Oct	
  2015	
  
Java Mixed-Mode
Flame Graphs
Brendan Gregg Senior Performance Architect

Understanding Java CPU usage
quickly and completely

Quickly
• Via SSH and open source tools (covered in this talk)
• Or using Netflix Vector GUI (also open source):
1. Observe high CPU usage
2. Generate a flame graph

Completely
Java Mixed-Mode Flame Graph
via Linux perf_events:
Kernel
Java
JVM

Messy House Fallacy
Fallacy:	
  my	
  code	
  is	
  a	
  mess,	
  I	
  bet	
  yours	
  is	
  
immaculate,	
  therefore	
  the	
  bug	
  must	
  be	
  mine	
  
	
  
Reality:	
  everyone's	
  code	
  is	
  terrible	
  and	
  buggy	
  
• Don't overlook system code: kernel, libraries, etc.

Context

• Over 60 million subscribers
– Just launched in Spain!
• AWS EC2 Linux cloud
• FreeBSD CDN
• Awesome place to work

Cloud
• Tens of thousands of AWS EC2 instances
• Mostly running Java applications (Oracle JVM)
Linux	
  (usually	
  Ubuntu)	
  
OpMonal	
  Apache,	
  
memcached,	
  Node.js,	
  
…	
  
Atlas,	
  S3	
  log	
  rotaMon,	
  
sar,	
  Trace,	
  perf,	
  stap,	
  
perf-­‐tools	
  
Vector,	
  pcp	
  
Java	
  (JDK	
  8)	
  
GC	
  and	
  
thread	
  
dump	
  
logging	
  
Tomcat	
  
ApplicaMon	
  war	
  files,	
  
plaYorm,	
  base	
  servlet	
  
hystrix,	
  metrics	
  (Servo),	
  
health	
  check	
  

Why we need CPU profiling
• Improving performance
• Cost savings
– ASGs often scale on load
average (CPU), so CPU
usage is proportional to cost
Auto	
  Scaling	
  
Group	
  
Scaling	
  Policy	
  
CloudWatch,	
  Servo	
  
– Identify tuning targets
– Incident response
– Non-regression testing
– Software evaluations
– CPU workload
characterization
loadavg,	
  latency,	
  …	
  
	
  
Instance	
  
Instance	
  
Instance	
  

The Problem with Profilers

Java Profilers
Kernel,
libraries,
JVM
Java

Java Profilers
• Visibility
– Java method execution
– Object usage
– GC logs
– Custom Java context
• Typical problems:
– Sampling often happens at safety/yield points (skew)
– Method tracing has massive observer effect
– Misidentifies RUNNING as on-CPU (e.g., epoll)
– Doesn't include or profile GC or JVM CPU time
– Tree views not quick (proportional) to comprehend
• Inaccurate (skewed) and incomplete profiles

System Profilers
Kernel
TCP/IP
Java
JVM
Locks
Time
epoll
Idle
thread

System Profilers
• Visibility
– JVM (C++)
– GC (C++)
– libraries (C)
– kernel (C)
• Typical problems (x86):
– Stacks missing for Java
– Symbols missing for Java methods
• Other architectures (e.g., SPARC) have fared better
• Profile everything except Java

Workaround
• Capture both Java and system profiles, and examine
side by side
Java
System
• An improvement, but Java context is often crucial for
interpreting system profiles

Solution
Java Mixed-Mode Flame Graph
Kernel
Java
JVM

Solution
• Fix system profiling
– Only way to see it all
Kernel
• Visibility is everything:
– Java methods
– JVM (C++)
– GC (C++)
– libraries (C)
– kernel (C)
Java
JVM
• Minor Problems:
– 0-3% CPU overhead to enable frame pointers (usually 
slide 18:
Simple Production Example
1. Poor performance,
and one CPU at 100%
2. perf_events flame
graph shows JVM
stuck compiling

slide 19:
Another System Example
Exception handling consuming CPU

slide 20:
DEMO
FlameGraph_tomcat01.svg

slide 21:
Exonerating The System
• From last week:
- Frequent thread creation/
destruction assumed to be
consuming CPU resources.
Recode application?
- A flame graph quantified this
CPU time: near zero
- Time mostly other Java methods

slide 22:
Profiling GC
GC internals, visualized:

slide 23:
CPU Profiling

slide 24:
CPU Profiling
• Record stacks at a timed interval: simple and effective
– Pros: Low (deterministic) overhead
– Cons: Coarse accuracy, but usually sufficient
stack
samples:
syscall
on-CPU
time
off-CPU
block
interrupt

slide 25:
Stack Traces
• A code path snapshot. e.g., from jstack(1):
$ jstack 1819
[…]
"main" prio=10 tid=0x00007ff304009000
nid=0x7361 runnable [0x00007ff30d4f9000]
java.lang.Thread.State: RUNNABLE
running
at Func_abc.func_c(Func_abc.java:6)
codepath
at Func_abc.func_b(Func_abc.java:16)
start
at Func_abc.func_a(Func_abc.java:23)
at Func_abc.main(Func_abc.java:27)
running
parent
g.parent
g.g.paren

slide 26:
System Profilers
• Linux
– perf_events (aka "perf")
• Oracle Solaris
– DTrace
• OS X
– Instruments
• Windows
– XPerf
• And many others…

slide 27:
Linux perf_events
• Standard Linux profiler
– Provides the perf command (multi-tool)
– Usually pkg added by linux-tools-common, etc.
• Features:
– Timer-based sampling
– Hardware events
– Tracepoints
– Dynamic tracing
• Can sample stacks of (almost) everything on CPU
– Can miss hard interrupt ISRs, but these should be near-zero. They can
be measured if needed (I wrote my own tools)

slide 28:
perf record Profiling
• Stack profiling on all CPUs at 99 Hertz, then dump:
# perf record -F 99 -ag -- sleep 30
[ perf record: Woken up 9 times to write data ]
[ perf record: Captured and wrote 2.745 MB perf.data (~119930 samples) ]
# perf script
[…]
bash 13204 cpu-clock:
459c4c dequote_string (/root/bash-4.3/bash)
465c80 glob_expand_word_list (/root/bash-4.3/bash)
466569 expand_word_list_internal (/root/bash-4.3/bash)
465a13 expand_words (/root/bash-4.3/bash)
43bbf7 execute_simple_command (/root/bash-4.3/bash)
435f16 execute_command_internal (/root/bash-4.3/bash)
one
435580 execute_command (/root/bash-4.3/bash)
stack
43a771 execute_while_or_until (/root/bash-4.3/bash)
sample
43a636 execute_while_command (/root/bash-4.3/bash)
436129 execute_command_internal (/root/bash-4.3/bash)
435580 execute_command (/root/bash-4.3/bash)
420cd5 reader_loop (/root/bash-4.3/bash)
41ea58 main (/root/bash-4.3/bash)
7ff2294edec5 __libc_start_main (/lib/x86_64-linux-gnu/libc-2.19.so)
[… ~47,000 lines truncated …]

slide 29:
perf report Summary
• Generates a call tree and combines samples:
# perf report -n -stdio
[…]
# Overhead
Samples Command
Shared Object
Symbol
# ........ ............ ....... ................. .............................
20.42%
bash [kernel.kallsyms] [k] xen_hypercall_xen_version
--- xen_hypercall_xen_version
check_events
call tree
summary
|--44.13%-- syscall_trace_enter
tracesys
|--35.58%-- __GI___libc_fcntl
|--65.26%-- do_redirection_internal
do_redirections
execute_builtin_or_function
execute_simple_command
[… ~13,000 lines truncated …]

slide 30:
Flame Graphs

slide 31:
perf report Verbosity
• Despite summarizing, output is still verbose
# perf report -n -stdio
[…]
# Overhead
Samples Command
Shared Object
Symbol
# ........ ............ ....... ................. .............................
20.42%
bash [kernel.kallsyms] [k] xen_hypercall_xen_version
--- xen_hypercall_xen_version
check_events
|--44.13%-- syscall_trace_enter
tracesys
|--35.58%-- __GI___libc_fcntl
|--65.26%-- do_redirection_internal
do_redirections
execute_builtin_or_function
execute_simple_command
[… ~13,000 lines truncated …]

slide 32:
Full perf report Output

slide 33:
… as a Flame Graph

slide 34:
Flame Graphs
git clone --depth 1 https://github.com/brendangregg/FlameGraph
cd FlameGraph
perf record -F 99 -a –g -- sleep 30
perf script | ./stackcollapse-perf.pl | ./flamegraph.pl >gt; perf.svg
• Flame Graphs:
– x-axis: alphabetical stack sort, to maximize merging
– y-axis: stack depth
– color: random (default), or a dimension
• Currently made from Perl + SVG + JavaScript
– Multiple d3 versions are being developed
• Easy to get working
– http://www.brendangregg.com/FlameGraphs/cpuflamegraphs.html
– Above commands are Linux; see URL for other OSes

slide 35:
Linux perf_events Workflow
list events
count events
capture stacks
perf list
perf stat
perf record
Typical
Workflow
perf.data	
  
text UI
dump profile
perf report
flame graph
visualization
perf script
stackcollapse-perf.pl
flamegraph.pl

slide 36:
Flame Graph Interpretation
g()
e()
f()
d()
c()
i()
b()
h()
a()

slide 37:
Flame Graph Interpretation (1/3)
Top edge shows who is running on-CPU,
and how much (width)
g()
e()
f()
d()
c()
i()
b()
h()
a()

slide 38:
Flame Graph Interpretation (2/3)
Top-down shows ancestry
e.g., from g():
g()
e()
f()
d()
c()
i()
b()
h()
a()

slide 39:
Flame Graph Interpretation (3/3)
Widths are proportional to presence in samples
e.g., comparing b() to h() (incl. children)
g()
e()
f()
d()
c()
i()
b()
h()
a()

slide 40:
Flame Graph Colors
• Randomized by default
• Can be used as a dimension. e.g.:
– Mixed-mode flame graphs
– Differential flame graphs
– Search

slide 41:
Mixed-Mode Flame Graphs
• Hues:
– green == Java
– red == system
– yellow == C++
• Intensity randomized
to differentiate frames
– Or hashed based on
function name
Mixed-Mode
Kernel
Java
JVM

slide 42:
Differential Flame Graphs
• Hues:
Differential
– red == more samples
– blue == less samples
• Intensity shows the
degree of difference
• Used for comparing
two profiles
• Also used for showing
other metrics: e.g., CPI
more
less

slide 43:
Flame Graph Search
• Color: magenta to show matched frames
search
button

slide 44:
Flame Charts
• Final note: these are useful, but are not flame graphs
• Flame charts: x-axis is time
• Flame graphs: x-axis is population (maximize merging)

slide 45:
Stack Tracing

slide 46:
System Profiling Java on x86
• For example,
using Linux perf
• The stacks are
1 or 2 levels
deep, and have
junk values
# perf record –F 99 –a –g – sleep 30
# perf script
[…]
java 4579 cpu-clock:
ffffffff8172adff tracesys ([kernel.kallsyms])
7f4183bad7ce pthread_cond_timedwait@@GLIBC_2…
java
4579 cpu-clock:
7f417908c10b [unknown] (/tmp/perf-4458.map)
java
4579 cpu-clock:
7f4179101c97 [unknown] (/tmp/perf-4458.map)
java
4579 cpu-clock:
7f41792fc65f [unknown] (/tmp/perf-4458.map)
a2d53351ff7da603 [unknown] ([unknown])
java
4579 cpu-clock:
7f4179349aec [unknown] (/tmp/perf-4458.map)
java
4579 cpu-clock:
7f4179101d0f [unknown] (/tmp/perf-4458.map)
[…]

slide 47:
… as a Flame Graph
Broken Java stacks
(missing frame pointer)

slide 48:
Why Stacks are Broken
• On x86 (x86_64), hotspot uses
the frame pointer register (RBP)
as general purpose
• This "compiler optimization"
breaks (simple) stack walking
• Once upon a time, x86 had fewer
registers, and this made much more sense
• gcc provides -fno-omit-frame-pointer to avoid
doing this, but the JVM had no such option…

slide 49:
Fixing Stack Walking
Possibilities:
A. Fix frame pointer-based stack walking (the default)
– Pros: simple, supported by many tools
– Cons: might cost a little extra CPU
B. Use a custom walker (likely needing kernel support)
– Pros: full stack walking (incl. inlining) & arguments
– Cons: custom kernel code, can cost more CPU when in use
C. Try libunwind and DWARF
– Even feasible with JIT?
Our current preference is (A)

slide 50:
Hacking OpenJDK (1/2)
• As a proof of concept, I hacked hotspot to support an
x86_64 frame pointer
--- openjdk8clean/hotspot/src/cpu/x86/vm/x86_64.ad 2014-03-04 …
+++ openjdk8/hotspot/src/cpu/x86/vm/x86_64.ad 2014-11-08 …
@@ -166,10 +166,9 @@
// 3) reg_class stack_slots( /* one chunk of stack-based "registers" */ )
-// Class for all pointer registers (including RSP)
+// Class for all pointer registers (including RSP, excluding RBP)
reg_class any_reg(RAX, RAX_H,
RDX, RDX_H,
RBP, RBP_H,
Remove RBP from
RDI, RDI_H,
RSI, RSI_H,
register pools
RCX, RCX_H,
[...]

slide 51:
Hacking OpenJDK (2/2)
--- openjdk8clean/hotspot/src/cpu/x86/vm/macroAssembler_x86.cpp 2014-03-04…
+++ openjdk8/hotspot/src/cpu/x86/vm/macroAssembler_x86.cpp 2014-11-07 …
@@ -5236,6 +5236,7 @@
// We always push rbp, so that on return to interpreter rbp, will be
// restored correctly and we can correct the stack.
push(rbp);
mov(rbp, rsp);
Fix x86_64 function
// Remove word for ebp
prologues
framesize -= wordSize;
--- openjdk8clean/hotspot/src/cpu/x86/vm/c1_MacroAssembler_x86.cpp …
+++ openjdk8/hotspot/src/cpu/x86/vm/c1_MacroAssembler_x86.cpp …
[...]
• We used this patched version successfully for some limited
(and urgent) performance analysis

slide 52:
-XX:+PreserveFramePointer
• We shared our patch publicly
– See "A hotspot patch for stack profiling (frame pointer)" on the
hotspot complier dev mailing list
– It became JDK-8068945 for JDK 9 and JDK-8072465 for JDK 8,
and the -XX:+PreserveFramePointer option
• Zoltán Majó (Oracle) took this on, rewrote it, and it is now:
– In JDK 9
– In JDK 8 update 60 build 19
– Thanks to Zoltán, Oracle, and the other hotspot engineers for
helping get this done!
• It might cost 0 – 3% CPU, depending on workload

slide 53:
Broken Java Stacks (before)
# perf script
[…]
java 4579 cpu-clock:
ffffffff8172adff tracesys ([kernel.kallsyms])
7f4183bad7ce pthread_cond_timedwait@@GLIBC_2…
java
4579 cpu-clock:
7f417908c10b [unknown] (/tmp/perf-4458.map)
java
4579 cpu-clock:
7f4179101c97 [unknown] (/tmp/perf-4458.map)
java
4579 cpu-clock:
7f41792fc65f [unknown] (/tmp/perf-4458.map)
a2d53351ff7da603 [unknown] ([unknown])
java
4579 cpu-clock:
7f4179349aec [unknown] (/tmp/perf-4458.map)
java
4579 cpu-clock:
7f4179101d0f [unknown] (/tmp/perf-4458.map)
java
4579 cpu-clock:
7f417908c194 [unknown] (/tmp/perf-4458.map)
[…]
• Check with "perf
script" to see stack
samples
• These are 1 or 2
levels deep (junk
values)

slide 54:
Fixed Java Stacks
# perf script
[…]
java 8131 cpu-clock:
7fff76f2dce1 [unknown] ([vdso])
7fd3173f7a93 os::javaTimeMillis() (/usr/lib/jvm…
7fd301861e46 [unknown] (/tmp/perf-8131.map)
7fd30184def8 [unknown] (/tmp/perf-8131.map)
7fd30174f544 [unknown] (/tmp/perf-8131.map)
7fd30175d3a8 [unknown] (/tmp/perf-8131.map)
7fd30166d51c [unknown] (/tmp/perf-8131.map)
7fd301750f34 [unknown] (/tmp/perf-8131.map)
7fd3016c2280 [unknown] (/tmp/perf-8131.map)
7fd301b02ec0 [unknown] (/tmp/perf-8131.map)
7fd3016f9888 [unknown] (/tmp/perf-8131.map)
7fd3016ece04 [unknown] (/tmp/perf-8131.map)
7fd30177783c [unknown] (/tmp/perf-8131.map)
7fd301600aa8 [unknown] (/tmp/perf-8131.map)
7fd301a4484c [unknown] (/tmp/perf-8131.map)
7fd3010072e0 [unknown] (/tmp/perf-8131.map)
7fd301007325 [unknown] (/tmp/perf-8131.map)
7fd301007325 [unknown] (/tmp/perf-8131.map)
7fd3010004e7 [unknown] (/tmp/perf-8131.map)
7fd3171df76a JavaCalls::call_helper(JavaValue*,…
7fd3171dce44 JavaCalls::call_virtual(JavaValue*…
7fd3171dd43a JavaCalls::call_virtual(JavaValue*…
7fd31721b6ce thread_entry(JavaThread*, Thread*)…
7fd3175389e0 JavaThread::thread_main_inner() (/…
7fd317538cb2 JavaThread::run() (/usr/lib/jvm/nf…
7fd3173f6f52 java_start(Thread*) (/usr/lib/jvm/…
7fd317a7e182 start_thread (/lib/x86_64-linux-gn…
• With -XX:
+PreserveFramePointer
stacks are full, and
go all the way to
start_thread()
• This is what the
CPUs are really
running: inlined
frames are not
present

slide 55:
Fixed Stacks Flame Graph
Java stacks
(but no symbols)

slide 56:
Stacks & Inlining
• Frames may be missing (inlined)
• Disabling inlining:
No inlining
– -XX:-Inline
– Many more Java frames
– Can be 80% slower!
• May not be necessary
– Inlined flame graphs often make
enough sense
– Or tune -XX:MaxInlineSize and
-XX:InlineSmallCode a little to reveal more frames
• Can even improve performance!
• perf-map-agent (next) has experimental un-inline support

slide 57:
Symbols

slide 58:
Missing Symbols
• Missing symbols may show up as hex; e.g., Linux perf:
71.79%
sed
sed
|--11.65%-- 0x40a447
0x40659a
0x408dd8
0x408ed1
0x402689
0x7fa1cd08aec5
12.06%
sed
sed
--- re_search_internal
|--96.78%-- re_search_stub
rpl_re_search
match_regex
do_subst
execute_program
process_files
main
__libc_start_main
[.] 0x000000000001afc1
broken
[.] re_search_internal
not broken

slide 59:
Fixing Symbols
For JIT'd code, Linux perf already looks for an
externally provided symbol file: /tmp/perf-PID.map, and
warns if it doesn't exist
# perf script
Failed to open /tmp/perf-8131.map, continuing without symbols
[…]
java 8131 cpu-clock:
7fff76f2dce1 [unknown] ([vdso])
7fd3173f7a93 os::javaTimeMillis() (/usr/lib/jvm…
7fd301861e46 [unknown] (/tmp/perf-8131.map)
[…]
This file can be created by a Java agent

slide 60:
Java Symbols for perf
• perf-map-agent
– https://github.com/jrudolph/perf-map-agent
– Agent attaches and writes the /tmp file on demand (previous
versions attached on Java start, wrote continually)
– Thanks Johannes Rudolph!
• Use of a /tmp symbol file
– Pros: simple, can be low overhead (snapshot on demand)
– Cons: stale symbols
• Using a symbol logger with perf instead
– Patch by Stephane Eranian currently being discussed on
lkml; see "perf: add support for profiling jitted code"

slide 61:
Stacks & Symbols
Java Mixed-Mode Flame Graph
Kernel
Java
JVM

slide 62:
Stacks & Symbols (zoom)

slide 63:
Instructions

slide 64:
Instructions
1. Check Java version
2. Install perf-map-agent
3. Set -XX:+PreserveFramePointer
4. Profile Java
5. Dump symbols
6. Generate Mixed-Mode Flame Graph
Note these are unsupported: use at your own risk
Reference: http://techblog.netflix.com/2015/07/java-in-flames.html

slide 65:
1. Check Java Version
• Need JDK8u60 or better
– for -XX:+PreserveFramePointer
$ java -version
java version "1.8.0_60"
Java(TM) SE Runtime Environment (build 1.8.0_60-b27)
Java HotSpot(TM) 64-Bit Server VM (build 25.60-b23, mixed mode)
• Upgrade Java if necessary

slide 66:
2. Install perf-map-agent
• Check https://github.com/jrudolph/perf-map-agent for the
latest instructions. e.g.:
$ sudo bash
# apt-get install -y cmake
# export JAVA_HOME=/usr/lib/jvm/java-8-oracle
# cd /usr/lib/jvm
# git clone --depth=1 https://github.com/jrudolph/perf-map-agent
# cd perf-map-agent
# cmake .
# make

slide 67:
3. Set -XX:+PreserveFramePointer
• Needs to be set on Java startup
• Check it is enabled (on Linux):
$ ps wwp `pgrep –n java` | grep PreserveFramePointer

slide 68:
4. Profile Java
• Using Linux perf_events to profile all processes, at 99
Hertz, for 30 seconds (as root):
# perf record -F 99 -a -g -- sleep 30
• Just profile one PID (broken on some older kernels):
# perf record -F 99 -p PID -g -- sleep 30
• These create a perf.data file

slide 69:
5. Dump Symbols
• See perf-map-agent docs for updated usage
• e.g., as the same user as java:
$ cd /usr/lib/jvm/perf-map-agent/out
$ java -cp attach-main.jar:$JAVA_HOME/lib/tools.jar \
net.virtualvoid.perf.AttachOnce PID
• perf-map-agent contains helper scripts. I wrote my own:
– https://github.com/brendangregg/Misc/blob/master/java/jmaps
• Dump symbols quickly after perf record to minimize stale
symbols. How I do it:
# perf record -F 99 -a -g -- sleep 30; jmaps

slide 70:
6. Generate a Mixed-Mode Flame Graph
• Using my FlameGraph software:
# perf script >gt; out.stacks01
# git clone --depth=1 https://github.com/brendangregg/FlameGraph
# cat out.stacks01 | ./FlameGraph/stackcollapse-perf.pl | \
./FlameGraph/flamegraph.pl --color=java --hash >gt; flame01.svg
– perf script reads perf.data with /tmp/*.map
– out.stacks01 is an intermediate file; can be handy to keep
• Finally open flame01.svg in a browser
• Check for newer flame graph implementations (e.g., d3)

slide 71:
Automation

slide 72:
Netflix Vector

slide 73:
Netflix Vector
Select Instance
Select
Metrics
Flame Graphs
Near real-time,
per-second metrics

slide 74:
Netflix Vector
• Open source, on-demand, instance analysis tool
– https://github.com/netflix/vector
• Shows various real-time metrics
• Flame graph support currently in development
– Automating previous steps
– Using it internally already
– Also developing a new d3 front end

slide 75:
DEMO
d3-flame-graph

slide 76:
Advanced Analysis

slide 77:
Linux perf_events Coverage
… all possible with Java stacks

slide 78:
Advanced Flame Graphs
• Examples:
– Page faults
– Context switches
– Disk I/O requests
– TCP events
– CPU cache misses
– CPI
• Any event issued in synchronous Java context

slide 79:
Synchronous Java Context
• Java thread still on-CPU, and event is directly triggered
• Examples:
– Disk I/O requests issued directly by Java à yes
• direct reads, sync writes, page faults
– Disk I/O completion interrupts à no*
– Disk I/O requests triggered async, e.g., readahead à no*
* can be made yes by tracing and associating context

slide 80:
Page Faults
• Show what triggered main memory (resident) to grow:
# perf record -e page-faults -p PID -g -- sleep 120
• "fault" as (physical) main memory is allocated ondemand, when a virtual page is first populated
• Low overhead tool to solve some types of memory leak
RES column in top(1)
grows
because

slide 81:
Page Fault Flame Graph
Java code
epoll

slide 82:
Context Switches
• Show why Java blocked and stopped running on-CPU:
# perf record -e context-switches -p PID -g -- sleep 5
• Identifies locks, I/O, sleeps
– If code path shouldn't block and looks random, it's an involuntary context switch. I
could filter these, but you should have solved them beforehand (CPU load).
• e.g., was used to understand framework differences:
rxNetty
Tomcat

slide 83:
Context Switch Flame Graph (1/2)
rxNetty
epoll
futex

slide 84:
Context Switch Flame Graph (2/2)
Tomcat
sys_poll
futex

slide 85:
Disk I/O Requests
• Shows who issued disk I/O (sync reads & writes):
# perf record -e block:block_rq_insert -a -g -- sleep 60
• e.g.: page faults in GC? This JVM has swapped out!:

slide 86:
TCP Events
• TCP transmit, using dynamic tracing:
# perf probe tcp_sendmsg
# perf record -e probe:tcp_sendmsg -a -g -- sleep 1; jmaps
# perf script -f comm,pid,tid,cpu,time,event,ip,sym,dso,trace >gt; out.stacks
# perf probe --del tcp_sendmsg
• Note: can be high overhead for high packet rates
– For the current perf trace, dump, post-process cycle
• Can also trace TCP connect & accept (lower overhead)
• TCP receive is async
– Could trace via socket read

slide 87:
TCP Send Flame Graph
Only one code-path
taken in this example
kernel
Java
JVM
ab (client process)

slide 88:
CPU Cache Misses
• In this example, sampling via Last Level Cache loads:
# perf record -e LLC-loads -c 10000 -a -g -- sleep 5; jmaps
# perf script -f comm,pid,tid,cpu,time,event,ip,sym,dso >gt; out.stacks
• -c is the count (samples
once per count)
• Use other CPU counters to
sample hits, misses, stalls

slide 89:
One Last Example
• Back to a
mixed-mode
CPU flame graph
• What else can we
show with color?

slide 90:
CPI Flame Graph
• Cycles Per
Instruction!
– red == instruction
heavy
– blue == cycle
heavy (likely mem
stall cycles)
zoomed:

slide 91:
Links & References
Flame Graphs
– http://www.brendangregg.com/flamegraphs.html
– http://techblog.netflix.com/2015/07/java-in-flames.html
– http://techblog.netflix.com/2014/11/nodejs-in-flames.html
– http://www.brendangregg.com/blog/2014-11-09/differential-flame-graphs.html
Linux perf_events
– https://perf.wiki.kernel.org/index.php/Main_Page
– http://www.brendangregg.com/perf.html
– http://www.brendangregg.com/blog/2015-02-27/linux-profiling-at-netflix.html
Netflix Vector
– https://github.com/netflix/vector
– http://techblog.netflix.com/2015/04/introducing-vector-netflixs-on-host.html
JDK tickets
– JDK8: https://bugs.openjdk.java.net/browse/JDK-8072465
– JDK9: https://bugs.openjdk.java.net/browse/JDK-8068945
hprof: http://www.brendangregg.com/blog/2014-06-09/java-cpu-sampling-using-hprof.html

slide 92:
Oct	
  2015	
  
Thanks
Questions?
http://techblog.netflix.com
http://slideshare.net/brendangregg
http://www.brendangregg.com
[email protected]
@brendangregg