Scalability and High Availability are most decisive

Besides a solid network infrastructure, the system concept is most important:

  • server redundancy: all important tasks run on multiple servers (e.g. loadbalancers, databases, backend and frontend servers)
  • component redundancy: multiple power supplies and RAID concepts per server
  • multiple power supply circuits
  • high level of security of easy to administer linux boxes
  • avoidance of any kind of "Single-Point-of-Failures"
  • external CDN (content delivery network) for failover, loadbalancing or geographic delivery
  • capacity bottlenecks can easily bypassed through adding any amount of virtual server instances
  • cloud computing connection