As companies are processing more data than ever before to enable growth and technology advancements, finding cost-effective methods for storing data can be challenging.
Fortunately, there is an emergence of ML-based tools to ensure the ongoing stability, performance, and cost optimization of these large data processing workloads.
To understand why ML is so useful in supporting data storage, we spoke with Maxim Melamedov, CEO of Zesty – a company offering machine learning-based solutions that optimize the allocation of computing and storage resources.
How did Zesty come to be? What has your journey been like?
Zesty was founded in 2019 by Maxim Melamedov and Alexey Baikov. They were addressing the challenges and limitations found in today’s FinOps toolset. Companies are struggling with skyrocketing cloud costs and never-ending pressure to dramatically reduce those costs fast. Yet no matter what they do, the majority of businesses eventually reach a cloud savings glass ceiling. They're unable to get beyond a certain savings threshold.
Recognizing the need for a better solution, Maxim and Alexey created Zesty. Zesty uncovers hidden cloud savings with the power of ML that were once unreachable. Zesty’s Machine Learning-based solution automatically optimizes the allocation of compute and storage resources in real time to dramatically reduce cloud costs without any manual work or risk to application performance.
Can you tell us what you do? What are the main challenges you help navigate?
FinOps professionals are often challenged with the task of reducing cloud costs. Especially during difficult financial periods when there is a lot of pressure to keep a healthy gap between cost margins.
This is far from simple. While there are many cost analysis tools providing insight and recommendations into how to optimize cloud infrastructure, the time taken to crawl through this data and make the necessary optimizations is time-consuming. It takes engineers away from key development tasks.
Furthermore, the situation is never static. A resource that is under-utilized today may be over-utilized tomorrow. This happens as new features get added, user behavior changes, migrations occur, and customers come and go. All these shifts make anticipating future resource consumption extremely difficult.
FinOps professionals invest hours calculating future usage, only to encounter a misallocation of resources. For instance, some workloads are under-resourced, risking downtime and data loss. While others are over-resourced, wasting precious budget allocations.
Zesty changes this paradigm by using ML to adjust cloud resources to real-time demand. This way, users can provision resources without spending hours trying to determine predictions of future usage. The advanced ML model analyzes workload usage patterns and adjusts cloud resources fast and flawlessly. Further advancing the capabilities of human effort, allowing FinOps and platform engineers to achieve cloud savings previously considered impossible to reach.
Users can optimize cloud resources in real-time, achieving the best possible cost margins without any human effort. Zesty is at work 24/7 to ensure cloud resources are always used as efficiently as possible. This ensures application performance while dramatically reducing waste.
How did the recent global events affect your field of work?
These tough economic times have proven the value and need for Zesty’s products. Our solutions enable customers to save up to 50% off EC2 and up to 90% off block storage. Some of our customers have saved millions of dollars.
It's enabled them to not only get through recession periods but to also expand their cloud footprint to further speed up innovation. As a result of these benefits, our customer base has grown by 20% despite the economic downturn.
How does your platform address the needs of enterprise-level clients compared to small businesses or individuals?
Both small businesses and enterprises can benefit from Zesty’s products. Organizations of all sizes generally get significant value from Commitment Manager, which helps companies automate the management of their public cloud-provided discount commitments. An intelligent ML algorithm gets used to adapt AWS Reserved Instances according to changing workload demand. This enables organizations to maximize their discount coverage and ensure they’re paying a better rate for more of their workloads.
For many organizations, there is fear of vendor lock-in due to the financial risk of being unable to adjust cloud infrastructure for one or three years. This commitment causes organizations to forgo cost savings available through AWS discount commitments. Commitment Manager eliminates these risks by minimizing the long-term commitment period and enabling the flexible allocation of RIs across workloads.
For small organizations, Commitment Manager allows them to start leveraging these discount programs. Large enterprises tend to use Commitment Manager in combination with their self-managed commitments, helping them to cover the 10 or 20% of instances that they struggle to cover on their own.
Zesty Disk, our solution for block storage, is preferred by large enterprises with mature FinOps looking for new frontiers to further reduce their cloud spend. The solution automatically shrinks and expands block storage file systems according to real-time application needs.
Large businesses tend to produce copious volumes of data, which is subject to great fluctuations due to sales seasons and business events. It’s easy to get stuck paying for TB of storage capacity across thousands of workloads, that's no longer needed and wastes millions of dollars as a result.
By using Zesty Disk to shrink and expand storage file systems according to real-time needs, businesses always have the right amount of storage capacity for their workloads. It's eliminating waste and provides a significant improvement to their cloud’s ROI. With savings compounded for every megabyte under management, we’ve noticed the more data a company stores, the greater its total EBS savings.
Is your data protected against cyberattacks, and how do you ensure that it's secure?
Zesty does not collect PII or any data at all on the disk for that matter. To train the ML algorithm, only metadata and usage metrics are collected. I.e. the I/O count, throughput, previous usage levels, patterns, etc.
Our zero level of exposure to any sensitive consumer data makes Zesty a relatively risk-free solution with minimal security implications. When optimization actions should be taken, they're executed through highly limited IAM role permissions. Zesty is SOC 2, and GDPR compliant.
And finally, what’s next for Zesty?
Zesty’s mission is to help more and more businesses leverage ML to capture hidden cloud savings. To achieve this goal, we will expand our product line to different cloud providers and support a wider range of OS.
Zesty looks forward to expanding on its vision of helping cloud Ops teams maximize efficiency and reduce wasteful practices in their technology stack.