What challenge with cloud computing is FittedCloud trying to solve?
PM: A significant challenge public cloud customers face is wasted spending on cloud resources. An estimated 30 to 45 percent of the money IT organizations spend on cloud infrastructure is wasted. As noted by the analyst firm, Storage Switzerland, this happens because public customers buy more cloud services than they need to support peak workloads. Public cloud customers are typically charged based on the amount of infrastructure provisioned, not what is actually used.
How does FittedCloud reduce wasted spending?
PM: Many cloud management companies are attempting to solve this challenge with cost management solutions. FittedCloud emphasizes automation and applies machine learning to its solutions. A few years ago, we recognized the impact machine learning is having on data security, financial trading, healthcare, marketing personalization, and smart cars, among other industries. We are using machine learning and predictive analytics to help Amazon Web Services (AWS) customers “fit the cloud” to their needs, matching resource provisioning to application utilization – automatically and transparently.
How do you “fit the cloud” to a customer’s needs?
PM: Very few things, if any, in the world are truly elastic in nature – meaning that they grow or expand when you want more of them and contract or reduce when you want less of them – thereby making sure that you as the owner of the resource get maximum utilization of the investment.
An easy to understand example is your clothing. Wouldn’t it be wonderful if all your clothes were truly elastic? Imagine if your pants knew it’s Thanksgiving and could expand an inch or so to make sure you are comfortable after the delicious feast. What if your dresses could know that you like to exercise a lot in the summer and thereby would shrink by a size or two and then revert back after Labor Day? You would rarely need to buy new clothes – they would automatically adjust according to your needs. In fact, they would learn your patterns and habits over time and adjust without your asking.
That is the concept FittedCloud is applying to cloud computing. By using patented, sophisticated machine learning algorithms to learn a customer’s historical use of cloud services and match resource provisioning to application utilization, there is no wasted spending. Our cost management solutions not only identify opportunities for AWS customers to save money, their Actionable Advisories also provide one-click, semi-automated remediation. Customers love how fast and easy it is to identify, review and execute the solutions’ recommended actions for cost savings. We also offer full automation for customers who prefer a completely lights out solution.
How does machine learning help manage cloud computing costs?
PM: The key idea underlying machine learning-driven methods is that we are able to learn patterns of customers’ resource usage from historical data records. With these patterns, we are able to predict future needs.
We usually have two stages while using machine learning for prediction: a training stage and a prediction stage. At the training stage, given a set of collected data records, we train a mathematical model which is usually a mapping function from the data to the outcome. At the prediction stage, we apply this model to predict the outcome for a given new input data.
What cost management solutions do you offer and how do they work?
PM: We offer solutions that help AWS customers find and fix wasted spending, right-size their infrastructure and detect anomalies with resource provisioning. Let’s look at Amazon EBS to start. The typical Amazon EBS customer is charged for the resources provisioned and not for what is used. For example, if a customer provisions a 1TB EBS volume and uses only 100GB, they are still charged for 1TB capacity. If a customer configures a provisioned IOPS (IO1) volume capable of 5,000 IOPS and only uses 1,000 IOPS, the customer is still charged for the full 5,000 IOPS. FittedCloud EBS Optimizer automatically and transparently manages EBS storage so that you only pay for the storage capacity or performance used by applications.
Next, let’s examine Amazon EC2. Often AWS customers over-provision cloud compute resources by a factor of 2 or 3. FittedCloud EC2 Cost Optimization assists in identifying Amazon EC2 compute resources that are underutilized, ensuring that customers only provision and pay for used compute resources. The solution provides customers with full control over the extent and timing of the optimization of their resources. The solution also assists customers in scaling up their resources when applications need additional compute power or memory. All of the actions are performed seamlessly and in a secure way without any performance degradation or application disruption.
For example, you may have provisioned an ‘m4.4xlarge’ instance (16 cores, 64GB of RAM) and your application may only use part of the CPU or memory provisioned. Our software will monitor your application usage and depending on the utilization will automatically switch it to an m4.2xlarge at a cost savings of 50% per instance or even to an m4.xlarge at a cost savings of 75%.
FittedCloud recently added anomaly detection to its AWS cost management solutions. What does the new solution do?
PM: In today’s connected world, chances of intrusion and malicious attacks are constant. Public cloud environments, such as AWS, are not immune. For example, if a company’s identity and access management (IAM) keys get exposed or stolen they can be used by an attacker to create a large number of public cloud compute and storage resources that can cause financial harm. Unfortunately, there are no mechanisms that exist in public cloud infrastructures to prevent such a malicious act.
Human or programming errors in automation scripts can also cause abnormal changes to resource provisioning. These types of errors can create a significant number of resources not only causing direct financial harm but also restricting resources needed for mission critical applications.
We enhanced our predictive analytics capabilities to now find anomalous data patterns and inform AWS customers about changes to their infrastructure resources that are out of the norm. For example, our cost management solutions will identify when a higher than the usual number of EC2 resources are created, when EBS capacity is greater than normal, or when a greater than the standard amount of data is uploaded to S3 buckets. Our Anomaly Detection alerts provide the location of the attack and credentials used for attack, among other pertinent details.
You currently focus on AWS. Google and Microsoft are also strong players on the public cloud IaaS market. Do you also intend to provide similar cost management solutions for Google Cloud Platform and Microsoft Azure?
PM: We started with AWS because they are the market leader according to Gartner and other industry analysts. As our solutions mature we will explore opportunities to create versions that will help customers of Google Cloud Platform and Microsoft Azure realize the same level of cost savings.