Application scaling is easy, except when it isn't

Melpomene - Fotolia


IT application support demands hard ops data and artful finesse

The secret to successful application resource management is coordinated capacity decisions between developers and operations -- but also the application owners.

Let's be honest -- good planning between applications and infrastructure is just planning. The real test of how well development prepared a deployment is when they turn the workload over to the application operations support team to handle day-to-day events.

Editor's note: If you're still in the planning stages for an application update or new deployment, read how to get the applications and infrastructure teams on the same page to evaluate resource needs and scalability.

Operations has a much bigger part in an application's lifecycle once it is up and running. When IT application support is done properly, ops can help to reclaim resources and provide more efficient infrastructure to dev's applications. IT monitoring is a key operations goal and not just to find faults. Resource usage trend monitoring is vital. Usage trends from IT monitoring systems establish the consumption patterns of key applications and can negate any or all predeployment application design and planning. Don't ignore the facts coming from your monitoring systems. While ops crunches the numbers ahead of time, that's no guarantee of accuracy. Adjust designs based on the data that IT application support technicians receive in production. Live apps are not static, so the underlying infrastructure must be fluid in its design and operations.

How to create flexible IT resources

You can increase resources assigned to an application quickly and without pushback from the app owners. Going in the opposite direction is a challenge.

Stakeholders will offer every excuse in the book -- and many more invented right on the spot -- to prevent application operations support teams from removing memory or CPU from the app's server. Why change something that is not broken? Because a move that frees up resources without hindering performance benefits the overall IT deployment -- operations' prerogative.

Stakeholders will offer every excuse in the book -- and many more invented right on the spot -- to prevent application operations support teams from removing memory or CPU from the app's server.

The first step of IT application support is to do no harm. Application owners expect any resource reduction to cause harm, so pare away excess with caution:

  • Don't start altering an application's resources without accurate information. Perform trend monitoring over time, not just a few hours or days. Application peak usage can show daily, weekly or even monthly spikes in use, so ensure you have enough data to make an informed decision.
  • Never vary resource allocations while another change, such as an update or patch, is scheduled. If anything goes wrong with the application, its team will immediately see the decrease in resources at fault. The software change, patch or update will be the absolute last thing that the application owner considers; they may simply refuse to look at those effects until resource allocation returns to original levels.
  • Test actual consumption with the help of resource limits in the virtual infrastructure. Resource restriction on the virtual plane lets IT application support teams try before they buy an adjustment. It's easy to restrict or free up virtual resources, backing out quickly if problems start to occur. Virtual restrictions also generate a data trend of expected application performance with the new settings in place.
  • Be flexible and work with application stakeholders. All the trend data and reports might show that the app is overallocated and needs a cutback – sometimes, you still can't do it. Chargeback options, or simply internal politics, factor into IT application support decisions. You can set soft limits and restrictions in virtualized environments, where they will not affect what the application owners see inside their virtualized guests.

Even the savviest administrators and IT operations personnel struggle to reclaim resources during IT application support in production, because the task is not purely technical. Hard data helps make the case for scaling back an application's available hardware, but blunt force won't get you far. Show why this adjustment benefits the application owners: They could gain more servers for other projects. Highlight the positive results of resource modification and encourage application owners to take ownership of reclamation efforts. Then everyone wins.

Next Steps

Modern IT infrastructure finds ways to waste money

What do hyperscale cloud providers do differently?

Sprawl management among top 10 trends in IT

Dig Deeper on Matching IT Resources to Application Requirements