In theory it all promises to be so easy from now on. We don't need to plan because if we hit a problem, we just add in more capacity from this infinitely elastic Cloud. There will always be limits to that elasticity though. Don't get carried away with the hype, capacity planning is more vital than ever. You'll still be doing it in some form, it's just a question of how well you do it. All those good practice guidelines from ITIL and elsewhere are still true, we just have to apply them to a different environment - the Cloud, whether it be Public, Private or Hybrid. You just can't escape some truisms: for example, all businessmen know that at some stage revenue must exceed expenses. Likewise we still need to make sure we have the right capacity available, whether it is on a day to day basis or added in at the drop of a hat - this is capacity planning.
Capacity planning does need to evolve to support the Cloud. The technical, predicting what resources you will need and the financial, what and when you pay for them, need to be more closely aligned. Buying your ticket on the door for a popular event always costs more than buying in advance. With Cloud, that burst of extra resource will cost you dear, so you need capacity planning to minimize that hit. Cloud providers, external or internal, need to over-provision to cater for those bursts and someone has to pay for his. Planning what you need and when, as much as you can, is your way to make sure it isn't you. The alternative is that suppliers will use the airline model and over-book. Better from a cost perspective, but a risk factor that needs to be included in your plans. No one likes
turning up and finding out they've been bounced to the next flight - especially your users. 
We've all seen the films where the barman refuses to give the client another drink - 'time to go home now, you've had enough'. It's not often however that your IT supplier tells you when you've had enough. It's still down to you to make that call about your IT infrastructure. Only the SaaS(Software as a Service) model of cloud computing movestheresponsibility for capacity planning to your supplier. If you're buying a platform,infrastructure or an application as a service, it is down to you to make sure you ask for the right resources. Buy too much and you areperpetuating the over-provisioning mistakes of the distributed era. Buy too little and youbounce from performance crisis to performance crisis and pay the costs of short term fixes. Traditional capacity planning techniques allied to some new considerations are the way to make sure you only pay for what you use.
Being based around relevant service levels is one of the new elements that capacity planning must deliver. Hardly new in theory, as tight coupling of capacity planning and service level management has been a recommendation from ITIL for a long time. Cloud computing takes many factors previously within your control away, so what has been theory in the past now needs to be practice. Effective capacity management spanning complex environments of internal, external or hybrid Clouds needs to migrate from being a technical and tactical activity to become a strategic business-orientated process. The starting point is what level of service for a given business requirement, service or application is needed, then one needs to plan how best to support that from a capacity perspective, both now, allowing for future trends and changes. Coupling this consideration to cost factors will then help assess whether that capacity should be bought in traditional mode, in advance of demand, or 'just in time' using the elasticity of supply that Cloud offers.
How can you monitor resource usage and predict resource requirements
across
your
internal and external Clouds? |
|
 |
Achieving this at a technical level is more challenging. Businesses will need bothcapacity analysts and capacity tools that have a widerspan than has been the norm. Staff will need to be able to talk to the business about service requirements, whilst also being able to understand the technological options available and the relative cost implications of those options. On the technical front, tools will need to be able to integrate all data sources, not be 'silo specific'. They will also need to be able to incorporate that service level and cost data. Especially with public Cloud, it won't be possible to measure everything. Any good analyst will tell his patient not to worry about things outside his control. Measure what you can, manage that as effectively as you can, and use other measures such as contracts and service penalties to manage the things you can't measure. One thing that needs to be included as part of capacity planning and management that has often been neglected to date, is the workload view, be that job, application, service or other. Simple utilization views of resources will be insufficient when the unit that
you might want to move around the Cloud is the workload, not the technology
underneath such as processor, disc or network segment. |
Capacity management and particularly capacity planning move from the technical silo to become a more strategic business task with Cloud implementations. Rather than look at utilization of one technical area, it must be about aggregation and analysis. You will need to be aggregating data from the business, all your resource areas and cost and service requirements information. Service levels lose value if they are only used to get compensation after the fact - better to use them to manage a smooth and consistent service over time. No resource is infinitely scalable and even with the best and most flexible Cloud providers, it will still take time to configure in and utilize additional resource or cloud bursts, for example getting data into place. Good capacity planning will enable you to plan what resources you need and minimize those costly disruptions. IT provides a framework for managing the service you get from external suppliers and when necessary, assessing the value in switching from one to another. |
of keeping your feet on the ground
