KrakenD API Gateway Product overview @devopsfaith Microservices - - PowerPoint PPT Presentation

KrakenD API Gateway

Product overview

@devopsfaith

Microservices are challenging

The need for agility and scalability has driven Microservices adoption, but: 91% of Internet businesses are using or have plans to use microservices 86% expect microservices to be the default architecture within five years

* Independent research study conducted by Dimensional Research, April 2018

99% report challenges in using microservices

What problem do we solve?

Consume use-cases in a microservice-based ecosystem

Catalog Promotions Basket Payments Orders Pricing Stock Authentication

?

KrakenD offloads shared needs

Catalog Basket Orders Promotions Stock Payments Pricing Authentication Manipulation Filtering Circuit Breaker Metrics/Tracing Aggregation Security Authorization Service Discovery Encoding Logging Rate Limit Monitoring Load Balancer Pub/Sub Transport adapter Stub Data Traffic Mirroring Queues Serverless functions Lambda Bot detection I/O validation

Use cases

Common scenarios for KrakenD

MONOLITH

Database

Catalog Promotions Basket Payments Orders Pricing Stock Authentication

Transition to microservices

1 2

JWT

Authentication

3

Service 1

Aggregate

Service 2

"endpoint": "/basket", "extra_config": { "github.com/devopsfaith/krakend-jose/validator": { "alg": "HS256", "audience": ["http://api.example.com"], "roles_key": "roles",

"roles": ["user", "admin"],

"jwk-url": "https://monolith/jwk/symmetric.json" } }, "output_encoding": "no-op", "headers_to_pass": ["Authentication"], "backend": [{ "url_pattern": "/bar", "encoding": "no-op" }]

Authorization granularity

krakend.json

Authentication

/checkout JWT token

Catalog Promotions Basket Payments Orders Pricing Stock

MONOLITH

MONOLITH

Orders Pricing Stock Basket Payments Promotions Catalog

MONOLITH

🎊

Event-driven architecture

Notifications

/notify

QUEUE

4

Client Apps (consume a use-case)

Catalog Promotions Basket Payments Orders Pricing Stock Authentication

/frontpage { "catalog": {}, "promos": {}, "pricing": {} }

Aggregation

<id_product>2</id_product> <name>Wireless Headset</name> <date fmt="Y-m-d">2019-06-04</date> { "code": "PromoChristmas", "discount": 0.15, "products": [1,2,15] }

+

{ "id_product": 2, "name": "Wireless Headset", "date": "2019-06-04", "code": "PromoChristmas", "discount": 0.15, "products": [1,2,15] }

Aggregated

}

Catalog Promotions

KrakenD as a 3rd party provider

{ "3dparty": {}, "internal": {} } 3rd party services Company services

Transformation API

rename id_user => user_id rename alias => username remove password add static_data Services

Manipulation/Filtering/Grouping/Transformation

<id_product>2</id_product> <name>Wireless Headset</name> <date fmt="Y-m-d">2019-06-04</date>

Catalog

{ "code": "PromoChristmas", "discount": 0.15, "products": [1,2,15] }

Promotions +

{ "catalog": { "id_product": 2, "name": "Wireless Headset", "date": "2019-06-04", }, "promotions": { "code": "PromoChristmas", "savediscount": 0.15, "products": [1,2,15], } }

Aggregated

}

TL;DR - Instant creation of high-performance API

through a configuration file

More components and features… (open source)

https://github.com/devopsfaith/krakend-contrib

Enterprise components

• KrakenD Studio
• KrakenD Developer Portal
• KrakenD Management Portal
• KrakenD Enterprise Plugins
• KrakenD gRPC plugin generator
• KrakenD Dashboard

Differences

With similar products

Photo by @voyagefervor, Instagram

Service Mesh API Gateways Proxies with GW GraphQL API Managers

Proxy with GW 1:1 mapping endpoint-backends - No business logic - Offload cross-cutting

• concerns. No aggregation

Products with overlapping features

GraphQL HTTP only - Single Endpoint - Allows the client to choose exactly the data in the response. ~BFF. E.g: you provide an API to developers out of your

• rganization

API Gateway Services aggregation - Business logic - API Contract - No coupling to backend - Offload cross-cutting concerns. Can implement the BFF pattern. Service Mesh Internal communication between services (not for the end-user). No business logic API Managers Access management (generate API Keys), billing, developer portal, usage statistics

API GW API GW: North-south traffic Mesh: east-west traffic

Proxy with API Gateway capabilities

Catalog Promotions Basket Payments Orders Pricing Stock Authentication

PROXY

Stateless vs Stateful

Stateful

Stateless

A gateway is not the new monolith

★ Coordination required ★ Data synchronization ★ Datastore as source of truth ★ Complexity ★ Multi-region lag ★ Mutable configuration NON-LINEAR SCALABILITY

Stateless Stateful

★ No node coordination ★ No synchronization ★ Zero complexity ★ No challenges for Multi-region ★ Declarative configuration ★ Immutable infrastructure LINEAR SCALABILITY

DB-LESS != stateless

Deployment

Simple deployment (stateless)

FROM devopsfaith/krakend COPY krakend.json \ /etc/krakend/krakend.json

+ ≃

70MB

Deploy anywhere

Orchestration Platforms

Assign a KrakenD to each team (client type)

Catalog Promotions Basket Payments Orders Pricing Stock Authentication

Assign a KrakenD to each team (micro frontends)

} } }

Not necessarily the single point of entry

Catalog Promotions Payments Orders Pricing Stock Authentication

Performance

Benchmarks

Benchmark Overview

# Hardware specs Requests second Average response 1. Amazon EC2 (c4.2xlarge) 10126.1613 reqs/s 9.8ms 2. Amazon EC2 (c4.xlarge) 8465.4012 reqs/s 11.7ms 3. Amazon EC2 (m4.large) 3634.1247 reqs/s 27.3ms 4. Amazon EC2 (t2.medium) 2781.8611 reqs/s 351.3ms 5. Amazon EC2 (t2.micro) 2757.6407 reqs/s 35.8ms 6. MacBook Pro (Aug 2015) 2,2 GHz Intel Core i7 18157.4274 reqs/s 5.5ms

18.000 reqs/sec

On a raspberry 2

Without KrakenD

With KrakenD (concurrence=3)

Observability

Visualize the entire ecosystem from a central place

1-click export of logging, metrics and traces

Enable monitoring

Metrics and Tracer exporters for every taste

KrakenD in production

KrakenD Users

Open-source adoption

120,000 KrakenD sessions (12mo)

1 session = 1 cluster with at least 12 hours uptime

Takeaways

➔ Automatic API generation and integration ➔ Consumers (iOS, Android, Web, Server devs) in control

• f the API

➔ Homogeneous consumption

• f data formats and

encodings ➔ Reduced bandwidth and errors ➔ Increased speed ➔ Better quality of service

Questions?

Let’s have a beer!

@devopsfaith | @alombarte