Blog

Proper Cloud Server Monitoring

February 08, 2018 | Ken Leoni

Cloud Server Monitoring is quite different from on-premises server monitoring and requires a fundamental change in mindset.  The situation is further complicated when working with hybrid cloud infrastructures or pursuing a multi-provider cloud strategy. Two key questions that need to be answered are...

Proper Server Monitoring

How does IT ensure they are delivering the desired performance for cloud services while at the same time operating within an appropriate cost structure?

What are the attributes of a proper cloud server monitoring strategy?



How are cloud servers different from on-premises?

Organizations continue to wrestle with what applications should move to cloud servers and which should remain on-premises. The differences between the two server deployment models impact the approach organizations take to server monitoring.

  • Cost - Cloud servers are pay-as-you-go; meaning you pay for only what you need and when you need it (when properly configured).

    Cloud costs are variable, because price is based on the fluctuating number of servers deployed over a period of time. Because costs are variable they can escalate quickly, especially if server workloads aren’t properly allocated or accounted for.

    On-premises/private cloud server costs are fixed as the server count is bounded by the physical capacity of the host(s) – capacity can only increase with more hardware. It is a much more difficult proposition for IT to obtain Capex for additional hosts than it is to get Opex to add more cloud server capacity.


  • Flexible and adjustable – The ability to quickly size and scale the number of cloud servers provides IT with access to a nearly unlimited amount of resources. However with great power comes great responsibility, while cloud computing comes with all kinds of flexibility it does not preclude someone from making poor sizing decisions that in turn result in costs getting quickly out of hand.

    Let’s consider auto scaling - the automatic adjustment of server capacity based on workload.  The ability to modify server counts and/or server resources makes the handling of volatile workloads a much easier and cost-effective proposition.

    However, one of the challenges with auto scaling is determining how many and what type of servers are needed for a given workload. There are multiple factors that come into play including properly sizing CPU, Memory, Network, and Disk.


  • Workload volatility matters - Applications where usage peaks and then subsides are likely to benefit most from a cloud deployment because the costs are dramatically reduced during lulls in usage. 

    However, applications with a more continuous workload may benefit less. It is the servers that run constant workloads where proper sizing of the resources is going to be critical to maximizing the economic benefit. Depending on the cloud service provider the organization could be making a long term (3 year) commitment to a specific instance type.



Attributes for proper cloud monitoring

Monitor with no discernible overhead – The gathering of key performance indicators (KPIs) is an integral part of any server monitoring initiative.  Monitoring should be non-intrusive, meaning there should be no discernible overhead generated by the monitoring technology in the targeted environment. The last thing an IT organization wants to do is implement server monitoring that adversely affects cloud or application performance. 

 

  • Target critical KPIs - It is important that any cloud server monitoring strategy understand what KPI’s to target, how often to target them, and what to evaluate as a problem. Ideally you will want to collect and evaluate different KPI’s at different intervals of time based on how critical and volatile the metrics are. 

  • Keep it agentless - Another characteristic that keeps system overhead to a minimum is to collect KPIs without deploying any additional technology on targeted servers -  this is accomplished by utilizing the most appropriate protocols to poll for critical KPI’s. An agentless approach also delivers the additional benefit of keeping IT administration to a minimum.

    Agentless doesn’t equate to lack of resilience.  In fact, a properly architected solution can reliably consolidate the monitoring of IT assets deployed across multiple cloud environments - be it public, private or hybrid. Resilience means guaranteed collection of KPI’s even when connectivity to a remote location is compromised.

 

 

 

Proactive and Intelligent Alerting -  The goal of server monitoring is to anticipate and then alert IT staff to server and application performance issues.  Keeping the information sent to IT relevant and timely goes a long way towards a server monitoring strategy that saves time rather than takes time.

Organizations can pay a heavy price for mis-sizing cloud resources - both in terms of lost productivity and revenue as well as the time and effort required to counteract any resistance towards future cloud initiatives. 

An even more costly problem occurs when cloud resources are over allocated, as even though the servers and accompanying applications are performing to expectations they are operating at a significantly higher and wasteful cost structure.

Over allocating server resources is a more insidious problem to detect because it requires either a bit of sleuthing by IT or the implementation of technology with intelligent alerting.

The problem can only be uncovered when actual resource usage is correlated with available cloud resources.

 

Longitude Dashboard  - Hybrid Cloud Infrastructure
Longitude dashboard showing hybrid cloud health



Capacity Planning
-  Collecting and correlating server performance behavior is an invaluable tool that helps predict what the future holds.

Server capacity planning is ultimately about getting advanced warning of resource problems that will not only impact the business, but also affect IT’s  reputation.  Remember, while many a success are taken for granted, it only takes a high-profile mistake to quickly destroy credibility. IT needs to be proactive about adjusting the IT infrastructure to accommodate changes in the business.

An advantage of cloud computing is the speed at which server resources can be acquired and deployed.  In the interest of expediency -  it can be all too easy to give in to a hunch or to simply  mirror an existing configuration as part of a deployment.

Proper capacity planning means looking at critical KPIs by going both wide (across multiple servers) and deep (multiple metrics within servers) and leveraging that information to determine the optimal resource allocation.

Capacity planning is never a one-and-done process as server workloads are constantly changing. While the easiest resource issues to spot are those that are sudden and immediately impactful, the more challenging diagnosis occurs when resource depletion problem is slow and gradual.  It is critical that IT maintain a constant watch even when performance doesn't seem to be an issue.

 

Longitude Report Summarizing Cloud Based Server Utilization
Longitude Summary Report  showing server health 


Conclusion

Ultimately it is about ensuring consistent and cost-efficient service levels across multiple environments. This can only be accomplished by aggregating and evaluating public, private, and hybrid resources.

Although the deployment of cloud resources can be quick and easy there can be a significant trade off between expediency and cost.  Proper cloud server monitoring needs to deliver to IT  the decisive information they need to accurately size and optimize their deployments.

Implemented properly, server monitoring is not heavy in terms of system overhead or arduous in terms of the effort required of IT to deploy, manage, and interpret.

The challenge is optimizing the time required of IT to manage and monitor cloud resources against the actual costs of the cloud infrastructure itself and maximizing IT's return-on-effort. This can only be accomplished with automated monitoring, reporting, and capacity planning.

Start Your Free 14 Day Trial of  Longitude Today! 

 

Sign Up for the Blog

Heroix will never sell or redistribute your email address.