Deployment Considerations

Deployment of the components for web application hosting depends on the capabilities of the cloud service(s) which are chosen. The DNS and CDN usually live in the public network - these are typically purchased as services from a suitable provider. For IaaS:

- The Firewall and Load Balancer are deployed in the cloud service. Many cloud service providers have Firewall services available, either in the form of specialized hardware devices or in the form of dedicated systems running suitable software. Multiple redundant Firewalls can be deployed to avoid a single point of failure, each handling a unique IP address, where the capabilities of DNS servers to map one URL to multiple IP addresses can be used. The Load Balancer can be run on one or more separate servers within the cloud service, or it can run on the same system(s) as the Firewall.

- It is typical for the Web Application Servers to be run in multiple instances, all serving the same web pages. This can be both to deal with the expected throughput of requests and also to provide resilience against failure of a single instance. The location of instances can be a consideration, where physically remote instances can help deal with problems that affect a single data center.

- The User Directory, the Cache and the File Repository are run as instance VMs in the IaaS service, with redundancy and failover for each.

- The databases are run as VMs, with multiple instances and replicated data stores for capacity and resilience purposes.

In the case of a typical PaaS platform, the Firewall and Load Balancer are part of the platform and simply require configuration. The Web Application Servers are also part of the platform - the application code must be loaded onto them, but the running of multiple instances is handled by the platform and all that is required is configuration of the number and location(s) to use. Similarly, the PaaS usually supplies the database as a service, and it is common for the database service to provide replication of data and scalability of instances.

Regardless of where components are deployed - public, private or hybrid – lifecycle, operations, and governance requirements need to be considered and addressed. Where components are deployed will strongly affect how management and governance are done. Private deployments may be able to use existing internal management and governance tools if they have access to the cloud infrastructure. For public, hybrid, and externally hosted private deployments, lifecycle operations – instantiate, initiate, terminate – for components outside the firewall need to be negotiated with the hosting parties.