Toast Masters

Become a toast master!

There are three kinds of toast machines, each for a different kind of toast.

	Plain	Chocolate	Strawberry
Service Id	1	2	3
Process time	3s	2.5s	1s

Note: Each kind of toast machine have different process time.

Architecture

Project Objectives

Main objective: Practice applying horizontal scaling and load balancing strategies.

Side effect: Become a toast master!

Challenges

Part 1 (Low Concurrency Test): Choose a Suitable Load Balancing Strategy

Test description: 50 requests in total, with 9 requests at the same time at most.

Each toast machine is limited to processing 3 toasts (requests) concurrently, if you go over this number, then the toast machine will burn out and produce failures.

For this part, you will be restricted to 1 replica of each kind of toast machine.

Please adjust the load balancing strategy through nginx.conf in order to successfully process all of the toasts without failures.

Examples

Round Robin (default)

Evenly dispatch to each server.

http {
  upstream h1servers {
    server plain:10000;
    server chocolate:10000;
    server strawberry:10000;
  }
  server {
    listen 80 default_server;
    listen [::]:80 default_server;
    location /h1 {
      proxy_pass http://h1servers;
    }
  }
}

Weighted Round Robin

e.g. for every 6(3+2+1) request, plain will get 3, chocolate will get 2, strawberry will get the last one.

http {
  upstream h1servers {
    server plain:10000 weight=3;
    server chocolate:10000 weight=2;
    server strawberry:10000 weight=1;
  }
  server {
    listen 80 default_server;
    listen [::]:80 default_server;
    location /h1 {
      proxy_pass http://h1servers;
    }
  }
}

Least Connection

Choose the server that currently holds least connections.

http {
  upstream h1servers {
    least_conn;
    server plain:10000;
    server chocolate:10000;
    server strawberry:10000;
  }
  server {
    listen 80 default_server;
    listen [::]:80 default_server;
    location /h1 {
      proxy_pass http://h1servers;
    }
  }
}

Read more: https://nginx.org/en/docs/http/load_balancing.html

Part 2 (High Concurrency Test): Experiment with Different Replica Settings

Test description: 50 requests in total, with 50 requests at the same time at most.

Process 50 toasts (requests) in the shortest amount of time with the least failure (failed requests).

You can adjust the load balancing strategy and number of replicas of each kind of toast machines freely.

Click to see the leaderboard

Score formula: Time taken to process 50 toasts (requests) + number of failures (failed requests as penalty)

The lower the score, the better you are!

Getting Started

1. Clone this repository

git clone https://github.com/candy02058912/toastmasters.git

2. Test on local machine

Prerequisites: Docker should be installed

scripts/deploy_swarm.sh: For deploying the service

Usage

scripts/deploy_swarm.sh

scripts/scale_swarm.sh: For setting replicas for each service

Usage

scripts/scale_swarm.sh [-s <service id>] -r <number of replicas>

Example

# set service 1 to 1 replica and set service 2 to 2 replicas
srcipts/scale_swarm.sh -s 1 -r 1 -s 2 -r 2

# set all services to 3 replicas
scripts/scale_swarm.sh -r 3

scripts/run.sh: Run ApacheBench for load testing

Usage

# Test 1
scripts/run.sh -t 1

# Test 2
scripts/run.sh -t 2

# Test 1 & 2
scrupts/run.sh

Monitor the service:

docker stats
CONTAINER ID        NAME                                          CPU %               MEM USAGE / LIMIT     MEM %               NET I/O             BLOCK I/O           PIDS
723109f3b483        demo_strawberry.1.yugi4i6j8g07j4ak09vq2uc2o   0.00%               1.16MiB / 5MiB        23.20%              1.78kB / 0B         0B / 0B             5
c3a277e1093c        demo_plain.1.728f24hul2e0177auwfqp4rhk        0.00%               1.148MiB / 5MiB       22.97%              1.78kB / 0B         0B / 0B             6
379b48c1fe86        demo_chocolate.1.l3980tsvgqp8araf8c2y7f6tl    0.00%               1.145MiB / 5MiB       22.89%              1.78kB / 0B         0B / 0B             6
2c17680b2ab1        demo_nginx.1.iq8w5mabcum8yuitskt69y89w        0.00%               4.418MiB / 7.779GiB   0.06%               263kB / 335kB       0B / 0B             5
860ed21cbdb1        demo_tester.1.uf5u6822lowrtll2wry2hikmy       0.00%               296KiB / 7.779GiB     0.00%               2.07kB / 42B        0B / 0B             1

The metrics you should take a look:

Time taken for tests:   56.109 seconds
...
Non-2xx responses:      49

Edit src/nginx/nginx.conf

Edit src/nginx/nginx.conf for adjusting load balancing strategies, please run scripts/deploy_swarm.sh when the file is changed, documentation: https://nginx.org/en/docs/http/load_balancing.html

4. Submit Project

Files involved: nginx.conf, scale.conf

• Edit nginx.conf for your desired load balancing strategy

• Edit scale.conf with your desired number of replicas

• Open a Pull Request, documentation: https://docs.github.com/en/github/collaborating-with-issues-and-pull-requests/creating-a-pull-request

• Once submission finished successfully, you will receive a comment in your PR

Example

Low concurrency test:

Server Software:		nginx/1.17.8
Server Hostname:		nginx
Server Port:		80
Document Path:		/h1
Document Length:		Variable
Concurrency Level:		9
Time taken for tests:		26.097 seconds
Complete requests:		50
Failed requests:		0
Non-2xx responses:		34
Total transferred:		13633 bytes
HTML transferred:		6081 bytes
Requests per second:		1.92
Transfer rate:		0.51 kb/s received
Connnection Times (ms)
	min	avg	max
Connect:	0	0	2
Processing:	0	2195	12633
Total:	0	2195	12635

High concurrency test:

Server Software:		nginx/1.17.8
Server Hostname:		nginx
Server Port:		80
Document Path:		/h1
Document Length:		Variable
Concurrency Level:		50
Time taken for tests:		5.517 seconds
Complete requests:		50
Failed requests:		0
Non-2xx responses:		47
Total transferred:		15108 bytes
HTML transferred:		7517 bytes
Requests per second:		9.06
Transfer rate:		2.67 kb/s received
Connnection Times (ms)
	min	avg	max
Connect:	0	2	3
Processing:	447	771	3011
Total:	447	773	3014