Skip to content

Latest commit

 

History

History
73 lines (58 loc) · 4.01 KB

load-balancers.md

File metadata and controls

73 lines (58 loc) · 4.01 KB
Raw

Load Balancers

Distribution of tasks over set of resources.

Fundamental feature: distribute incoming requests over cluster of backend servers according to a scheduling algorithm

Features:

  • acts as reverse proxy - not expose IPs of servers
  • increase performance - faster response
  • no single point of failure
  • scability
  • increase reliability - fault tolerence (makes possible to continue operating despite failures or malfunctions)
  • better security
  • perform continous heath checks of servers (Two kinds: 1. Shallow: pings servers 2. Deep: health and status of server)
  • caching and compression

Types of Load Balancers - based on function:

1. Network Load Balancer / Layer 4 (L4) Load Balancer:

  • 4th OSI layer: only [IP address + port] known

  • How it works: 1. client send request to balancer (request: client IP -> balancer IP) 2. balancer choses a server according to a scheduling algorithm 3. balancer changes the addresses of the request to (request: balancer IP -> server IP) with NAT 4. balancer sends the request to the server 5. connect made

  • single TCP connection made btw client and server (client->balancer->server)

  • no data reading - only passes packets of data to destination according to IP address

  • Pros:

    • simpler
    • efficient / faster
    • one TCP connection - one chain btw source and destination

  • Cons:

    • no smart load balancing based on data type - because no data reading
    • not for microservices - xyz.com/service1 and xyz.com/service2 can be completely different services
    • no caching - because no data reading

  • Implement with: HA Proxy

2. Application Load Balancer / Layer 7 (L7) Load Balancer:

  • 7th OSI Layer: autherized to see the data

  • How it works: 1. client sends a request to balancer 2. balancer decrypts data 3. balancer picks a server logically assigned to deal with that data 3. balaner encrypts and makes a connection request to the server (with maybe new headers) 4. connection made

  • two connections btw client and server (client->balancer and balancer->server)

  • Pros:

    • smart load balancing - look at data and send it to logcally assigned servers
    • great for microservices
    • caching possible

  • Cons:

    • expensive
    • decrypts

Load Balancing Algorithms

  • Round robin - requests are distributed across the group of servers sequentially
  • Least connections - request sent to the server with the fewest current connections to clients
  • Least Time - requests set to the server selected by a formula that combines the fastest response time and fewest active connections
  • IP Hash - hashing of IP decides which server (needed when requests of same client needs to go to single server)

Real-life implementation

  • Good idea to have a back-up load balancer (so balancers don't become the single point of failure) using floating IPs. One active at a time.
  • Load balancers can be placed between: a) the user and the web server b) web servers and application servers/cache servers c) application servers and database

More to know

  • DNS load balancing
  • Direct routing
  • IP tunneling
  • Weighted

References