No More Tiers

When planning a programming project, it helps to have a high-level, abstract plan to simplify things. The nice thing about planning server architecture is that it is somewhat standardized. You can visualize the project as a series of tiers. We’ll go over them here

Ground Rules

  • If someone else does it better, use it - It is usually more cost effective to have third-parties handle services, such as outgoing mail, incoming mail, monitoring, metrics, etc.
  • If you can generalize, do it - Infrastructure as a Service (IaaS) providers can be very effective, but make your infrastructure plan in a way that you can easily move from one provider to another for a given service. The challenge here usually stems from high level services (DNS, CDN, Private Networking, etc.). Stick to internal standards, and have a major plan if possible for major IaaS providers.

Public DNS Tier

The public DNS tier is usually reserved for production and production-like environments. Generally, this tier is used in concert with the CDN tier, as many CDN providers have discovered they can use their resources in speed/redundancy for DNS as well.

CDN Tier

Again, the CDN tier is usually reserved for production and production-like environments. Generally, you will want to optimize for loading time and bandwidth savings at this level.

Origin Tier

The origin tier is usually a load balancer, the layer that terminates HTTPS traffic. It may not though, if you require full security within the private network. It is useful to have your non-production environments start here to ensure that load balancing is working correctly.

Web Tier

In some cases, the Origin Tier can be merged with this layer, because most major load balancers are reverse HTTP/S proxies and can be used as such. However, some organizations might find it useful to operate their own Web Tier to ensure that actions such as redirects and custom HTTP -> APP connectors (mod_jk for example) are used.

App Tier

This usually comprises of an application server that is running the code for the application(s) that your devs will be developing. Generally, a local test environment will start here so developers can rapidly verify their changes.

Cache Tier

This tier usually consists of a caching application (memcache, redis, etc.) that can keep session data even if Application servers are removed/added, ensuring that a user’s ephemeral data stays on the application as long as possible. It makes sense to keep short term backups of cache data in case of failure.

Data Tier

This is often regarded as the most important tier and most impossible to keep stateless. This is usually a full relational database, although newer applications may use more simplified models (MongoDB, etc).

Service Tier

This is generally considered as more of an ephemeral tier, as it covers a litany of side services, such as:

  • Outgoing Mail
  • Monitoring
  • Log processing You will usually want to use third-party services where possible and cost-effective

Message Tier

This is another ephemeral tier. In larger groups of applications, a messaging system such as ActiveMQ, RabbitMQ, etc. is used as a buffer between services. This is generally utilized to improve redundancy by allowing messages from a higher tier to a lower one to be queued and kept in place if the lower tier is unavailable.

Published: September 13 2015

blog comments powered by Disqus