JVM Instances |
jvm_Ctr_1(1), jvm_Backend_1(8), jvm_TxInjector_1(8)
|
OS Image Description |
os_1
|
Tuning |
- tuned-adm profile=throughput-performance
- ulimit -n 65535
- echo 16 > /proc/sys/kernel/sched_nr_migrate
- echo 16000000 > /proc/sys/kernel/sched_latency_ns
- echo 30000000 > /proc/sys/kernel/sched_min_granularity_ns
- echo 990000 > /proc/sys/kernel/sched_rt_runtime_us
- echo 1000 > /proc/sys/kernel/sched_migration_cost_ns
- echo 10000 > /proc/sys/vm/dirty_expire_centisecs
- Added 'processor.max_cstate=1 intel_idle.max_cstate=0' to GRUB_CMDLINE_LINUX_DEFAULT
|
Notes |
None
|
Parts of Benchmark |
Controller
|
JVM Instance Description |
jvm_1
|
Command Line |
-server -Xms2g -Xmx2g -Xmn1536m -XX:+UseParallelOldGC
|
Tuning |
None
|
Notes |
Used numactl to interleave memory on all CPUs
|
Parts of Benchmark |
Backend
|
JVM Instance Description |
jvm_1
|
Command Line |
-showversion -server -XX:AllocatePrefetchInstr=2 -XX:+UseParallelOldGC -XX:ParallelGCThreads=28 -XX:LargePageSizeInBytes=2m -XX:+AggressiveOpts -XX:-UseAdaptiveSizePolicy -XX:+AlwaysPreTouch -XX:-UseBiasedLocking -XX:+UseLargePages -Xms30g -Xmx30g -Xmn27g -XX:SurvivorRatio=28 -XX:TargetSurvivorRatio=95 -XX:MaxTenuringThreshold=15 -Xnoclassgc -XX:+UseRTMLocking -XX:InlineSmallCode=10k -XX:MaxGCPauseMillis=300 -XX:+UseRTMDeopt -XX:UseAVX=0 -XX:LoopUnrollLimit=100 -XX:AdaptiveSizeMajorGCDecayTimeScale=12 -XX:AdaptiveSizeDecrementScaleFactor=2
|
Tuning |
None
|
Notes |
Used numactl to affinitize each Backend JVM to physical cores in a NUMA node. - numactl --physcpubind=0-13,112-125 --membind=0
- numactl --physcpubind=14-27,126-139 --membind=1
- numactl --physcpubind=28-41,140-153 --membind=2
- numactl --physcpubind=42-55,154-167 --membind=3
- numactl --physcpubind=56-69,168-181 --membind=4
- numactl --physcpubind=70-83,182-195 --membind=5
- numactl --physcpubind=84-97,196-209 --membind=6
- numactl --physcpubind=98-111,210-223 --membind=7
|
Parts of Benchmark |
TxInjector
|
JVM Instance Description |
jvm_1
|
Command Line |
-Xms2g -Xmx2g -Xmn1536m -XX:+UseParallelOldGC
|
Tuning |
None
|
Notes |
Used numactl to affinitize each Transaction Injector JVM to a physical core in a NUMA node. - numactl --physcpubind=0 --membind=0
- numactl --physcpubind=14 --membind=1
- numactl --physcpubind=28 --membind=2
- numactl --physcpubind=42 --membind=3
- numactl --physcpubind=56 --membind=4
- numactl --physcpubind=70 --membind=5
- numactl --physcpubind=84 --membind=6
- numactl --physcpubind=98 --membind=7
|