ms-meets-nosql-examples

This is the repository for the code examples shown in my Microservices meets NoSQL Talk

Run it yourself

1. After checking out the repository install all node dependencies using npm install
2. For the database part you need to install the following databases in their default configuration, please follow their instructions:
3. Arangodb Note: The app has been written for ArangoDB 2.6 which is currently in alpha version.
4. Mongodb
5. Neo4J in neo.js please adjust username/password
6. Deploy the Foxx App to arangodb with foxx-manager install foxxapp /arangodb
7. Install HAProxy
8. Start HAProxy with haproxy -f haproxy.cfg. If you want to modify the config and reload it use haproxy -f haproxy.cfg -st $(<haproxy.pid).
9. Fire up all node processes:

node app.js -p 3000 &
node cpuService.js -p 3001 &
node cpuService.js -p 3002 &
node cpuService.js -p 3003 &
node cpuService.js -p 3004 &
node memService.js -p 3005 &
node productService.js -p 4000 &
node productServiceArango.js -p 4001 &
node neo.js -p 4002 &
node mongo.js -p 4003 &

Now you have the microservices running. The first route available to you is http://localhost:9000/app for the micro-service only example. Try to replace app.js with monolith.js here. The behaviour is identical, but monolith is harder to scale. The other route available to you is http://localhost:9000/products/{id}. This one switches between ArangoDB and MongoDB & Neo4J setup. You can notice the difference by the printed internal ids.

Example Data

Disclaimer: This dataset is neither meant to represent any real world use case, nor did i introduce any indexing in the databases. It is only a simple showcase to transport my idea.

MongoDB

Connect to MongoDB with the mongo shell and execute the followng commands:

unix> mongo

db.createCollection("products");
db.products.save({id: "098", item: "Product1"})
db.products.save({id: "765", item: "Product2"})
db.products.save({id: "432", item: "Product3"})

db.createCollection("users");
db.users.save({id: "123", name: "Alice"})
db.users.save({id: "456", name: "Bob"})
db.users.save({id: "789", name: "Charly"})

Neo4J

Connect to Neo4J's webinterface at http://localhost:7474 There execute the following cypher statement:

CREATE p = (p1:Product {id: "098"})<-[:Bought]-(f1:Users {id: "456"})<-[:Friend]-(me:Users {id: "123"})-[:Friend]->(f2:Users {id: "789"})-[:Bought]->(p2:Product {id: "765"})
RETURN p

ArangoDB

Connect to ArangoDB with arangosh (the arangodb shell) and execute the following statements:

unix> arangosh

var gm = require("org/arangodb/general-graph");
var g = gm._create("ecom", [gm._relation("relations", ["Users", "Products"],["Users", "Products"])]);
db.Products.save({_key: "098", item: "Product1"})
db.Products.save({_key: "765", item: "Product2"})
db.Products.save({_key: "432", item: "Product3"})

db.Users.save({_key: "123", name: "Alice"})
db.Users.save({_key: "456", name: "Bob"})
db.Users.save({_key: "789", name: "Charly"})

db.relations.save("Users/123", "Users/456", {});
db.relations.save("Users/123", "Users/789", {});
db.relations.save("Users/456", "Products/098", {});
db.relations.save("Users/789", "Products/765", {});

API calls

This calls can be used to test the services we have just defined.

First of all Monolith vs. Microservices: To contact the Monolith use curl http://127.0.0.1:2000/app

To contact the Microservice version use curl http://127.0.0.1:9000/app Note: The only difference here is the port to access the process. This means we can replace a running monolith with a microservice version without the users noticing. The result will be identical with respect to the random number at the end.

In order to compare the performance you can use the apachebench tool which will execute several requests agains the endpoints. If you do so play around in the microservice version by starting/stopping processes and look how it effects the performance, or even try processes on different machines. Remember to adjust haproxy.cfg

ab -c 10 -n 10000 http://127.0.0.1:2000/app
ab -c 10 -n 10000 http://127.0.0.1:9000/app

Now lets come to the NoSQL services. You can directly access the MongoDB and Neo4J version using curl http://127.0.0.1:4000/products/123 | json_pp

And the Foxx application with: curl http://127.0.0.1:8529/arangodb/products/123 | json_pp

Here the results are also identical but includes the internal attributes. These are included on purporse to see in the Load Balanced version which database has actually returned the result. Now trigger any of these databases through the Load Balancer, they should nicely alternate: curl http://127.0.0.1:9000/products/123 | json_pp