Moving infrastructure and apps to the cloud promises agility that better aligns resources to demand. There’s also a promise for cost savings depending on what you’re planning to do.

However, if you do not manage the consumption of resources—especially in public clouds like Amazon Web Services (AWS) and Microsoft Azure—you could end up paying a lot more than expected. Being able to monitor and plan for the cost of running infrastructure in the cloud vs on-premises helps you to make good business decisions.

The return on investment (ROI) for moving to the cloud vs on-premises is also tied to outcomes. If you move applications to the cloud and end up losing visibility and control, an outage could be very costly.

Two of the most important questions for IT operations to consider for monitoring include:

• What can and cannot be monitored in the cloud versus on-premises?
• Will you lose any visibility or control if you move your app to the cloud?

Here are several factors to consider when moving apps and infrastructure to the cloud.

Monitoring cloud-hosted apps and infrastructure

When you consider monitoring hosted apps and infrastructure in the cloud, agents or probes can be installed in the same cloud environment. The agents or probes provide key performance metrics to the monitoring software. You’ll get the typical infrastructure metrics for performance and availability including CPU, memory, and processor utilization as well as application performance metrics for transactions. You’re instrumenting the same infrastructure as you would on-premises, just in a different environment.

Moving to the cloud should not be a problem in this case.

You should consider whether to run your monitoring software that takes in the agent or probed data hosted on-premises, or alongside the infrastructure and apps in the cloud. In some cases, you might want to consider a hybrid approach where you could have monitoring instances running in the cloud environment collecting and doing analytics and then storing the data back at your private data center. Organizations who do this prefer not to store data in public cloud environments for security reasons.

You’ll want to make sure that the monitoring solution you choose can be architected with components that could span from your data center to private and public clouds. We have a large retail customer who has done exactly that. They were able to architect a solution that took advantage of scaling their infrastructure and monitoring in the public cloud while maintaining the security of their sensitive data on-premises.

Legacy monitoring tools

If you have legacy monitoring tools on hardware appliances installed in your data center, then moving to the cloud is not going to work out so well for monitoring. If there’s a software version of the appliance that you can install on a virtual machine (VM) in the cloud, then you can get the same or comparable monitoring coverage as you had on-premises.

Newer architectures

It gets more complicated when the technology used in the cloud runs the apps differently than what happened on-premises. For example, running apps in Docker containers requires a different approach to instrumentation. You need to monitor the hosting environment of the Docker containers as well as what’s going on inside the containers themselves

Here’s a great example from one of our TrueSight Solution Engineers who shows us how to achieve the instrumentation for Java apps on Docker:

TrueSight App Visibility in Docker Containers

In a Docker environment, the ability to scale and have a dynamic monitoring solution that can keep up with the changing conditions is crucial.

To maintain the same visibility and control, you will want to make sure that you can monitor the newer technologies as well as instrument the app environment.

Monitoring end-user experience

Getting real-time, end-user experience metrics from apps that run in the cloud is a relatively standard activity when you have control of the web servers that are serving the apps. You just need to add a snippet of code to the pages being served by the web server and then end-user experience monitoring is enabled. For example, a JavaScript snippet can be automatically injected to the headers of web pages at run time, or you can manually copy them to the pages. You can then monitor the real user experience of the web pages as well as the Ajax calls.

Most monitoring tools include this type of monitoring so that your cloud monitoring decision will not be impeded by this concern.

What differentiates beyond front-end, end-user experience monitoring?

Going beyond the basics of end-user monitoring requires tying end-user experience monitoring to infrastructure monitoring and downstream analytics. This helps to troubleshoot an issue much faster than going to one console for user monitoring and another one for the infrastructure.

Cloud providers typically provide a level of monitoring natively. For example, for AWS, there are CloudWatch metrics available for the running infrastructure. These can also be integrated into your environment so that you have the infrastructure health from the cloud provider as well as deeper server diagnostics as necessary that you can instrument with agents.

Having a single console that shows this data in an application performance context will help reduce mean time to repair (MTTR) instead of having to flip through multiple sources of monitoring data. End-user experience monitoring for SaaS apps cannot be done as easily if you don’t have the ability to instrument the web pages of the service provider of the app.

Downstream analytics using AIOps approaches for your complete IT environment can also have a huge impact on the value of your end-user monitoring in the context of a larger strategy of IT operations.

Monitoring SaaS apps

You cannot monitor SaaS apps that you subscribe to very easily unless the provider gives you access to their environment.

You would be tied to whatever the SaaS provider offers you in terms of service levels. If you have degraded performance, you’d typically enter a service request to the provider. There might also be an SLA in place to mitigate any dissatisfaction with performance levels.

However, this does not stop you from doing some monitoring. You could do synthetic monitoring of the web pages that you’re accessing, particularly if there are disputes between your service provider and your end-user experience. They might be using a less granular approach to monitoring. If they only report on hourly averages, the fact that several users have slow response times might not be captured if their average response time did not reach a critical threshold. You could pinpoint what times and pages were slow based on your synthetic monitoring.

Note—if you’re hosting the SaaS apps as a provider, you’d consider the same things you would consider in the previously discussed examples for monitoring cloud-hosted apps and infrastructure. You’d have access to instrument all or parts of the environment yourself and provide the SaaS offering to your customers.

TrueSight for cloud-hosted apps and infrastructure

Whether you’re monitoring from the back-end legacy components in your data center environments to the nimble Docker-hosted apps in the cloud, TrueSight can help with:

• End-user experience monitoring on-premise or in the cloud
• Software-based packet capture that can be deployed in the cloud
• Integrated metrics from AWS, Azure, or OpenStack cloud providers
• Synthetic monitoring for simulated end-user experience for SaaS apps

For more information, see
http://www.bmc.com/it-solutions/truesight.html

Digital transformation has changed how we do almost everything. Our consumer expectations have evolved so that established services like banking now have mobile apps that work in the palm of our hands to win or keep our business. Hailing a ride or purchasing just about anything happens in seconds—the apps must run without a hitch.

Optimizing for infrastructure capacity in this fast-paced digital environment offers a competitive advantage for any business. Whether it’s financial, online commerce, ticketing or anything else IT intensive, go-to-market initiatives must be fast and powered by less resources.

ROI is key

Optimizing for infrastructure capacity means having “just enough” resources required to run applications and services without interruptions in desired performance.

No organization wants to purchase resources that aren’t necessary. And with on-demand resources available from numerous cloud providers, the costs can be much greater than expected without doing some up-front planning.

Enterprise IT has two options:

Option 1—Invest in identifying the IT resources they need to align with demand

Option 2—Risk degraded performance and costly expenditures for unnecessary IT resources

The return on investment (ROI) for capacity planning and optimization can significantly advance the business if done with the right approach.

I must admit I’m biased toward Option 1 because I’ve seen it in action.

Consider a typical scenario of “VM sprawl”

In one of my first technical marketing assignments, I needed to showcase monitoring SharePoint on Windows for an upcoming conference in Las Vegas.

A lot of enterprises use VMware’s vSphere for provisioning and maintaining virtual machines (VMs), or something similar – like Microsoft’s Hyper-V. Hosts run on desired operating systems with specific applications and services for development and production activities.

I installed SharePoint servers in a VM environment along with our monitoring software on separate VMs to run the collection and analytics, and another VM for the web server that provided dashboards for users. vSphere was provisioned to me by a senior person on my team. Occasionally, things stopped working and he provided more resources with his superpowers.

Both the SharePoint environment and the monitoring software were new releases so there were lots of bugs. I was new to tech marketing and wanted to impress my team with “I got this.”

I created VM “snapshots” of incremental versions of stability as backups so that I could revert to a prior state at any time without having to start over. If I introduced some buggy software or demo scenario that hosed up the system, I could revert to a dated snapshot.

It was a great strategy in theory but in practice there were many times when I had to start all over—uninstalling and reinstalling both the SharePoint software and our monitoring software. The result was a lot of VMs in my allocated vSphere space in various states as snapshots that were irrelevant in the end.

Fortunately, everything went well for the demo with the help of lots of back and forth with R&D as well as a lot of Knowledge Base articles from Microsoft professionals. I ended up leaving many of the snapshots on disc as I went on to my next assignment.

Let me recap. Yes, I left all those snapshots on our vSphere environment at full tilt—taking up disc space with a lot of memory, CPU, and dual processor capability.

What if the scale of what I was doing was much larger with hundreds of developers or other IT professionals doing the same thing? This could be significant depending on the environment and who’s controlling the allocation of resources. My guy with the superpowers for vSphere would not be enough—having effective strategies in place could make all the difference.

If these VMs were deployed in a public cloud, imagine what the bill would be if I left them idle!

Consider using strategies for Capacity Optimization as digital maturity for your enterprise.

Strategies for capacity optimization

Visualization—Having a complete view of IT resources in multi-cloud environments that span on-prem data centers, data centers in the cloud, storage from many different vendors, along with a view of users desiring consumer-grade performance 24×7 helps IT make key decisions about who’s using capacity, how much, and where resources can be pulled back or re-allocated.

Forecasting—Being able to plan for capacity can make all the difference. Outages can be prevented and decisions can be made to accommodate growth or a reduction in IT resources. This requires predictive analytics based on several interdependent resources that include the network, the number of users, and the computing environment including storage.

Cost control—Having the ability to shop around for on-demand resources at the right time to fit a need or to invest in on premises infrastructure helps to control cost. This control requires the combination of visualization and forecasting as well the analytics to calculate costs in many different virtualization, private and public cloud environments.

Proof of ROI

Now more than ever, the race for innovation and efficient use of IT resources separates winners from losers. Finding out what others are doing and what results they are getting can help.

Here are a couple ways to investigate further:

Capacity Optimization Reviews https://www.bmc.com/blogs/truesight-capacity-optimization-reviews-ratings

Customer Testimonial https://www.bmc.com/customers/incontact.html

Clouds—A playing field for innovation

For years, we’ve been innovating with cloud technologies. Within minutes we spin up infrastructure that is publicly accessible to users across the globe for just about any web-based application or service.

The promise of agility and lower cost drives many innovators to self-provision cloud resources instead of going through the dedicated IT operations processes in place. Upper management often drives innovation and rewards this behavior when it goes well. When IT operations processes are left by the wayside though, what’s the risk?

My cloud story

For my role in technical marketing at BMC, I need to have environments to test and demo our solutions as they are being released more frequently—and in some cases—continuously. We have a dedicated team that can spin up infrastructure to support this. However, for one of my smaller projects, having to go through that process did not seem as efficient as using Amazon Web Services (AWS) or Microsoft Azure. I was able to start my project right away at no initial cost. I also had a hunch that as it got underway and incurred expenses, it would be at a lower cost overall for the company. We ultimately used my team member’s corporate credit card to submit for expenses.

What’s to lose?

At any large company, a lot of other folks are probably doing the same thing and going directly to a public cloud for a relatively small amount of resources. This can generate large surprises with material consequences. I must admit that in my case, there were times that I did not check on the state of my running instances. I had stuff consuming data and resources in the cloud that was not necessary, especially after trade shows were over. Stay tuned for a future post on how we are managing this companywide to enable fast access to resources while controlling costs.

For today, lets consider how IT Operations can champion innovation like this on the edges within the management and oversight of infrastructure and operations (I/O).

IT Operations as multi-cloud champion

Some innovators might argue that IT Operations might not be best suited to champion multi-cloud initiatives inspired by innovation. Before weighing in on the discussion, consider what championing would mean in a multi-cloud playing field and then be the judge. Here are just a few ways that IT Operations can help:

• Planning—IT Operations can have a companywide view of all the cloud resources being utilized, or planned for utilization, along with their associated costs. In a best case, they would be able to simulate usage from various deployment options. The options could include virtualization on premises, or private cloud, or public cloud. This data would then inform decisions and align resources with business goals instead of just following hunches. This information might help prevent folks like me from going out on our own. Partnering with IT Operations and getting these more responsive insights helps us make better decisions.
• Optimization—With central IT Operations as the referee or overseer of cloud resources, infrastructure can be optimized to fit the workload. Without jeopardizing performance, idle and underutilized infrastructure can be discovered by IT Operations oversight then redirected towards other projects that need resources or have a higher priority. With the ability to programmatically control virtualized infrastructure hosted in the cloud—again from a central location—workloads can be dynamically shifted to available infrastructure in real time.
• Monitoring—This gets complicated if you have an assortment of monitoring tools all over the place and then you add multi-cloud sources. Services and apps running in the cloud often include some form of native monitoring. Normalizing this monitoring across a consistent platform has great benefits. The data that is used and shared across teams becomes a trusted standard. With everyone using the same platform of data, there’s a reduction in complexity that makes it easier to prioritize and address issues faster.

When IT Operations has a complete view of cloud adopted resources across the organization, finding that sweet spot of just enough infrastructure to run workloads while not compromising performance is much easier to achieve.

What’s next?

Consider how multi-cloud strategies have changed the way organizations are doing business compared to the past, even five years ago. Some of the innovation that was introduced then is now part of core IT. This balance of innovation on the edges with central IT Operations as the champion of these initiatives will make a huge difference, particularly with multi-cloud players competing for business in a fast-paced digital environment.

It’s not uncommon for performance monitoring strategies to leverage thresholds to alert operators when certain metrics for key performance indicators (KPIs) go outside what is generally considered acceptable for IT apps and services to run smoothly.

To better understand this, consider when an app on your computer starts consuming a lot of memory and everything else seems to slow down.

While most of us don’t have thresholds and alerts configured to mitigate our slow computer issues, enterprise-level computing requires this exact kind of notification and much more to keep IT operations at optimal performance. Think of it this way, instead of just you having a bad experience while an app is using up memory from your computer, imagine thousands of users accessing an app that is consuming memory from a bunch of servers. When IT is alerted with an issue from a threshold violation, they want the fastest mean time to repair (MTTR).

Helpline scenario

We call our bank’s or healthcare provider’s helpline and the customer service professional says, “My computer is slow today.” He or she is accessing one or more of the mission-critical apps within their enterprise system. They are being affected one way or another by some element in their IT environment that causes your request for help to slow down. Slow service coupled by a bad experience might make you consider other options. What if hundreds or thousands of customers are having this same bad experience?

Complexity

What makes these issues particularly complex is that enterprise-level app performance requires tip-top performance across all tiers of web servers, app servers, and database servers in traditional architectures as well as in more modern, highly-distributed environments with microservices. For most enterprises, the complexity behind mission-critical apps span data centers, private and public clouds, and even brokered services from other providers. If the apps are not running smoothly, chances are the business is losing revenue, productivity, and even brand loyalty.

If IT operations staff get an alert every time a performance measurement reaches a threshold for even a brief period, managing the volume of these alerts can be a nightmare. The issue might not even have a significant or sustained impact on end-user experience. This situation can be very inefficient to address.

That’s why modern IT Ops solutions provide ways to go beyond basic monitoring to reduce the noise associated with event storms.

How machine learning and analytics can help

Instead of relying on static thresholds, here are several machine learning and analytics approaches to fix issues based on priority and impact so that IT operations can reduce MTTR:

Behavioral Learning—Use a combination of baseline data from machine learning and a threshold value to determine when an issue is more likely to be a problem. You only get notified when a measurement is outside of what is learned as normal for the time of the baseline data and is above the critical performance value of a threshold that you specify. A typical scenario for this would be during expected high demand usage at the beginning of the day or work week. CPU utilization for servers supporting apps could spike for an expected length of time every Monday morning between 9 AM and 10 AM.

For example, you have set an absolute threshold of 80 percent utilization for CPU—generally considered as a good value to start being concerned about performance of your servers. If utilization spikes to 90 percent against a learned baseline for this period between 85 – 95 percent, you would not get an alert. If, however, there’s another time during the week when usage is at 90 percent with the baseline between 65 – 75 percent, you would get an alert that you need to address. In the first case, the actual usage does not go above the threshold and outside the learned baseline. In the second case, the actual usage is above what is considered normal and is above the absolute threshold.

This helps you prioritize the most important issues first.

Predictive Events—Use a combination of hourly baseline data and a critical threshold value to receive an alert or event notification when a measurement is about to reach the critical threshold (typically about three hours). Machine learning provides the hourly baseline data and you set the critical threshold. When a measurement starts to go outside the baseline value but before it reaches the critical threshold, you essentially get a warning. If you don’t do something in time, performance can be significantly impacted.

Someone on your IT team might change a configuration that could unexpectedly flood a certain resource during peak times. A predictive event would give you plenty of time to address the issue before users even notice.

Probable Cause Analysis—Use a process that analyzes data and displays all relevant events and anomalies when troubleshooting an issue. In complex IT environments, there are typically many factors that can impact the performance of an application or service. By having the system narrow down what is the likely cause of an issue, IT operations can pinpoint the root cause of the problem faster. Relevant events to consider are ranked and displayed based on their relationship to the initial event, the timeframe of consideration, and abnormalities captured by behavioral learning.

You could be alerted that a server in your environment has processor time that has spiked above what is considered normal and is related to a memory issue. Because of the spike in memory, the server does not respond to requests for data as quickly. By addressing the memory issue, the processor time goes back down to normal.

Pattern Matching—With a little upfront planning, you can identify probable cause scenarios that typically happen in your environment as knowledge patterns so that troubleshooting issues can go even faster over time. In the case of processor time and memory in the previous example, you could create a knowledge pattern that identifies critical sets of infrastructure that you need to address when both processor time and memory are an issue. You would get notified of the probable cause event rather than having to run a new probable cause analysis.

If the processor time is unusually high but there’s no memory issue, then you could do further investigation for other issues.

Log Analytics—With machine learning applied to volumes of log files, the system can baseline what is considered normal for like log entries. That means what is considered “normal” is learned by the system. At any time when the number of log entries that match the pattern change by an unusual amount, you can then quickly identify an issue in your IT environment.

This is what is considered as an “out-of-bound” anomaly. The issue is occurring far more or far less frequently than what is normal. This would be particularly useful if you suddenly have a lot of users with access to a machine who did not before or in the opposite case, a lot of users don’t have access when they did before. There were configuration changes related to access lists (ACLs) that you need to address.

A long history at BMC and more to come

These machine learning and analytics techniques have been a part of the TrueSight portfolio for years and it’s just getting better. Innovating with cross-platform and 3rd party data support, TrueSight now offers service ticket analytics and correlation with time-series data from almost anywhere. You can quickly ingest volumes of real-time and historical data in minutes to uncover patterns and find root causes for what used to take many days and hours for IT operations staff to address.

Going beyond basic monitoring with TrueSight to reduce the mean time to repair—respond, remediate, rescue (MTTR) any IT operations issue.

IT services now sprawl from a wider range of resources to a wider range of audiences in most digital business environments—thanks to the cloud.  A lot of IT services are now offered with just three or four clicks on the Internet and a credit card to make it easier than ever.

For example, let’s take a group of developers who have been tasked to make one of their mission critical apps run faster, with more functionality, and be delivered quickly. Why would they want to take the time to provision their infrastructure with traditional IT approaches when they can get started right now—turning to AWS, Azure, OpenStack and others—to quickly have an environment in place right away?

The win here is that this team is up and running quickly and producing results, typically in a development environment. But what happens when it’s time to scale in production? They have not been too concerned with monitoring in their original environment because they just wanted to get the functionality in place to deliver the faster app with enhancements to functionality.

Once they go to production, everything gets messy. Their line of business owner, although pleased with the results, starts clamoring about budgets. Which cloud provider is going to be the best solution for production? Are their costs associated with run-time environments that could sky rocket without the proper visibility? What happens when the infrastructure is maxed out or part of the application delivery fails?

After some analysis and partnering with the IT team, they decide to host parts of the application in a public cloud and the storage on premise in a more traditional database server farm. Some organizations like this group might also have some backend processing that is still running on mainframe technology.

In one scenario, they could rely on the native monitoring that might be available on some of these technologies or they could have some “point” solutions to gather logs or monitor application performance. Some of the cloud environment might not have the same visibility they are used to and they are left with the fear that their production environment could fail. They will not be able to address the issues quickly with a patchwork of legacy tools and environments.

In a better scenario—as one of our BMC customers at Ensono, a leading IT infrastructure provider, put it—“This patchwork of diverse tools worked well in the past. To support today’s complex IT needs, however, we need a sophisticated, holistic monitoring strategy that covers mainframe to cloud and everything in between.”

For IT initiatives like these, clouds just won’t go away in today’s digital business. Whether it’s SaaS, PaaS, IaaS, or you-name-it as a service, cloud provides us with nimble, scalable infrastructure so that anybody and any organization can quickly do just about any digital activity by giving up some control and paying as they go.

Check out the hottest new cloud initiatives for TrueSight Operations Management:

• Comprehensive monitoring for both AWS and Azure infrastructure environments with TrueSight Operations Management
• Real end-user monitoring with a software edition of TrueSight APM that can be installed directly in any hosted or private cloud environment
• Agile, second-by-second monitoring and analytics with TrueSight Pulse and TrueSight Intelligence

Get started today.

TrueSight Operations Management Free Trial – BMC

Where to next?

• TrueSight Operations Management – BMC
• IT Operations Analytics (ITOA) – BMC

You’ve probably heard the buzz phrase “single pane of glass” more times than you care to admit. It’s typically used to convey the benefits of an enterprise-level solution that unifies a bunch of data into a single dashboard interface.

Questions to Ask

When you are talking to someone extolling the virtues of their “single pane”, here are some questions to ask:

• Does “single pane” mean one dashboard where all data resides or where all workflows begin?
• If a right-click launches into another interface is it no longer a single pane?
• Is it one solution that meets the needs of IT with multiple interfaces?

We should consider banning the term “single pane of glass” and concentrate more on what it is we’re actually talking about: bringing disparate sources of data together to gain actionable insight. That’s the value, regardless of what we end up calling it.

Current State of IT

“Managing IT has gotten so complex that data streams of performance, availability, and business insight metrics now span an ever-increasing, widely-distributed architecture—through private and public clouds, data centers, remote sites, mobile users, and even ‘things.’  There’s no secret about how to address this data volume. The right data must be in the hands of those who make critical decisions about the data and the faster, the better.” -IT Pro, BMC Software

Most IT shops are simplifying their approach in order to avoid piling up monitoring tools and to reduce fire drills and war rooms. It’s not uncommon for different departments—for example, and operations—to get defensive, lean on their own tools to justify their performance, and blame issues on others. Promises that a “single pane of glass” can solve this problem are appealing. But are those promises realistic?

Operational Dashboards

IT staff design operational dashboards based on devices, events, infrastructure and/or applications. They then set static or dynamic alerts for their KPI’s.

If a business critical application is down even for just a few seconds, users click away and revenue drops. If the app is down longer brand loyalty is at risk.  The face of many companies now is their application presence and their user’s experience. Think of Airbnb, Uber, Turo, Amazon, and others. Having operational dashboards in place that can catch and mitigate outages is essential.

Business Value Challenge

Once you master “find-and-fix” workflows with operational dashboards, you create the space to innovate and undertake strategic projects. Innovating and driving strategic change require you to “up-level” your operational dashboards to a business value dashboard that enables you to look more broadly than operations to all the value drivers in your business.

Business value dashboards reflect metrics that drive the whole business, not just IT operations.  Business drivers are unique for each enterprise and IT staff need to look beyond out-of-the-box, boiler plate products. The ideal is a dynamic solution with a common framework that can bring data in from anywhere, layer it with analytics and intelligence and visualize it in an actionable, unified view.  BMC TrueSight is such a solution.

Real World Scenario

When Sarah Palin announced her endorsement of Donald Trump, she was wearing a striking sweater that got a lot of attention on social media. Within a few days that sweater was sold out from all suppliers.

Let’s say you had real-time access to the trending social media sentiment on Sarah Palin’s sweater and you tied it to your online ordering process. You could dynamically adjust pricing to capture more of the market value and preemptively order stock before your competitors. This is an example of how a business value dashboard empowers business owners to react to dynamic market data and increase their business bottom line.

Achieving Actionable, Business Value Dashboards

Using BMC TrueSight solutions you can configure the kind of business value dashboards that provide actionable insight from the data you collect from applications and infrastructure and combine it with data from other sources such as social media and end user behavior.

By moving beyond the idea of a “single pane of glass” to a business value dashboard powered by BMC TrueSight solutions, you can help business line owners to make more informed business decisions, reduce the amount of time you spend on day-to-day operations, and free up time for innovating.  Manage your datacenter and  industrial IT infrastructure while expanding into modern, digital IT service delivery.