Upgrade to Pro
— share decks privately, control downloads, hide ads and more …
Speaker Deck
Speaker Deck
PRO
Sign in
Sign up
for free
Ruby & Friends - Taking Go as an example
Richard Lee
April 26, 2014
Technology
3
140
Ruby & Friends - Taking Go as an example
Presented at RubyConf TW 2014
Richard Lee
April 26, 2014
Tweet
Share
More Decks by Richard Lee
See All by Richard Lee
dlackty
0
5
dlackty
0
24
dlackty
0
15
dlackty
7
360
dlackty
8
500
dlackty
1
110
dlackty
10
330
dlackty
80
7.1k
dlackty
23
700
Other Decks in Technology
See All in Technology
harshbothra
0
130
sat
40
29k
koukyo1994
3
500
gobeyond20xx
0
230
masakazu
0
180
karamem0
1
750
gamella
3
1.5k
900groove
2
450
soracom
0
160
chandi
0
120
nisshii0313
1
170
hmatsu47
4
600
Featured
See All Featured
vanstee
117
4.9k
tenderlove
53
3.5k
tammielis
237
23k
chriscoyier
145
20k
lara
172
9.6k
danielanewman
1
520
ufuk
56
5.4k
jlugia
217
16k
iamctodd
19
2k
rasmusluckow
318
18k
keathley
20
710
eileencodes
113
25k
Transcript
Ruby & Friends Taking Go as an example
... In this talk, I'm going to present several ways
to make such two-way communications between Ruby & Go language.
Richard Lee • CTO @ Polydice, Inc • iCook.tw dev
& ops • Open source lover • @dlackty everywhere
Agenda 1. Why Go language? 2. Go with Ruby using
asynchronous workers 3. Remote Procedure Code 4. Advanced tools
Go Why choose another language when you have Ruby? 1.
Static typing 2. Compiled language 3. Higher level concurrency abstraction (goroutines) 4. No object oriented design
Use cases 1. Command line tools - Heroku's hk 2.
System softwares - Docker 3. DevOps tools - ngrok / packer
Heroku's CLI benchmark ## version $ time heroku version >/dev/null
real 0m1.813s $ time hk version >/dev/null real 0m0.016s ## list $ time heroku apps >/dev/null real 0m3.830s $ time hk apps >/dev/null real 0m0.785s
You don't really understand a language until you use it
in production
Then you might want to use Go in your system
with Ruby
My First Law of Distributed Object Design: Don't distribute your
objects — Martin Fowler
Async processing Use Redis or any message queue based processing.
Resque / Sidekiq friends: • From Ruby to Go • go-workers • goworkers • go-sidekiq • From Go to Ruby • go-resque
go-workers for Resque in action func myFunc(queue string, args ...interface{})
error { fmt.Printf("From %s, %v", queue, args) return } func init() { goworker.Register("MyClass", myFunc) }
...and in Resque class MyClass @queue = :myqueue end Resque.enqueue
MyClass, ['hi', 'there']
Async processing (cont'd) • Pros: • Not fast enough but
quite reliable • easy to scale out • Cons: • Async • Additional setup for queues
Be careful about typing // Expecting (int, string) func myFunc(queue,
args ...interface{}) error { id, ok := args[0].(int) if !ok { return errorInvalidParam } name, ok := args[1].(string) if !ok { return errorInvalidParam } doSomething(id, name) return nil }
Compared to Ruby code class MyClass @queue = :myqueue def
self.perform(i, str) doSomething(i, str) end end
Performance boost • Benchmarked using matrix multiplication a = Matrix[...]
b = Matrix[...] c = a * b puts a * b • 9x faster than Resque • 4x faster than Sidekiq
In some cases, you might need immediate result Then you
need RPC
Remote Procedure Calls Just like function invokation but remotely •
Dynamically-typed • JSON-RPC • MsgPack-RPC • Statically format • Protobuf • Thrift
Dynamically-typed Which means: 1. No need to predefine data format
2. No generated codes for both client & server 3. Quite easy to migrate
JSON-RPC 1. Everybody love JSON! 2. Golang has built-in library
3. Ruby has nice wrappers
JSON-RPC Protocol Can be built on simple TCP connection or
HTTP Request: {"method": "echo", "params": ["Hello JSON-RPC"], "id": 1} Response: {"result": "Hello JSON-RPC", "error": null, "id": 1}
It's demo time.
Client sock = TCPSocket.new "localhost", "5566" call = { method:"RPCMethods.Say",
params:[{"text"=>"Hello, world!"}], id:rand(100) } sock.write JSON.dump call JSON.load sock.readline
Server func (m *RPCMethods) Say(args Args, results *Results) error {
results.Text = args.Text fmt.Println(results.Text) return nil // no error } func (m *RPCMethods) Ping(args Args, results *Results) error { results.Text = "Pong" fmt.Println(results.Text) return nil // no error }
MsgPack 1. MsgPack is efficient binary serialization format 2. Quite
similar to JSON but with binary type 3. Nice packages for most language
MsgPack-RPC 1. MsgPack-RPC is ...RPC using MsgPack 2. Again, nice
packages for different languages
It's demo time.
Client code cli = MessagePack::RPC::Client.new("127.0.0.1", 5566) cli.timeout = 5 v
= cli.call(:echo, "Ruby Conference Taiwan!") cli.close
Statically-typed Which means: 1. Predefine format with special language (IDL)
2. Usually with generated codes 3. Good when you have nice plan
Protobuf 1. Widely used by Google 2. Built in with
Golang 3. However, there's no official RPC mechanism
Thrift 1. Open sourced by Facebook 2. Officially support a
wide range of language 3. Have built in RPC mechanism
Thrift IDL RpcService.thrift namespace go demo.rpc namespace ruby demo.rpc enum
TweetType { TWEET, // 1 RETWEET // 2 } struct Tweet { 1: required i32 userId; 2: required string userName; 3: required string text; 4: optional Location loc; 5: optional TweetType tweetType = TweetType.TWEET }
Thrift in aciton 1. Usually your first start from defining
you service in IDL 2. Generate server / client code using thrift commands 3. Copy that to both side and integrate
But wait, why not just use REST?
RPC v.s. HTTP / RESTful APIs • Persistent connection •
Smaller in size • Easier to implement
Conclusion
Most importantly, learning a new language is always fun and
inspiring
Thank you! Let's Keep in touch Twitter / GitHub: @dlackty