Introduction to Microservices

In

Introduction

  • Characteristics

    • decentralized - no unifying schema
    • polyglot - heterogenous approach
    • independent - they don’t affect each other
    • do 1 thing well - if it is too complex, split it
    • black box - each microservice is a box, communication is done via API
    • you build it, you run it

  • Benefits

    • Agility - different teams own different services
    • Innovation - any in any language
    • Quality - gets POO advantages: reusability, composability, maintenaibility
    • Scalability - vertical (bigger machines) horizontal (more machines) + change failing components
    • Availability - easier failure isolation

  • Challenges

    • distributed systems (bandwith, network usage)
    • migration (define patterns)
    • versions (define pipeline release)
    • organization (build effective team structure)

Migration to Microservices

  • Brownfield applications are common (old not build with latest approaches)
  • Use structure of web application (individual URIs) to divide them in functional domains, and replace those domains with new microservices
  • Strangler application patterns: a vine strangles the tree that it is wrapped around
    • Strangler application steps:
      • 1- Transform: create paralel site
      • 2- Coexist: leave the old one, but redirect every new functionallity as it is implemented incrementally
      • 3- Eliminate: remove old functionality
  • Pre-requisites
    • Web or API based monolith: base on URL structure
    • Standarized URL structure: careful with intermediate layers
    • Meta UIs: larger chucnk so UI constructed on the fly
  • Bad patterns to apply
    • No 1 page at a time: it may result on consitency problems: the smallest sliver is a microservice
    • All at once
  • Good patterns
    • Refactor back (Inside part)
      • identify bouned contexts in application design (domain driven design)
      • chooe the smalles bounded bounded context, easier to refactor
      • conceptual plan of microservices within the context (rough URL structure)
    • Refactor front to accommodate the microservices on back (Outside part)
      • analize relationships between existing screens
      • apply principle of least astonishment to the aspects of model manipulation
      • conceptually plan the microservices within the context
      • choose wether to release an entire chunk at a time, or each chunk as a series os slivers
  • Relase process (Agile)
    • MVP (Minimal Viable Product) -> User Stories
    • User Stories may be grouped in EPICS (high level fencionality)
    • Squads implements EPICS (samlles element of implementation)

Complexity

  • Architectural
Model Complexity Meassure
Monolithic Code complexity Dependencies
Microservices Complexity in interactions Interaactions of the individual services’ domain
  • Operational
    • scaling + cost-efficient
    • operate 100 microservice components without multuplying effort
    • keeop track of pipelines
    • monitor system health
    • track debug interactions
    • analyze high ammounts of log data in distributed applications
    • deal with lack of standards
    • value diversity without locking technologies
    • versioning
    • ensure are in services use without consitent pattern usage
    • ensure decoupling abd communication

Microservices and the Cloud

  • AWS: Cloud services provider
    • Advantages
      • High availability (access from any device)
      • Fault tolerant (version storage)
      • Scalabiliy (add/grow services easily)
      • Elasticity (remove/shrink services easily)
    • Services provided
      • Amazon S3 (large “unlimited” storage)
      • Amazon EC2 (computer power) - part of VPC, the user connects here
      • Amazon RDS (databases) - part of VPC
    • Resolves most important challenges
      • On demand-resources
      • Programmability
      • Experiment with low cost and risk
      • Infrastructure as code
      • Continuous delivery
      • Service orientation
      • Managed services
      • Poliglot

Simple Microservices Architecture

Split functionalities into cohesive “verticals”

User Interface Microservices Datastore
Cloud front Application load balancer ElastiCache
(Static Content) ECS RDS
S3 DynamoDB

User Interface (CDN and WAF)

  • Route 53 provides features that can be leveraged for service discovery (user -> DNS resolver (doesn’t have the user) -> Route 53 (authority))

    • Adds health checks and simple failover recovery (if health check fails, activate secoundary server, either calculated or with latency measurement health checks)
    • Routing policies: multiple records in a single DNS (different weight)

  • Virtual private clouds (VPCs)

  • Cloud front is global Content Delivery Network that accelerates delivery of websites, APIs, video content… (example: on JS calls, images and CSS are on CDNs)

    • Fastest access
    • Reduce risks
    • Possibility to cache contents in an easy way
    • Use of CDN is an standard among companies with selling capabilities on Internet
    • Empower possibiliy to use HTT/2 which is much performant protocol adapted to smartphone and devices bandwith consume and features pattern

  • WAF (Web Application Firewall) - control traffic + custom rules for common attack patterns + API to handle rules, AWS WAF payment: price depends on number of rules.

    • Part of CDM solutions or Application Load Balancer (ALB) that fronts web servers or origin servers running on EC2.
    • Increased protection against web attacks
    • Security integrated with how you develop applications
    • Ease of payment and maintenance
    • Improved web traffic visibility
    • Cost effective web application protection

Microservices

  • OSI model (remember)
    • 7: Application
    • 6: Presentation
    • 5: Session
    • 4: Transport
    • 3: Network
    • 2: Data-link
    • 1: Physical

graph LR
A[Amazon Route 53]
B[ELB 1]
C[ELB2]
D[EC2 instance 1]
E[EC2 instance 2]
subgraph ELB-VPC
  B
  C
end
subgraph Customer-VPC
  D
  E
end
A --> B;
B --> D;
A --> C;
C --> E;
  • Components used
    • API of a microservice: REST
    • ELB (Elastic Load Balancer) distributes traffic
    • detects which instances are online and distributes trafic among them, mitigates DDOS
    • Idle timeouts (60-36000 seconds): close connections by load balancer when they are not longer used
    • Listeners (1-10 per ELB) with routung rules (actions to forward requests, use path patterns formats)
    • Multiple availability zones: 2 or more subnets in deiiferent zones with the load balancer
    • ELB sends latency information to cloudwatch -> which will request auto-scaling(AS) via AS Policy -> which will send action to ELB
      • ELB metrics available
      • Healthy Host Count: healthy instances in each availability zone
      • Latency: elapsed time from when the request leaves the load balancer until the response is received
      • Rejected Connection count: when ELB can not establish connection with healthy target in order to route the request
    • ALB Application Load Balancer) -> level 7 OSI model: rules to redirect `+ target groups of VPCs
    • ECS (Container Service) + autoscaling
graph TD
A[Load balancer]
B[Listener]
C[Rule 1]
D[Rule 2]
E[EC2]
F[EC2]
G(Health check)
H(Health check)
I[Listener]
J[Rule 3]
K[EC2]
L(Health check)
subgraph Listener layer A
  B
  C
  D
end
subgraph Login
  E
  G
end
subgraph Img
  F
  H
end
subgraph Listener layer B
  I
  J
end
subgraph Payment
  K
  L
end
A --> B;
B --> C;
C --> E;
B --> D;
D --> F;
A --> I;
I --> J;
J --> K;

  • Key features

    • High availability
    • Health checks
    • Security groups
    • SSL Termination
    • Sticky sessions
    • VPCs
    • Idle connection timeout
    • Connection draining
    • Dynamic Por Mapping
    • Protocols
    • Backend server Auth (on classic, not ALB)
    • Cloudwatch Metrics
    • Access Logs
    • Path-based routing
    • Deletion Protection

  • EC2 (Amazon ECS) or Amazon EC2 Container Service: processing activities

    • Supports Docker containers and allows to easily run applications on a managed cluster of amazon EC2 instances

    • Scaled in/out, use Auto-scaling

    • Amazon EKS provides container services for instance orchestration via Kubernetes

    • Docker on AWS

      • Configuration and deployment
      • Microservices

    • Batch processing

    • Amazon ECS Container Agent: checks the clusters

    • Clusters: regional resource poll for grouping container instances, which start empty and can be scaled

    • Typical workflow

      • 1- User pushes an image to DockerHub
      • 2- User creates task definition on Amazon ECS (declare source requirements)
      • 3- User runs instances on EC2 (custom AMI with Docker support and ECS Agent. Instances will be registered with default cluster)
      • 4- User describes cluster on Amazon ECS (get information about cluster state and available resources)
      • 5- User runs task on amazon ECS (using task definition from step 2, which schedules it)
      • 6- User describes cluster again and checks the cluster has the docker image loaded

    • Ways to start a task

      • StartTask 
        1
        aws ecs starttask -cluster default -task-definition sleep360:1 -container-instance <instance arn>
      • RunTask 
        1
        aws ecs run-task --cluster default -task-definition sleep360:1 --count 1
      • Bring your own scheduler (Mesos, Marathon, custom)

    • Use of AWS CLI:

Action Example
List all clusters aws ecs list-clusters -- profile myProfile
List all container instances in cluster aws ecs list-container-instances --cluster default --profile myProfile
List all task definition aws ecs list-task-definitions --profile myProfile
List tasks running aws ecs list-tasks --cluster dafult -- profile myProfile
Describe task via taskArn aws ecs describe-tasks --cluster dafult -- profile myProfile -- tasks f94dbd87-7d84-4e27-ab70-0461d455d1ba
Describe container instance via containerInstanceArn aws ecs describe-container-intances --cluster dafult -- profile myProfile -- container_instances 25340-a6ff-45de-b3eb-fa43a88e9313
Run task aws ecs run-task --task-definion sleep360:2 --cluster default --profile myProfile
  • Amazon Elastic File System (Amazon EFS): storage for EC2, can be mouted AS-IS when connected to Amazon VPC (Virtual Private Cloud)
    • NFSv4 protocol, multiple EC2s can share it

Data storage

Persist data:

  • Amazon ElastiCache Service
  • Amazon Relational Database Service (MS SQL server, Oracle, MySQL, MariaDB, PostgreSQL, Amazon Aurora)
  • Amazon DynamoDB: NoSQL: no strict schema (can not join table, info must be merged in applications)

Learning about RDS

  • AWS proviisioned database storage
  • Amazon Dynamo DB
    • read consistency -> strong consistent
    • provisioned throughput capacity (read/write throughputs)

Tools

Springboot

Spring framework

  • Dependency injection and integration (@Component, @Autowire, @Context)

  • Super easy central configuration service creation (@EnableConfigServer creates new SpringBoot Application)

  • Straightforward service discovery registry (Registry = also a microservice, other services can register themselves)

    • Integration with Eureka REST service registry by Netflix
    • Spring Security: athentication and authorization
    • Spring Boot: top of the Framework: (SpringBoot = (Spring Framework) + (Embedded HTTP Servers) - (XML <bean> Configuration or @Configuration))

  • Benefits

    • Easy to develop Java or Groovy
    • Reduces dev time
    • Avoid boilerplate code, annotations, configuration
    • Easy integration with other Spring resources
    • Provides CLI to develop and test Springboot apps
    • Provides plugins to test, build (Maven, Gradle) or work with embedded DBs

Docker

“Docker doesn’t create containers, but it does containers for you making it easier by using an standard”

  • Linux server
  • Docker daemon
  • Login/interact DockerHub
  • Use Docker Client to connect to Docker Daemon

Example: go to DockerHub, find apache image,then send it to daemon which will send it to the Linux Kernel

1
docker run apache

  • Advantages

    • Uses LXCs (Linux Containers), which are user space interface for the Linux Kernel Containment
    • Several LXCs run in one control hist LXCs
    • LXCs are an alternative to hypervisors (Virtualbox, VMware…)
    • Prevents “Runs in your machine, but not in mine”

  • Benefits

    • Portability
    • Productivity
    • Efficiency
    • Control

Therefore, on Amazon AWS:

  • Amazon ECS eliminates the eed to install, operate and scale cluster management infrastructure
  • Use API to launch Docker-enabled applications, query the status…
  • Amazon EDS is scalable and elastic
  • Amazon EKS leverages and replaces ECS to be used under the context of Kibernetes orchestrator

Kubernetes

  • Introduction
    *Container orchestration to deploy containers

    • Portable
    • Extensible
    • Self-healing

  • Benefits

    • Deploy is fast and predictive
    • Scale on the fly
    • Roll out new features seamlessly
    • Limit hardware usage

  • Common needs satified: it provides simplicity of Platform as a Service (PaaS) with felixibility of Infrastructure as a Service (IaaS)

    • Co-location helper processes
    • Mounting storage systems
    • Distributing secrets
    • Checking application health
    • Monitoring resources
    • Using horizonatl pod autoscaling
    • Naming and discovery
    • Balancing loads
    • Rolling updates
    • Accessing and ingesting logs
    • Debugging applications
    • Providing authentication and authorization
    • Replicating application instances

  • Amazon EKS

    • Amazon Elastic Container Service for Kubernetes: install and manage Kubernetes clusters

      • Integrated with many AWS services
      • Elastic load balancing
      • IAM authen tication
      • Amazon VPC for isolation
      • AWS PrivateLink for private network access
      • AWS CloudTrail for logging

    • Benefits

      • Fully managed and highly available
      • Secure
      • Fully compatible with Kubernetes Community Tools