tobert · January 31, 2022 15:22 · Apr 1, 2015 · Apr 1, 2015 · Apr 1, 2015 · Apr 1, 2015
diff --git a/0000README.md b/0000README.md
@@ -42,4 +42,11 @@ cassandra-stress \
 In the course of tuning for these workloads I've experimented on and off with the following settings. I'll
 need to break down which is providing what benefit later.
 
-`-XX:+AlwaysPreTouch -XX:+ResizeTLAB -XX:-UseBiasedLocking`
+`-XX:+AlwaysPreTouch -XX:+ResizeTLAB -XX:-UseBiasedLocking`
+
+## stress logs
+
+(http://tobert.org/logs/cms-2Gnew-with-other.log.gz)
+(http://tobert.org/logs/cms-2Gnew-without-other.log.gz)
+(http://tobert.org/logs/g1-1000ms-with-other.log.gz)
+(http://tobert.org/logs/g1-1000ms-without-other.log.gz)
diff --git a/0000README.md b/0000README.md
@@ -12,10 +12,11 @@ G1 with "other" settings (see below) can beat CMS on latency and throughput on a
 
 | Settings      | Op/s    | latency mean | latency median | latency p95 | latency p99 | latency p99.9 | latency max |
 | ------------- | ------- | ------------ | -------------- | ----------- | ----------- | ------------- | ----------- |
-| G1+other    | 74379   | 2.7          | 1.7            | 4.9         | 7.5         |  21.7         | 2901.2      |
-| G1-other    | 72116   | 2.8          | 1.7            | 5.0         | 7.7         |  24.3         | 1088.2      |
-| CMS+other   | 65681   | 3.0          | 1.5            | 4.7         | 7.7         | 276.9         | 1848.0      |
-| CMS-other   | 62196   | 3.2          | 1.6            | 5.1         | 8.3         | 185.8         | 1615.3      |
+| G1+other      | 74379   | 2.7          | 1.7            | 4.9         | 7.5         |  21.7         | 2901.2      |
+| G1-other      | 72116   | 2.8          | 1.7            | 5.0         | 7.7         |  24.3         | 1088.2      |
+| CMS+other     | 65681   | 3.0          | 1.5            | 4.7         | 7.7         | 276.9         | 1848.0      |
+| CMS-other     | 62196   | 3.2          | 1.6            | 5.1         | 8.3         | 185.8         | 1615.3      |
+| CMS-other     | 64973   | 3.1          | 1.5            | 4.7         | 7.4         | 127.6         | 1645.7      |
 
 ## Hardware / Software
 

diff --git a/030cms-2Gnew-without-other.txt b/030cms-2Gnew-without-other.txt
@@ -1,5 +1,20 @@
 -XX:+UseParNewGC -XX:+UseConcMarkSweepGC
 
+Run 2: I wanted to add links to the stress logs and lost the one from above so I ran again and got better numbers:
+
+op rate                   : 64973 [WRITE:64973]
+partition rate            : 64973 [WRITE:64973]
+row rate                  : 64973 [WRITE:64973]
+latency mean              : 3.1 [WRITE:3.1]
+latency median            : 1.5 [WRITE:1.5]
+latency 95th percentile   : 4.7 [WRITE:4.7]
+latency 99th percentile   : 7.4 [WRITE:7.4]
+latency 99.9th percentile : 127.6 [WRITE:127.6]
+latency max               : 1645.7 [WRITE:1645.7]
+Total partitions          : 50000000 [WRITE:50000000]
+
+Run 1:
+
 op rate                   : 62196 [WRITE:62196]
 partition rate            : 62196 [WRITE:62196]
 row rate                  : 62196 [WRITE:62196]

diff --git a/010g1-1000ms-with-other.txt b/010g1-1000ms-with-other.txt
@@ -1,6 +1,5 @@
 -XX:+UseG1GC -XX:InitiatingHeapOccupancyPercent=60 -XX:MaxGCPauseMillis=1000 -XX:+AlwaysPreTouch -XX:+ResizeTLAB -XX:-UseBiasedLocking 
 
-Results:
 op rate                   : 74379 [WRITE:74379]
 partition rate            : 74379 [WRITE:74379]
 row rate                  : 74379 [WRITE:74379]

diff --git a/020g1-1000ms-without-other.txt b/020g1-1000ms-without-other.txt
@@ -1,3 +1,5 @@
+-XX:+UseG1GC -XX:InitiatingHeapOccupancyPercent=60 -XX:MaxGCPauseMillis=1000
+
 op rate                   : 72116 [WRITE:72116]
 partition rate            : 72116 [WRITE:72116]
 row rate                  : 72116 [WRITE:72116]

diff --git a/030cms-2Gnew-without-other.txt b/030cms-2Gnew-without-other.txt
@@ -1,6 +1,5 @@
-Stock CMS settings from Cassandra 2.0
+-XX:+UseParNewGC -XX:+UseConcMarkSweepGC
 
-Results:
 op rate                   : 62196 [WRITE:62196]
 partition rate            : 62196 [WRITE:62196]
 row rate                  : 62196 [WRITE:62196]

diff --git a/040cms-2Gnew-with-other.txt b/040cms-2Gnew-with-other.txt
@@ -1,6 +1,5 @@
-Stock settings with -XX:+AlwaysPreTouch -XX:+ResizeTLAB -XX:-UseBiasedLocking 
+-XX:+UseParNewGC -XX:+UseConcMarkSweepGC -XX:+AlwaysPreTouch -XX:+ResizeTLAB -XX:-UseBiasedLocking 
 
-Results:
 op rate                   : 65681 [WRITE:65681]
 partition rate            : 65681 [WRITE:65681]
 row rate                  : 65681 [WRITE:65681]

diff --git a/00g1gc-1000ms.txt → 010g1-1000ms-with-other.txt b/00g1gc-1000ms.txt → 010g1-1000ms-with-other.txt
diff --git a/05g1gc-1000ms-without-others.txt → 020g1-1000ms-without-other.txt b/05g1gc-1000ms-without-others.txt → 020g1-1000ms-without-other.txt
diff --git a/10cms-stock-2g-new.txt → 030cms-2Gnew-without-other.txt b/10cms-stock-2g-new.txt → 030cms-2Gnew-without-other.txt
diff --git a/15cms-other.txt → 040cms-2Gnew-with-other.txt b/15cms-other.txt → 040cms-2Gnew-with-other.txt
diff --git a/20g1-sprok.yaml → 110g1-1000ms-with-other.yaml b/20g1-sprok.yaml → 110g1-1000ms-with-other.yaml
diff --git a/20g1-sprok-without-others.yaml → 120g1-1000ms-without-other.yaml b/20g1-sprok-without-others.yaml → 120g1-1000ms-without-other.yaml
diff --git a/40cms-with-other-opts.yaml → 130cms-2Gnew-with-other.yaml b/40cms-with-other-opts.yaml → 130cms-2Gnew-with-other.yaml
diff --git a/30cms-sprok.yaml → 140cms-2Gnew-without-other.yaml b/30cms-sprok.yaml → 140cms-2Gnew-without-other.yaml
diff --git a/0000README.md b/0000README.md
@@ -4,6 +4,8 @@ Many people have asked me about using the G1 garbage collector with Cassandra. S
 of my customers are running 2.0 in production the test is with 2.0 for now. Once I script it
 up I'll re-run the numbers with 2.1.
 
+I also need to re-test with Java 7 and the Oracle JDKs.
+
 ## TL;DR
 
 G1 with "other" settings (see below) can beat CMS on latency and throughput on anemic CPUs if you set -XX:MaxGCPauseMillis=1000.

diff --git a/0000README.md b/0000README.md
@@ -10,10 +10,10 @@ G1 with "other" settings (see below) can beat CMS on latency and throughput on a
 
 | Settings      | Op/s    | latency mean | latency median | latency p95 | latency p99 | latency p99.9 | latency max |
 | ------------- | ------- | ------------ | -------------- | ----------- | ----------- | ------------- | ----------- |
-| G1 + other    | 74379   | 2.7          | 1.7            | 4.9         | 7.5         |  21.7         | 2901.2      |
-| G1 - other    | 72116   | 2.8          | 1.7            | 5.0         | 7.7         |  24.3         | 1088.2      |
-| CMS + other   | 65681   | 3.0          | 1.5            | 4.7         | 7.7         | 276.9         | 1848.0      |
-| CMS - other   | 62196   | 3.2          | 1.6            | 5.1         | 8.3         | 185.8         | 1615.3      |
+| G1+other    | 74379   | 2.7          | 1.7            | 4.9         | 7.5         |  21.7         | 2901.2      |
+| G1-other    | 72116   | 2.8          | 1.7            | 5.0         | 7.7         |  24.3         | 1088.2      |
+| CMS+other   | 65681   | 3.0          | 1.5            | 4.7         | 7.7         | 276.9         | 1848.0      |
+| CMS-other   | 62196   | 3.2          | 1.6            | 5.1         | 8.3         | 185.8         | 1615.3      |
 
 ## Hardware / Software
 
@@ -33,3 +33,10 @@ cassandra-stress \
                 -mode native cql3 \
                 -node 192.168.10.12
 ```
+
+## Other Settings
+
+In the course of tuning for these workloads I've experimented on and off with the following settings. I'll
+need to break down which is providing what benefit later.
+
+`-XX:+AlwaysPreTouch -XX:+ResizeTLAB -XX:-UseBiasedLocking`
diff --git a/0000README.md b/0000README.md
@@ -8,12 +8,12 @@ up I'll re-run the numbers with 2.1.
 
 G1 with "other" settings (see below) can beat CMS on latency and throughput on anemic CPUs if you set -XX:MaxGCPauseMillis=1000.
 
-| Settings             | Op/s    | latency mean | latency median | latency p95 | latency p99 | latency p99.9 | latency max |
-| -------------------- | ------- | ------------ | -------------- | ----------- | ----------- | ------------- | ----------- |
-| G1@1000ms + other    | 74379   | 2.7          | 1.7            | 4.9         | 7.5         |  21.7         | 2901.2      |
-| G1@1000ms - other    | 72116   | 2.8          | 1.7            | 5.0         | 7.7         |  24.3         | 1088.2      |
-| CMS + other          | 65681   | 3.0          | 1.5            | 4.7         | 7.7         | 276.9         | 1848.0      |
-| CMS - other          | 62196   | 3.2          | 1.6            | 5.1         | 8.3         | 185.8         | 1615.3      |
+| Settings      | Op/s    | latency mean | latency median | latency p95 | latency p99 | latency p99.9 | latency max |
+| ------------- | ------- | ------------ | -------------- | ----------- | ----------- | ------------- | ----------- |
+| G1 + other    | 74379   | 2.7          | 1.7            | 4.9         | 7.5         |  21.7         | 2901.2      |
+| G1 - other    | 72116   | 2.8          | 1.7            | 5.0         | 7.7         |  24.3         | 1088.2      |
+| CMS + other   | 65681   | 3.0          | 1.5            | 4.7         | 7.7         | 276.9         | 1848.0      |
+| CMS - other   | 62196   | 3.2          | 1.6            | 5.1         | 8.3         | 185.8         | 1615.3      |
 
 ## Hardware / Software
 

diff --git a/0000README.md b/0000README.md
@@ -14,7 +14,6 @@ G1 with "other" settings (see below) can beat CMS on latency and throughput on a
 | G1@1000ms - other    | 72116   | 2.8          | 1.7            | 5.0         | 7.7         |  24.3         | 1088.2      |
 | CMS + other          | 65681   | 3.0          | 1.5            | 4.7         | 7.7         | 276.9         | 1848.0      |
 | CMS - other          | 62196   | 3.2          | 1.6            | 5.1         | 8.3         | 185.8         | 1615.3      |
-| -------------------- | ------- | ------------ | -------------- | ----------- | ----------- | ------------- | ----------- |
 
 ## Hardware / Software
 

diff --git a/0000README.md b/0000README.md
@@ -8,7 +8,6 @@ up I'll re-run the numbers with 2.1.
 
 G1 with "other" settings (see below) can beat CMS on latency and throughput on anemic CPUs if you set -XX:MaxGCPauseMillis=1000.
 
-| -------------------- | ------- | ------------ | -------------- | ----------- | ----------- | ------------- | ----------- |
 | Settings             | Op/s    | latency mean | latency median | latency p95 | latency p99 | latency p99.9 | latency max |
 | -------------------- | ------- | ------------ | -------------- | ----------- | ----------- | ------------- | ----------- |
 | G1@1000ms + other    | 74379   | 2.7          | 1.7            | 4.9         | 7.5         |  21.7         | 2901.2      |

diff --git a/0000README.md b/0000README.md
@@ -4,6 +4,19 @@ Many people have asked me about using the G1 garbage collector with Cassandra. S
 of my customers are running 2.0 in production the test is with 2.0 for now. Once I script it
 up I'll re-run the numbers with 2.1.
 
+## TL;DR
+
+G1 with "other" settings (see below) can beat CMS on latency and throughput on anemic CPUs if you set -XX:MaxGCPauseMillis=1000.
+
+| -------------------- | ------- | ------------ | -------------- | ----------- | ----------- | ------------- | ----------- |
+| Settings             | Op/s    | latency mean | latency median | latency p95 | latency p99 | latency p99.9 | latency max |
+| -------------------- | ------- | ------------ | -------------- | ----------- | ----------- | ------------- | ----------- |
+| G1@1000ms + other    | 74379   | 2.7          | 1.7            | 4.9         | 7.5         |  21.7         | 2901.2      |
+| G1@1000ms - other    | 72116   | 2.8          | 1.7            | 5.0         | 7.7         |  24.3         | 1088.2      |
+| CMS + other          | 65681   | 3.0          | 1.5            | 4.7         | 7.7         | 276.9         | 1848.0      |
+| CMS - other          | 62196   | 3.2          | 1.6            | 5.1         | 8.3         | 185.8         | 1615.3      |
+| -------------------- | ------- | ------------ | -------------- | ----------- | ----------- | ------------- | ----------- |
+
 ## Hardware / Software
 
 * 6x Intel NUC 16GB RAM, quad-core @ 1.3Ghz, SSD

diff --git a/05g1gc-1000ms-without-others.txt b/05g1gc-1000ms-without-others.txt
@@ -0,0 +1,18 @@
+op rate                   : 72116 [WRITE:72116]
+partition rate            : 72116 [WRITE:72116]
+row rate                  : 72116 [WRITE:72116]
+latency mean              : 2.8 [WRITE:2.8]
+latency median            : 1.7 [WRITE:1.7]
+latency 95th percentile   : 5.0 [WRITE:5.0]
+latency 99th percentile   : 7.7 [WRITE:7.7]
+latency 99.9th percentile : 24.3 [WRITE:24.3]
+latency max               : 1088.2 [WRITE:1088.2]
+Total partitions          : 50000000 [WRITE:50000000]
+Total errors              : 0 [WRITE:0]
+total gc count            : 0
+total gc mb               : 0
+total gc time (s)         : 0
+avg gc time(ms)           : NaN
+stdev gc time(ms)         : 0
+Total operation time      : 00:11:33
+END
diff --git a/10cms-stock-2g-new.txt b/10cms-stock-2g-new.txt
@@ -1,4 +1,4 @@
-Stock settings from Cassandra 2.0
+Stock CMS settings from Cassandra 2.0
 
 Results:
 op rate                   : 62196 [WRITE:62196]

diff --git a/20g1-sprok-without-others.yaml b/20g1-sprok-without-others.yaml
@@ -0,0 +1,44 @@
+# G1GC without unrelated settings
+chdir: /data
+stdin: /dev/null
+stdout: /data/log/console.log
+stderr: /data/log/console.log
+uid: 1337
+gid: 1337
+argv:
+  - /usr/bin/java
+  - -ea
+  - -javaagent:{{ glob "" "/opt/cassandra/lib/jamm-*.jar" }}
+  - -Xmx8G
+  - -Xms8G
+  - -Xss256k
+  - -XX:+UseG1GC                          # use garbage-first collection
+  - -XX:InitiatingHeapOccupancyPercent=60 # don't do mixed mode until the heap is this full
+  - -XX:MaxGCPauseMillis=1000             # default is 200ms
+  - -XX:+UseTLAB                          # thread local allocation blocks
+  - -XX:StringTableSize=1000003
+  - -XX:CompileCommandFile=/data/conf/hotspot_compiler
+  - -Djava.net.preferIPv4Stack=true
+  - -Dcom.sun.management.jmxremote.port=7199
+  - -Dcom.sun.management.jmxremote.rmi.port=7199
+  - -Dcom.sun.management.jmxremote.ssl=false
+  - -Dcom.sun.management.jmxremote.authenticate=false
+  - -Dlog4j.defaultInitOverride=true
+  - -Dlog4j.configuration=log4j-server.properties # cassandra <= 2.0
+  - -Dlogback.configurationFile=logback.xml       # cassandra >= 2.1
+  - -Dcassandra.logdir=/data/log
+  - -Dcassandra-foreground=yes
+  - -XX:+HeapDumpOnOutOfMemoryError
+  - -XX:+PrintGCDetails
+  - -XX:+PrintAdaptiveSizePolicy
+  - -XX:+PrintGCApplicationStoppedTime
+  - -XX:+PrintPromotionFailure
+  - -Xloggc:/data/log/gc.log
+  - -XX:+UseGCLogFileRotation
+  - -XX:NumberOfGCLogFiles=10
+  - -XX:GCLogFileSize=10M
+  - -cp
+  - {{ glob ":" "/data/conf" "/data/lib" "/opt/cassandra/lib/*.jar" }}
+  - org.apache.cassandra.service.CassandraDaemon
+{{ range .ExtraArgs }}  - {{ . }}
+{{ end }}
diff --git a/0000README.md b/0000README.md
@@ -0,0 +1,24 @@
+## G1GC v.s. CMS for Cassandra 2.0
+
+Many people have asked me about using the G1 garbage collector with Cassandra. Since most
+of my customers are running 2.0 in production the test is with 2.0 for now. Once I script it
+up I'll re-run the numbers with 2.1.
+
+## Hardware / Software
+
+* 6x Intel NUC 16GB RAM, quad-core @ 1.3Ghz, SSD
+* Zulu OpenJDK 8
+* Ubuntu 14.04 LTS Docker image (tobert/cassandra:2.0.13 w/ zulu openjdk)
+* CoreOS 607.0 kernel 3.18.6
+* /data is on an Intel mSATA SSD using ext4
+
+## cassandra-stress
+
+```
+cassandra-stress \
+                write \
+                n=50000000 \
+                cl=ONE \
+                -mode native cql3 \
+                -node 192.168.10.12
+```
diff --git a/00g1gc-1000ms.txt b/00g1gc-1000ms.txt
@@ -1,3 +1,5 @@
+-XX:+UseG1GC -XX:InitiatingHeapOccupancyPercent=60 -XX:MaxGCPauseMillis=1000 -XX:+AlwaysPreTouch -XX:+ResizeTLAB -XX:-UseBiasedLocking 
+
 Results:
 op rate                   : 74379 [WRITE:74379]
 partition rate            : 74379 [WRITE:74379]

diff --git a/10cms-stock-2g-new.txt b/10cms-stock-2g-new.txt
@@ -1,3 +1,5 @@
+Stock settings from Cassandra 2.0
+
 Results:
 op rate                   : 62196 [WRITE:62196]
 partition rate            : 62196 [WRITE:62196]

diff --git a/15cms-other.txt b/15cms-other.txt
@@ -1,3 +1,5 @@
+Stock settings with -XX:+AlwaysPreTouch -XX:+ResizeTLAB -XX:-UseBiasedLocking 
+
 Results:
 op rate                   : 65681 [WRITE:65681]
 partition rate            : 65681 [WRITE:65681]

diff --git a/15cms-other.txt b/15cms-other.txt
@@ -0,0 +1,18 @@
+Results:
+op rate                   : 65681 [WRITE:65681]
+partition rate            : 65681 [WRITE:65681]
+row rate                  : 65681 [WRITE:65681]
+latency mean              : 3.0 [WRITE:3.0]
+latency median            : 1.5 [WRITE:1.5]
+latency 95th percentile   : 4.7 [WRITE:4.7]
+latency 99th percentile   : 7.7 [WRITE:7.7]
+latency 99.9th percentile : 276.8 [WRITE:276.8]
+latency max               : 1848.0 [WRITE:1848.0]
+Total partitions          : 50000000 [WRITE:50000000]
+Total errors              : 0 [WRITE:0]
+total gc count            : 0
+total gc mb               : 0
+total gc time (s)         : 0
+avg gc time(ms)           : NaN
+stdev gc time(ms)         : 0
+Total operation time      : 00:12:41
diff --git a/40cms-with-other-opts.txt → 40cms-with-other-opts.yaml b/40cms-with-other-opts.txt → 40cms-with-other-opts.yaml
diff --git a/01cms-stock-2g-new.txt → 10cms-stock-2g-new.txt b/01cms-stock-2g-new.txt → 10cms-stock-2g-new.txt
diff --git a/02g1-sprok.yaml → 20g1-sprok.yaml b/02g1-sprok.yaml → 20g1-sprok.yaml
@@ -1,3 +1,4 @@
+# G1GC settings for Cassandra 2.0 (or 2.1)
 chdir: /data
 stdin: /dev/null
 stdout: /data/log/console.log

diff --git a/03cms-sprok.yaml → 30cms-sprok.yaml b/03cms-sprok.yaml → 30cms-sprok.yaml
@@ -1,3 +1,4 @@
+# default Cassandra 2.0 settings
 chdir: {{ .DataDir }}
 stdin: /dev/null
 stdout: /data/log/console.log

diff --git a/40cms-with-other-opts.txt b/40cms-with-other-opts.txt
@@ -0,0 +1,50 @@
+# default Cassandra 2.0 CMS settings with the 3 other settings from the
+# G1 settings that might make a difference
+chdir: {{ .DataDir }}
+stdin: /dev/null
+stdout: /data/log/console.log
+stderr: /data/log/console.log
+uid: 1337
+gid: 1337
+argv:
+  - /usr/bin/java
+  - -ea
+  - -javaagent:{{ glob "" "/opt/cassandra/lib/jamm-*.jar" }}
+  - -XX:+CMSClassUnloadingEnabled
+  - -XX:+UseThreadPriorities
+  - -XX:ThreadPriorityPolicy=42
+  - -Xmx8G
+  - -Xms8G
+  - -Xmn2G
+  - -XX:+HeapDumpOnOutOfMemoryError
+  - -Xss256k
+  - -XX:+AlwaysPreTouch                   # allocate and zero (force fault) heap memory on startup
+  - -XX:+UseTLAB                          # thread local allocation blocks
+  - -XX:+ResizeTLAB                       # auto-optimize https://blogs.oracle.com/jonthecollector/entry/the_real_thing
+  - -XX:-UseBiasedLocking                 # disable biased locking for cassandra
+  - -XX:StringTableSize=1000003
+  - -XX:+UseParNewGC
+  - -XX:+UseConcMarkSweepGC
+  - -XX:+CMSParallelRemarkEnabled
+  - -XX:SurvivorRatio=8
+  - -XX:MaxTenuringThreshold=1
+  - -XX:CMSInitiatingOccupancyFraction=75
+  - -XX:+UseCMSInitiatingOccupancyOnly
+  - -XX:CompileCommandFile={{ .ConfDir }}/hotspot_compiler
+  - -XX:+UseCondCardMark
+  - -XX:+CMSParallelInitialMarkEnabled
+  - -XX:+CMSEdenChunksRecordAlways
+  - -Djava.net.preferIPv4Stack=true
+  - -Dcom.sun.management.jmxremote.port={{ .JmxPort }}
+  - -Dcom.sun.management.jmxremote.rmi.port={{ .JmxPort }}
+  - -Dcom.sun.management.jmxremote.ssl=false
+  - -Dcom.sun.management.jmxremote.authenticate=false
+  - -Dlog4j.configuration=log4j-server.properties
+  - -Dlog4j.defaultInitOverride=true
+  - -Dcassandra.logdir={{ .LogDir }}
+  - -Dcassandra-foreground=yes
+  - -cp
+  - {{ glob ":" .ConfDir .LibDir "/opt/cassandra/lib/*.jar" }}
+  - org.apache.cassandra.service.CassandraDaemon
+{{ range .ExtraArgs }}  - {{ . }}
+{{ end }}
diff --git a/g1gc-1000ms.txt → 00g1gc-1000ms.txt b/g1gc-1000ms.txt → 00g1gc-1000ms.txt
diff --git a/cms-stock-2g-new.txt → 01cms-stock-2g-new.txt b/cms-stock-2g-new.txt → 01cms-stock-2g-new.txt
diff --git a/g1-sprok.yaml → 02g1-sprok.yaml b/g1-sprok.yaml → 02g1-sprok.yaml
diff --git a/cms-sprok.yaml → 03cms-sprok.yaml b/cms-sprok.yaml → 03cms-sprok.yaml
diff --git a/cms-sprok.yaml b/cms-sprok.yaml
@@ -0,0 +1,45 @@
+chdir: {{ .DataDir }}
+stdin: /dev/null
+stdout: /data/log/console.log
+stderr: /data/log/console.log
+uid: 1337
+gid: 1337
+argv:
+  - /usr/bin/java
+  - -ea
+  - -javaagent:{{ glob "" "/opt/cassandra/lib/jamm-*.jar" }}
+  - -XX:+CMSClassUnloadingEnabled
+  - -XX:+UseThreadPriorities
+  - -XX:ThreadPriorityPolicy=42
+  - -Xmx8G
+  - -Xms8G
+  - -Xmn2G
+  - -XX:+HeapDumpOnOutOfMemoryError
+  - -Xss256k
+  - -XX:StringTableSize=1000003
+  - -XX:+UseParNewGC
+  - -XX:+UseConcMarkSweepGC
+  - -XX:+CMSParallelRemarkEnabled
+  - -XX:SurvivorRatio=8
+  - -XX:MaxTenuringThreshold=1
+  - -XX:CMSInitiatingOccupancyFraction=75
+  - -XX:+UseCMSInitiatingOccupancyOnly
+  - -XX:+UseTLAB
+  - -XX:CompileCommandFile={{ .ConfDir }}/hotspot_compiler
+  - -XX:+UseCondCardMark
+  - -XX:+CMSParallelInitialMarkEnabled
+  - -XX:+CMSEdenChunksRecordAlways
+  - -Djava.net.preferIPv4Stack=true
+  - -Dcom.sun.management.jmxremote.port={{ .JmxPort }}
+  - -Dcom.sun.management.jmxremote.rmi.port={{ .JmxPort }}
+  - -Dcom.sun.management.jmxremote.ssl=false
+  - -Dcom.sun.management.jmxremote.authenticate=false
+  - -Dlog4j.configuration=log4j-server.properties
+  - -Dlog4j.defaultInitOverride=true
+  - -Dcassandra.logdir={{ .LogDir }}
+  - -Dcassandra-foreground=yes
+  - -cp
+  - {{ glob ":" .ConfDir .LibDir "/opt/cassandra/lib/*.jar" }}
+  - org.apache.cassandra.service.CassandraDaemon
+{{ range .ExtraArgs }}  - {{ . }}
+{{ end }}
diff --git a/g1-sprok.yaml b/g1-sprok.yaml
@@ -0,0 +1,46 @@
+chdir: /data
+stdin: /dev/null
+stdout: /data/log/console.log
+stderr: /data/log/console.log
+uid: 1337
+gid: 1337
+argv:
+  - /usr/bin/java
+  - -ea
+  - -javaagent:{{ glob "" "/opt/cassandra/lib/jamm-*.jar" }}
+  - -Xmx8G
+  - -Xms8G
+  - -Xss256k
+  - -XX:+UseG1GC                          # use garbage-first collection
+  - -XX:InitiatingHeapOccupancyPercent=60 # don't do mixed mode until the heap is this full
+  - -XX:MaxGCPauseMillis=1000             # default is 200ms
+  - -XX:+AlwaysPreTouch                   # allocate and zero (force fault) heap memory on startup
+  - -XX:+UseTLAB                          # thread local allocation blocks
+  - -XX:+ResizeTLAB                       # auto-optimize TLAB size https://blogs.oracle.com/jonthecollector/entry/the_real_thing
+  - -XX:-UseBiasedLocking                 # disable biased locking for cassandra
+  - -XX:StringTableSize=1000003
+  - -XX:CompileCommandFile=/data/conf/hotspot_compiler
+  - -Djava.net.preferIPv4Stack=true
+  - -Dcom.sun.management.jmxremote.port=7199
+  - -Dcom.sun.management.jmxremote.rmi.port=7199
+  - -Dcom.sun.management.jmxremote.ssl=false
+  - -Dcom.sun.management.jmxremote.authenticate=false
+  - -Dlog4j.defaultInitOverride=true
+  - -Dlog4j.configuration=log4j-server.properties # cassandra <= 2.0
+  - -Dlogback.configurationFile=logback.xml       # cassandra >= 2.1
+  - -Dcassandra.logdir=/data/log
+  - -Dcassandra-foreground=yes
+  - -XX:+HeapDumpOnOutOfMemoryError
+  - -XX:+PrintGCDetails
+  - -XX:+PrintAdaptiveSizePolicy
+  - -XX:+PrintGCApplicationStoppedTime
+  - -XX:+PrintPromotionFailure
+  - -Xloggc:/data/log/gc.log
+  - -XX:+UseGCLogFileRotation
+  - -XX:NumberOfGCLogFiles=10
+  - -XX:GCLogFileSize=10M
+  - -cp
+  - {{ glob ":" "/data/conf" "/data/lib" "/opt/cassandra/lib/*.jar" }}
+  - org.apache.cassandra.service.CassandraDaemon
+{{ range .ExtraArgs }}  - {{ . }}
+{{ end }}
diff --git a/cms-stock-2g-new.txt b/cms-stock-2g-new.txt
@@ -0,0 +1,18 @@
+Results:
+op rate                   : 62196 [WRITE:62196]
+partition rate            : 62196 [WRITE:62196]
+row rate                  : 62196 [WRITE:62196]
+latency mean              : 3.2 [WRITE:3.2]
+latency median            : 1.6 [WRITE:1.6]
+latency 95th percentile   : 5.1 [WRITE:5.1]
+latency 99th percentile   : 8.3 [WRITE:8.3]
+latency 99.9th percentile : 185.8 [WRITE:185.8]
+latency max               : 1615.3 [WRITE:1615.3]
+Total partitions          : 50000000 [WRITE:50000000]
+Total errors              : 0 [WRITE:0]
+total gc count            : 0
+total gc mb               : 0
+total gc time (s)         : 0
+avg gc time(ms)           : NaN
+stdev gc time(ms)         : 0
+Total operation time      : 00:13:23
diff --git a/g1gc-1000ms.txt b/g1gc-1000ms.txt
@@ -0,0 +1,18 @@
+Results:
+op rate                   : 74379 [WRITE:74379]
+partition rate            : 74379 [WRITE:74379]
+row rate                  : 74379 [WRITE:74379]
+latency mean              : 2.7 [WRITE:2.7]
+latency median            : 1.7 [WRITE:1.7]
+latency 95th percentile   : 4.9 [WRITE:4.9]
+latency 99th percentile   : 7.5 [WRITE:7.5]
+latency 99.9th percentile : 21.7 [WRITE:21.7]
+latency max               : 2901.2 [WRITE:2901.2]
+Total partitions          : 50000000 [WRITE:50000000]
+Total errors              : 0 [WRITE:0]
+total gc count            : 0
+total gc mb               : 0
+total gc time (s)         : 0
+avg gc time(ms)           : NaN
+stdev gc time(ms)         : 0
+Total operation time      : 00:11:12
No results found