Slide 1
Slide 1 text
SERVERLESS
IS
UBY
FUTURE
NIKOLAY SVERCHKOV
@ssnickolay
Slide 2
Slide 2 text
No content
Slide 3
Slide 3 text
evilmartians.com
New
Slide 4
Slide 4 text
ME
Community
Speech
TWO
OPTIONS
Slide 5
Slide 5 text
ME
Community
Speech
VOTE
26 %
74 %
Slide 6
Slide 6 text
SERVERLESS AXIOMS
Event-Driven Architecture
Slide 7
Slide 7 text
SERVERLESS AXIOMS
Event-Driven Architecture
Stateless
Slide 8
Slide 8 text
SERVERLESS AXIOMS
Event-Driven Architecture
Stateless
♾ Infinite Scaling
Slide 9
Slide 9 text
SERVERLESS AXIOMS
Event-Driven Architecture
Stateless
♾ Infinite Scaling
Pay when the code is running
Slide 10
Slide 10 text
SERVERLESS AXIOMS
Event-Driven Architecture
Stateless
♾ Infinite Scaling
Pay when the code is running
Slide 11
Slide 11 text
SERVERLESS
UNDER
THE
MICROSCOPE
Slide 12
Slide 12 text
LOG
$ serverless invoke -f index -l
Slide 13
Slide 13 text
LOG
$ serverless invoke -f index -l
START RequestId: Version: $LATEST
END RequestId:
REPORT RequestId:
Duration: 1.48 ms Billed Duration: 100 ms
Memory Size: 1024 MB Max Memory Used: 31 MB
Slide 14
Slide 14 text
LOG
$ serverless invoke -f index -l
START RequestId: Version: $LATEST
END RequestId:
REPORT RequestId:
Duration: 1.48 ms Billed Duration: 100 ms
Memory Size: 1024 MB Max Memory Used: 31 MB
Slide 15
Slide 15 text
SERVERLESS AXIOMS
Event-Driven Architecture
Stateless
♾ Infinite Scaling
Pay when the code is running
Slide 16
Slide 16 text
LOG
$ serverless invoke -f index -l
START RequestId: Version: $LATEST
END RequestId:
REPORT RequestId:
Duration: 1.48 ms Billed Duration: 100 ms
Memory Size: 1024 MB Max Memory Used: 31 MB
Slide 17
Slide 17 text
LOG
$ serverless invoke -f index -l
START RequestId: Version: $LATEST
END RequestId:
REPORT RequestId:
Duration: 1.48 ms Billed Duration: 100 ms
Memory Size: 1024 MB Max Memory Used: 31 MB
Slide 18
Slide 18 text
WHAT ABOUT
CPU?
Slide 19
Slide 19 text
AWS Lambda allocates CPU power
proportional to the memory…
(с) Amazon Docs
Slide 20
Slide 20 text
AWS Lambda allocates CPU power
proportional to the memory…
For example, if you allocate 256 MB
memory, your Lambda function will receive
twice the CPU share than if you allocated
only 128 MB.
(с) Amazon Docs
Slide 21
Slide 21 text
(CPU, CST)
""<=>
MEMORY
http://bit.ly/2UZ7xDL
Completely Fair Scheduler (CFS)
Slide 22
Slide 22 text
https://aws.amazon.com/lambda/pricing/
6$ "=> 28.8M/128mb!
Slide 23
Slide 23 text
COST
SPEED ⏫
https://amzn.to/2Ix2t2O
Slide 24
Slide 24 text
COST
""<=>
LANGUAGE
Slide 25
Slide 25 text
$ wrk -t10 -c400 -d20s -R40000
$ curl http://endpoint.url/[ruby/go]
"=> {“success”: true}
VS
CHALLENGE:
Slide 26
Slide 26 text
0
250
500
750
1000
0 1 2 4 5
MAX
AVG
AVG ~80ms
31712 requests in 20.01s
Requests/sec: 1584.83
BATTLE (128 MB)
Slide 27
Slide 27 text
0
250
500
750
1000
0 1 2 4 5
0
250
500
750
1000
0 1 2 4 5
MAX
AVG
AVG ~1.8ms
AVG ~80ms
31712 requests in 20.01s
Requests/sec: 1584.83
50347 requests in 20.00s
Requests/sec: 2517.30
BATTLE (128 MB)
Slide 28
Slide 28 text
0
250
500
750
1000
0 1 2 4 5
0
250
500
750
1000
0 1 2 4 5
MAX
AVG
AVG ~1.8ms
AVG ~80ms
31712 requests in 20.01s
Requests/sec: 1584.83
50347 requests in 20.00s
Requests/sec: 2517.30
W
IN
BATTLE (128 MB)
Slide 29
Slide 29 text
0
250
500
750
1000
0 1 2 4 5
0
250
500
750
1000
0 1 2 4 5
MAX
AVG
AVG ~1.8ms
AVG ~80ms
31712 requests in 20.01s
Requests/sec: 1584.83
50347 requests in 20.00s
Requests/sec: 2517.30
W
IN
*
BATTLE (128 MB)
Slide 30
Slide 30 text
BATTLE (128 MB)
AVG ~1.8ms
AVG ~80ms
“Billed Duration: 100 ms” (c) Lambda Docs
Slide 31
Slide 31 text
WHEN YOU REWRITE EVERYTHING ON GO
EXPECTATION REALITY
Slide 32
Slide 32 text
No content
Slide 33
Slide 33 text
FREE!
Free Enterprise
Slide 34
Slide 34 text
FREE!
Free Enterprise
LYING
http://bit.ly/2lCt2dA
Slide 35
Slide 35 text
PREDICTABLY
Slide 36
Slide 36 text
PREDICTABLY
Slide 37
Slide 37 text
PREDICTABLY
Slide 38
Slide 38 text
http://serverlesscalc.com/
$2400
450M!!
$918
1GB
Slide 39
Slide 39 text
BATTLE (128 MB)
AVG ~1.8ms
AVG ~80ms
❓❓❓❓❓❓❓❓❓
❓❓❓❓❓❓❓❓❓❓
❓
❓
❓
❓
❓
Slide 40
Slide 40 text
http://rubyonjets.com/
Slide 41
Slide 41 text
RAILS
ON
SERVERLESS
Tung Nguyen
Slide 42
Slide 42 text
PURE RUBY LAMBDA
# handler.rb
require 'json'
def heartbeat(event:, context:)
{ statusCode: 200, body: JSON.generate(success: true) }
end
Slide 43
Slide 43 text
BATTLE (128 MB)
AVG ~39ms
AVG ~80ms
Slide 44
Slide 44 text
COST
""<=>
DEPENDENCIES
Slide 45
Slide 45 text
MINIMIZE THE COMPLEXITY OF
YOUR DEPENDENCIES
Slide 46
Slide 46 text
MINIMIZE THE COMPLEXITY OF
YOUR DEPENDENCIES
Slide 47
Slide 47 text
MINIMIZE THE COMPLEXITY OF
YOUR DEPENDENCIES
Slide 48
Slide 48 text
NO RAILS
GOOD!
Slide 49
Slide 49 text
CHEAPLY
NICE
NO RAILS
GOOD!
Slide 50
Slide 50 text
CHEAPLY
NICE
EVENTS
AWESOME
NO RAILS
GOOD!
Slide 51
Slide 51 text
CHEAPLY
NICE
EVENTS
AWESOME
NO RAILS
GOOD!
SCALING
AMAZING
Slide 52
Slide 52 text
TALK ABOUT
THE
PROBLEMS
TELL THE TRUTH
Sponsored by @stacey_om
Slide 53
Slide 53 text
IS
INFINITE
SCALING
ALWAYS
GOOD?
Slide 54
Slide 54 text
# before
$ puma -t 8:32 -w 1
# after
$ puma -t 8:32 -w 2
HOW RUBY DEV SCALES APP
Slide 55
Slide 55 text
500
Slide 56
Slide 56 text
SERVERLESS AXIOMS
Event-Driven Architecture
Stateless
♾ Infinite Scaling
Pay when the code is running
Slide 57
Slide 57 text
When you write your Lambda function code, do not
assume that AWS Lambda automatically reuses the
execution context for subsequent function invocations….
(c) AWS Docs
AWS LAMBDA
EXECUTION CONTEXT
Slide 58
Slide 58 text
EXECUTION CONTEXT
# 1M symbols
work = proc { 123_456_789 "** 123_456 }
def heartbeat(event:, context:)
{ statusCode: 200, body: … }
end
Slide 59
Slide 59 text
EXECUTION CONTEXT
# 1M symbols
work = proc { 123_456_789 !** 123_456 }
def heartbeat(event:, context:)
value = work.() # not cached
{
statusCode: 200,
body: JSON.generate(value: value)
}
end
Slide 60
Slide 60 text
EXECUTION CONTEXT
# 1M symbols
work = proc { 123_456_789 !** 123_456 }
value = work.() # cached
def heartbeat(event:, context:)
{
statusCode: 200,
body: JSON.generate(value: value)
}
end
Slide 61
Slide 61 text
CONNECTION POOL
pool = DB"::ConnectionPool.new()
def heartbeat(event:, context:)
connection = pool.get()
# use connection
connection.release()
end
Slide 62
Slide 62 text
https://stackoverflow.com/q/54110697
Slide 63
Slide 63 text
SOLUTIONS*
CLASSIC ($)
Slide 64
Slide 64 text
SOLUTIONS*
CLASSIC ($) MODERN ($$)
Serverless
Slide 65
Slide 65 text
SCALE
⬇
DOWSCALE
⬇
SCALE
⬇
…
Slide 66
Slide 66 text
BURSTY TRAFFIC
BURSTY LOAD
BURSTY ???
Slide 67
Slide 67 text
EXAMPLE
ML
Slide 68
Slide 68 text
EXAMPLE: ETL
Slide 69
Slide 69 text
EXAMPLE: ETL
Slide 70
Slide 70 text
EXAMPLE: BLOG
Slide 71
Slide 71 text
GOTO 0;
SERVERLESS
IS RUBY
FUTURE
Slide 72
Slide 72 text
GOTO 0;
SERVERLESS
IS RUBY
FUTURE
Slide 73
Slide 73 text
DREAMS
ABOUT
SERVERLESS
Slide 74
Slide 74 text
DREAMS
ABOUT
SERVERLESS
Incredible
Processing Power
Slide 75
Slide 75 text
SERVERLESS
Massive User Job
Slide 76
Slide 76 text
SERVERLESS
Backends
Massive User Job
DIFFERENT BACKENDS
Slide 77
Slide 77 text
SERVERLESS
Backends
Massive User Job
10_000.times do
Thread.new(job)
end
Slide 78
Slide 78 text
SERVERLESS
Backends
Burst-Parallel Job
Slide 79
Slide 79 text
SERVERLESS
Burst-Parallel Job
Video Encoding: VP8
Slide 80
Slide 80 text
Burst-Parallel Job
Video Encoding: Classic
encode([, , …, ]) "-> keyframe + interframe(2:n)
Slide 81
Slide 81 text
Burst-Parallel Job
Video Encoding: Existing
encode([, , …, ]) "-> keyframe + interframe(2:n)
1MB +25kb * n
Slide 82
Slide 82 text
Burst-Parallel Job
Video Encoding: Existing
encode([, , …, ]) "-> keyframe + interframe(2:n)
decode(keyframe + interframe[2:n]) "-> [, , …, ]
1MB +25kb * n
Slide 83
Slide 83 text
Burst-Parallel Job
Video Encoding: Existing
encode(i[1:10]) "-> keyframe1 + interframe(2:10)
encode(i[11:20]) "-> keyframe2 + interframe(12:20)
encode(i[21:30]) "-> keyframe3 + interframe(22:30)
1MB * N
Slide 84
Slide 84 text
VIDEO ENCODING: ALFALA
http://bit.ly/2lZFrbO
Slide 85
Slide 85 text
Burst-Parallel Job
preprocessor(compiler
(assembler(linker
(archiver(indexer
(strip(source_file)…)
Software Compilation: 7 Steps
Slide 86
Slide 86 text
Burst-Parallel Job
Software Compilation
Slide 87
Slide 87 text
Burst-Parallel Job
Software Compilation
Video Encoding
Slide 88
Slide 88 text
DIRECTED ACYCLIC GRAPH
BRO NOT A BRO
Slide 89
Slide 89 text
DIRECTED ACYCLIC GRAPH
Problem: Different Nodes
g
f
h
m
k
l
p
r
Slide 90
Slide 90 text
DAG "=> FUNCTIONS
preprocessor(compiler
(assembler(linker
(archiver(indexer
(strip(source_file)…)
Slide 91
Slide 91 text
f(g(h(p(…)…)
Slide 92
Slide 92 text
f(g(i)) !== λ(f, λ(g, i))
THE NEW ABSTRACTION: λ
Slide 93
Slide 93 text
f(g(i)) !== λ(f, λ(g, i))
f = proc { |a| a + 1 }
g = proc { |a| a + 2 }
f.call(g.call(0))
# "=> 3
THE NEW ABSTRACTION: λ
Slide 94
Slide 94 text
f(g(i)) !== λ(f, λ(g, i))
f = proc { |a| a + 1 }
g = proc { |a| a + 2 }
f.call(g.call(0))
# !=> 3
λ = proc { |f, i| f.(i) }
λ.call(f, λ.call(g, 0))
# "=> 3
THE NEW ABSTRACTION: λ
Slide 95
Slide 95 text
f(g(h(p(…)…) !==
λ(f, λ(g, λ(h, λ(p(…)…)
Slide 96
Slide 96 text
f(g(h(p(…)…) !==
λ(f, λ(g, λ(h, λ(p(…)…)
λ(ƒ,arg) !=> arg
WHERE
Slide 97
Slide 97 text
λ(ƒ,arg) !=> arg
Slide 98
Slide 98 text
λ(ƒ,arg) !=> arg
λ(ƒ,arg) "=> Θ,arg
Slide 99
Slide 99 text
λ(ƒ,arg) !=> arg
λ(ƒ,arg) !=> Θ,arg
λ(T) "=> T
Slide 100
Slide 100 text
DIRECTED ACYCLIC GRAPH
T2
T1
T3
T5
T4
T6
T7
T8
λ(T) "=> T, T{i} "<- T
Slide 101
Slide 101 text
SERVERLESS
Backends
Burst-Parallel Job
Slide 102
Slide 102 text
SERVERLESS
Backends
Burst-Parallel Job
λ(T) "=> T
Slide 103
Slide 103 text
SERVERLESS
Backends
Burst-Parallel Job
λ(T) "=> T
λ
Slide 104
Slide 104 text
SERVERLESS
Backends
Burst-Parallel Job
λ(T) "=> T
λ
IR
T
Slide 105
Slide 105 text
SERVERLESS
Backends
Burst-Parallel Job
IR
λ(T) "=> T
T
λ
STATIC
Slide 106
Slide 106 text
SERVERLESS
Backends
Burst-Parallel Job
λ
IR
T
Frontends
T generator
λ(T) "=> T
Slide 107
Slide 107 text
SERVERLESS
Backends
Burst-Parallel Job
λ
IR
T
Frontends
T generator
λ(T) "=> T
Slide 108
Slide 108 text
Burst-Parallel Job
IR
Frontends
Coordinator
HTTP
λ
Slide 109
Slide 109 text
SERVERLESS
Backend
Burst-Parallel Job
IR
Frontends
Coordinator
STATELESS
HTTP
Slide 110
Slide 110 text
Burst-Parallel Job
IR
Frontends
Coordinator
S3
HTTP
λ
Slide 111
Slide 111 text
THE FUTURE
IS HERE!
Slide 112
Slide 112 text
gg
Backends
Burst-Parallel Job
gg IR
Frontends
gg
Coordinator
C++
HTTP
ProtoBuf
S3
Stanford Systems
and
Networking Research
Slide 113
Slide 113 text
Fib(4)
fib(4)
fib(3) fib(2)
fib(2) fib(1)
fib(1) fib(0)
fib(1) fib(0)
+
+
+
+
Slide 114
Slide 114 text
Fib(4)
fib(4)
fib(3) fib(2)
fib(2) fib(1)
fib(1) fib(0)
fib(1) fib(0)
+
+
+
+
Hash#fib1
Hash#fib2
Slide 115
Slide 115 text
Нихуя не понял
Но очень интересно!
чего
Slide 116
Slide 116 text
gg FRAMEWORK
https://stanford.io/2mtmiiH
Slide 117
Slide 117 text
DEMO
Slide 118
Slide 118 text
DEMO
http://bit.ly/2m0G3Oo
https://asciinema.org/a/270154
Slide 119
Slide 119 text
RESULTS
gg
real
user
sys
1m17.670s
1m49.040s
0m7.820s
real
user
sys
0m43.354s
0m3.830s
0m5.490s
real
user
sys
0m29.376s
0m0.950s
0m1.400s
Slide 120
Slide 120 text
RESULTS
gg
real
user
sys
1m17.670s
1m49.040s
0m7.820s
real
user
sys
0m43.354s
0m3.830s
0m5.490s
real
user
sys
0m29.376s
0m0.950s
0m1.400s
Slide 121
Slide 121 text
RESULTS
gg
Total: 1m56s Total: 0m38s
~X3 faster
Slide 122
Slide 122 text
COMPARISON OF COLD-CACHE BUILD TIMES
https://cs.stanford.edu/~matei/papers/2019/usenix_atc_gg.pdf
Slide 123
Slide 123 text
DARK
MAGIC!
Slide 124
Slide 124 text
https://github.com/StanfordSNR/gg
Slide 125
Slide 125 text
GOTO 0;
SERVERLESS
IS RUBY
FUTURE
Slide 126
Slide 126 text
“As a computing substrate, we suspect cloud
functions are in a similar position to Graphics
Processing Units in the 2000s.
At the time, GPUs were designed solely for 3D
graphics, but the community gradually
recognized that they had become programmable
enough to execute some parallel algorithms
unrelated to graphics.” (c)
Slide 127
Slide 127 text
GOTO 0;
SERVERLESS
IS RUBY
FUTURE
Slide 128
Slide 128 text
SORBET?
Slide 129
Slide 129 text
IT saṃsāra
Java
Erlang
C
C++
"=> Scala
"=> Elixir
"=> Go
"=> Rust
Slide 130
Slide 130 text
RUBY
SELF-IDENTIFICATION
Slide 131
Slide 131 text
RUBY IS
Fast
Coding
Slide 132
Slide 132 text
RUBY IS
Startups
(Rails)
Fast
Coding
Slide 133
Slide 133 text
RUBY IS
Startups
(Rails)
GIL
Sensitive
Fast
Coding
Slide 134
Slide 134 text
RUBY IS
Startups
(Rails)
オブジェクト指向スクリプト⾔言語 Ruby
GIL
Sensitive
Fast
Coding
Slide 135
Slide 135 text
RUBY IS
Startups
(Rails)
GIL
Sensitive
The Object-oriented
Scripting Language Ruby
Fast
Coding
Slide 136
Slide 136 text
RUBY IS
Startups
(Rails)
GIL
Sensitive
The Object-oriented
Scripting Language Ruby
Fast
Coding
Slide 137
Slide 137 text
RUBY IS
Startups
(Rails)
GIL
Sensitive
The Object-oriented
Scripting Language Ruby
Fast
Coding
Slide 138
Slide 138 text
RUBY IS
Startups
(Rails)
GIL
Sensitive
The Object-oriented
Scripting Language Ruby
Fast
Coding
Slide 139
Slide 139 text
RUBY IS
Startups
(Rails)
GIL
Sensitive
The Object-oriented
Scripting Language Ruby
Fast
Coding
Slide 140
Slide 140 text
60 SECNDS
OF SUMMARY
Slide 141
Slide 141 text
@ssnickolay
Nikolay Sverchkov
@ssnickolay
THANK YOU!
@evilmartians
http://evl.ms/blog
http://evl.ms/telegram