DYNAMIC INFRASTRUCTURE MODELS

CHAPTER TWO: RESOURCE ALLOCATION

In this second edition of the Dynamic Infrastructure (DI) Series I’m going to examine some very basic forms of Automation. They might seem oversimplified for some of you, but they are important to consider if you’re aiming to get a taste of a DI model without breaking/re-writing too many Change Control Policies.

Going head to head with your service delivery manager isn't something I can help you with.

The key word is Allocation. We are not Dynamically Provisioning or Dynamically Bursting (at least, not this week); two models that induce wariness and scare off many from the benefits of adopting DI. Dynamic Allocation involves only the re-prioritisation of physical resources. Nothing new is introduced to the mix.

1 - Resource Allocation: Overflow

A process whereby all 'potential' Virtual Machines are ready-provisioned and running.

However, a small percentage of hardware is dedicated to the overflow requirements of existing service workloads. The overflow hosts are a cluster of servers that maintain copies of ALL the provisioned services (Virtual Machines) but exist in an idle state. This dedicated overflow hardware is only accessed when peak thresholds are reached triggering the need for greater resources.

This method relies on rules being put in place whereby only a specific number of Services/Applications can overflow at the same time, ensuring that the overflow hardware doesn't become a point of contention.

2 - Resource Allocation: Tiering

Tiering is simply the re-prioritisation of services and their access to available resources. It’s just like buying a standby ticket from an airline. You will get the seat you paid for as long as none of the airline's more ‘important’ (higher paying) customers need it.

For those boating inclined, consider the lower priority services as sacrificial anodes. For those not boating inclined, here's what the guys over at http://www.sacrificialanode.co.uk/  have to say:

"Metals on our keel, rudder, skeg, skin fittings and various other components, are all subject to corrosion. By connecting a metal of higher reactivity to these components, the metals can be protected. This is where the commercial anodes come into play. The metal which corrodes first is called the sacrificial anode."

Resource Tiering permits a sacrifice of CPU/RAM/Disk IO access to lower priority service so that higher priority services don't suffer...or sink your boat.

http://www.youtube.com/watch?v=GhrRlKcw3sI

Technorati Tags: Application Delivery, Dynamic Infrastructure, f5, Next Generation Datacenter