2.1
# RAMCloud multiRead performance for an approximately fixed number
# of 100 B objects with 30 byte keys
# distributed evenly across varying number of masters.
# Generated by 'clusterperf.py multiRead_general'
#
# Num Objs Num Masters Objs/Master Latency (us) Latency/Obj (us)
#----------------------------------------------------------------------------
5000 1 5000 2545.2 0.5
5000 2 2500 2353.3 0.5
4998 3 1666 2541.2 0.5
5000 4 1250 2989.2 0.6
5000 5 1000 2849.7 0.6
4998 6 833 3094.9 0.6
4998 7 714 5590.9 1.1
5000 8 625 5944.0 1.2
4995 9 555 3799.3 0.8
5000 10 500 3791.0 0.8
4994 11 454 3836.3 0.8
4992 12 416 3906.0 0.8
4992 13 384 6163.9 1.2
4998 14 357 4121.6 0.8
4995 15 333 4042.0 0.8
4992 16 312 4756.2 1.0
4998 17 294 6841.1 1.4
4986 18 277 4848.9 1.0
# RAMCloud multiRead performance for an approximately fixed number
# of 100 B objects with 30 byte keys
# distributed evenly across varying number of masters.
# Requests are issued in a random order.
# Generated by 'clusterperf.py multiRead_generalRandom'
#
# Num Objs Num Masters Objs/Master Latency (us) Latency/Obj (us)
#----------------------------------------------------------------------------
5000 1 5000 3196.3 0.6
5000 2 2500 2752.9 0.6
4998 3 1666 2870.1 0.6
5000 4 1250 3056.0 0.6
5000 5 1000 3279.5 0.7
4998 6 833 3249.6 0.7
4998 7 714 3664.7 0.7
5000 8 625 3925.1 0.8
4995 9 555 5874.5 1.2
5000 10 500 3871.4 0.8
4994 11 454 4155.5 0.8
4992 12 416 4411.9 0.9
4992 13 384 6118.0 1.2
4998 14 357 5809.7 1.2
4995 15 333 7381.6 1.5
4992 16 312 4857.4 1.0
4998 17 294 5848.5 1.2
4986 18 277 7730.7 1.6
# Cumulative distribution of time for a single client to read a
# single 100-byte object from a single server. Each line indicates
# that a given fraction of all reads took at most a given time
# to complete.
# Generated by 'clusterperf.py readDist'
#
# Time (usec) Cum. Fraction
#---------------------------
0.00 0.000
4.72 0.000
4.77 0.010
4.77 0.020
4.78 0.030
4.79 0.040
4.79 0.050
4.81 0.060
4.82 0.070
4.83 0.080
4.83 0.090
4.84 0.100
4.84 0.110
4.84 0.120
4.84 0.130
4.84 0.140
4.84 0.150
4.84 0.160
4.84 0.170
4.84 0.180
4.85 0.190
4.85 0.200
4.85 0.210
4.85 0.220
4.85 0.230
4.85 0.240
4.85 0.250
4.85 0.260
4.85 0.270
4.85 0.280
4.85 0.290
4.85 0.300
4.85 0.310
4.85 0.320
4.85 0.330
4.85 0.340
4.85 0.350
4.86 0.360
4.86 0.370
4.86 0.380
4.86 0.390
4.86 0.400
4.86 0.410
4.86 0.420
4.86 0.430
4.86 0.440
4.86 0.450
4.86 0.460
4.86 0.470
4.86 0.480
4.86 0.490
4.86 0.500
4.86 0.510
4.87 0.520
4.87 0.530
4.87 0.540
4.87 0.550
4.87 0.560
4.87 0.570
4.87 0.580
4.87 0.590
4.87 0.600
4.87 0.610
4.87 0.620
4.88 0.630
4.88 0.640
4.88 0.650
4.88 0.660
4.88 0.670
4.88 0.680
4.88 0.690
4.89 0.700
4.89 0.710
4.89 0.720
4.90 0.730
4.90 0.740
4.91 0.750
4.92 0.760
4.93 0.770
4.95 0.780
4.96 0.790
4.97 0.800
5.00 0.810
5.01 0.820
5.03 0.830
5.05 0.840
5.36 0.850
5.52 0.860
5.53 0.870
5.54 0.880
5.54 0.890
5.55 0.900
5.57 0.910
5.70 0.920
5.95 0.930
6.06 0.940
6.08 0.950
6.09 0.960
6.11 0.970
6.20 0.980
6.82 0.990
73.30 0.999
104.17 0.9999
108.52 1.000
# RAMCloud read performance for 100 B objects
# with keys of various lengths.
# Generated by 'clusterperf.py readVaryingKeyLength'
#
# Key Length Latency (us) Bandwidth (MB/s)
#----------------------------------------------------------------------------
1 5.2 0.2
5 4.9 1.0
10 4.9 1.9
15 5.0 2.9
20 5.0 3.8
25 5.0 4.8
30 5.1 5.7
35 5.2 6.4
40 5.1 7.5
45 5.1 8.4
50 5.1 9.3
55 5.3 9.8
60 5.3 10.8
65 5.4 11.5
70 5.3 12.5
75 5.1 13.9
80 5.2 14.8
85 5.4 15.1
90 5.3 16.1
95 5.4 16.6
100 5.4 17.6
200 6.2 30.6
300 6.5 44.3
400 6.9 55.2
500 7.1 67.3
600 7.2 79.0
700 7.2 92.5
800 7.4 103.7
900 7.3 117.4
1000 7.3 130.1
2000 8.0 237.4
3000 8.7 329.0
4000 9.6 396.5
5000 10.4 458.6
6000 11.0 521.2
7000 11.6 574.6
8000 12.3 620.5
9000 13.1 656.4
10000 13.7 696.3
20000 22.1 863.4
30000 29.7 962.6
40000 39.5 964.8
50000 47.9 994.9
60000 55.8 1026.3
# RAMCloud write performance for 100 B objects
# with keys of various lengths.
# Generated by 'clusterperf.py writeVaryingKeyLength'
#
# Key Length Latency (us) Bandwidth (MB/s)
#----------------------------------------------------------------------------
1 16.0 0.1
5 16.0 0.3
10 15.2 0.6
15 15.1 0.9
20 16.2 1.2
25 16.1 1.5
30 16.0 1.8
35 16.1 2.1
40 16.3 2.3
45 15.5 2.8
50 16.2 2.9
55 16.9 3.1
60 16.1 3.5
65 16.2 3.8
70 16.5 4.1
75 16.8 4.3
80 17.0 4.5
85 16.6 4.9
90 17.1 5.0
95 17.5 5.2
100 16.9 5.6
200 17.4 10.9
300 18.1 15.8
400 18.7 20.4
500 19.5 24.4
600 20.0 28.6
700 21.6 30.9
800 21.8 35.1
900 22.4 38.4
1000 24.2 39.4
2000 27.7 68.8
3000 32.7 87.4
4000 36.1 105.6
5000 38.0 125.4
6000 42.6 134.2
7000 46.2 144.5
8000 49.0 155.7
9000 55.1 155.8
10000 59.1 161.4
20000 101.3 188.3
30000 154.4 185.4
40000 209.3 182.3
50000 248.8 191.7
60000 301.3 189.9
# RAMCloud read performance as a function of load (1 or more
# clients all reading a single 100-byte object with 30-byte key
# repeatedly).
# Generated by 'clusterperf.py readLoaded'
#
# numClients readLatency(us) throughput(total kreads/sec)
#----------------------------------------------------------
1 5.2 192
2 5.5 364
3 6.0 497
4 6.6 602
5 8.1 619
6 9.4 638
7 11.3 617
8 12.2 654
9 14.7 613
10 15.2 659
11 17.0 649
12 18.5 648
13 20.6 630
14 21.7 646
15 24.7 608
16 25.3 632
17 31.7 537
18 34.3 525
19 37.9 501
20 33.8 591
# RAMCloud read performance when 1 or more clients read
# 100-byte objects with 30-byte keys chosen at random from
# 10 servers.
# Generated by 'clusterperf.py readRandom'
#
# numClients throughput(total kreads/sec) slowest(ms) reads > 10us
#--------------------------------------------------------------------
1 181 0.44 0.6%
2 342 0.61 0.5%
3 511 0.56 0.7%
4 665 0.68 0.8%
5 823 0.59 0.8%
6 980 0.60 0.8%
7 1096 0.63 1.4%
8 1187 0.75 2.1%
9 1370 0.58 1.7%
10 1468 0.57 2.2%
11 1629 3.09 1.4%
12 1788 0.63 1.7%
13 1839 0.65 2.5%
14 2074 0.61 1.6%
15 2157 1.79 2.0%
16 2324 0.56 2.4%
17 2492 0.64 2.1%
18 2698 0.51 2.2%
19 2762 0.52 2.9%
20 2887 0.60 3.2%
21 2995 0.61 3.7%
22 3027 0.55 4.8%
23 3228 0.90 5.0%
24 3302 0.61 5.6%
25 3151 7.33 6.2%
26 3232 6.46 7.4%
27 3113 10.09 8.2%
28 2836 21.78 8.8%
29 3533 3.67 10.8%
30 2701 14.82 12.0%
31 3641 5.01 13.2%
32 3690 2.43 16.7%
33 3422 12.33 19.5%
34 3554 12.53 19.6%
35 3891 2.46 22.4%
36 3894 2.45 23.7%
37 3933 2.61 24.9%
38 4035 6.08 23.0%
39 2837 41.37 25.1%
40 2709 25.48 29.4%
41 3844 12.33 29.8%
42 4227 10.01 31.3%
43 4275 10.01 32.1%
44 4173 2.54 34.7%
45 4160 2.59 33.5%
46 4196 2.66 35.2%
47 4196 2.71 36.5%
48 3695 12.43 35.5%
49 4245 2.73 38.5%
50 4265 2.92 37.0%
**** server4.rc07.log:
1376433692.862275100 src/CoordinatorClient.cc:563 in VerifyMembershipRpc default WARNING[24409:3]: verifying cluster membership for 8.0