Load Balancing

What is load balancing

Transcript

1. Shit happens! Shit happens!
2. Don't be PARANOID! Don't be PARANOID! • Are you scared of deploying new Are you scared of deploying new features? features? • Are you afraid of deploying changes? Are you afraid of deploying changes? • Does it take months to review changes Does it take months to review changes and confirm for deployment? and confirm for deployment? • Does it take ages to be comfortable about Does it take ages to be comfortable about new features to deploy? new features to deploy? • Is it a headache to recover broken service Is it a headache to recover broken service and your service is totally down? and your service is totally down?
3. Don't rush! Don't rush! • How often it happens that your service How often it happens that your service was not accessible for hours because of was not accessible for hours because of simple hardware fault? simple hardware fault? • ٌٌWhat about software issues? What about software issues? • Did someone say storages? Did someone say storages? • Revenue loss? Revenue loss? • Unhappy management? Unhappy management?
4. Think to grow! Think to grow! • Are you scalable? Are you scalable? • Vertically? :-P Vertically? :-P • How often do you suspend your services How often do you suspend your services to upgrade your hardware? to upgrade your hardware? • What happens if in one week your service What happens if in one week your service grows by 50% or 100% or even more? grows by 50% or 100% or even more? • Expanding to new Geo-Location is a Expanding to new Geo-Location is a headache, isn't it? headache, isn't it?
5. You think you are You think you are fault tolerant fault tolerant & & Scalable? Scalable? REALLY? REALLY? Think again! Think again!
6. If any or all of the above is your If any or all of the above is your problem or you can not tolerate them, problem or you can not tolerate them, Then I tell you what is missing in your Then I tell you what is missing in your infrastructure! infrastructure!
7. Be prepared! Be prepared!
8. Load Balancer Load Balancer
9. Don't be PARANOID Don't be PARANOID • Deploy changes to only one node with Deploy changes to only one node with least clients least clients • Deploy new feature to one node with Deploy new feature to one node with least clients and more testers. least clients and more testers. • Changes or Features that cause problem Changes or Features that cause problem to the system only affect very small group to the system only affect very small group of clients. of clients.
10. Don't rush! Don't rush! • Service on one server is down? Take your Service on one server is down? Take your time to fix it while other handling the time to fix it while other handling the requests for you. requests for you. • Low on disks? Don't worry! You'll have Low on disks? Don't worry! You'll have enough time to buy an additional disk :) enough time to buy an additional disk :) • Hardware faults? It's OK! You'll have Hardware faults? It's OK! You'll have enough time to find and fix the problem. enough time to find and fix the problem.
11. Think to Grow Think to Grow • Expanding or shrinking service capacities Expanding or shrinking service capacities is no longer an issue. is no longer an issue. • If tomorrow requests grows by 100% If tomorrow requests grows by 100% simple double your servers and all would simple double your servers and all would be fine. be fine. • No more costumers at off pick time? Why No more costumers at off pick time? Why wasting money, shrink your service by wasting money, shrink your service by turning off extra servers. turning off extra servers.
12. Into the cloulds Into the cloulds AWS AWS ELB (Elastic Load balancer) ELB (Elastic Load balancer) GAE GAE Just works! Just works! Azure Azure Server Groups Server Groups And many more... And many more...
13. Own Server farms Own Server farms • LVS LVS • Barracuda Barracuda • Piranha Piranha • HAProxy HAProxy • NginX (Proxy) NginX (Proxy) • ... ...
14. Network Layer Network Layer • RIP RIP • OSPF OSPF • BGP BGP • ... ...