DeepCloud AI is proud to be endorsed by one of the leading experts in the industry

DeepCloud AI
Aug 7, 2018 · 5 min read

Auto-tuning Workloads with AI enabled environment matching algorithm — A Case Study of DeepCloud AI

I always take pride in reflecting on my journey as a software engineer, and even more on my work on hypervisor architectures in early 2005, followed by my first-hand experimenting with Private and Public Clouds in 2009. The journey has since then been a long and cherished one while having some great privilege to work with leading cloud vendors, to build competitive infrastructure and platform solutions in the Cloud.

But one thing that has constantly intrigued me till-date is our ability to truly leverage the value of the Cloud. Instance right sizing and capacity planning has been a challenging task for any Cloud provider and it has almost seemed impossible to get it right, due to un-predictable workloads and abstract measurement strategies. Instance right sizing is not just about matching instance type and size with workloads for first-time provisioning, but also looking at deployed instances to identify opportunities for downsizing and optimization for reduced cost. Organizations typically tend to ignore this over other priority concerns and with a mind-set to correct later, after instances have been provisioned and deployed in live customer environments. This gives rise to largely over provisioned instances and leaking costs due to under-utilized infrastructure resources.

Traditional Right-Sizing and Capacity Planning Approaches

Then came in vendors like AWS who offered a wide selection of instances which were compute, memory and storage optimized, so users could pick and configure the best fit for their workloads before provisioning. AWS also provided tools such as Amazon CloudWatch, AWS Cost Explorer, EC2 Right Sizing and AWS Trusted Advisor to monitor and evaluate costs and utilization to identify opportunities for optimization.

All of these provided excellent ways to monitor and report utilization. They also provided recommendations for right-sizing but were never deterministic and auto-adjusting. AWS auto-scaling feature works well for workloads that originated in the Cloud. But for those applications that were conceptualized elsewhere and were largely brown-field in solutions that require new technology to be implemented alongside older tech, leveraging the auto-scaling capabilities was never a successful task, due to legacy architectures and sticky session management constraints.

Once applications are deployed in the Cloud production environment, teams invest considerable time in tuning workloads for proper performance.

What does it take to right size instances and plan capacity accurately?

  • Monitoring of compute resources to understand CPU Utilization, DiskI/O and Memoryutilization.
  • Monitoring network bandwidth utilization, throughput and response time.
  • Periodic sampling of data and analysis of workloads.
  • Identifying trends and pinpoint cost drivers.
  • Detect faulty systems and anomalies.
  • Monitoring system idle time and under utilization.
  • Monitoring overall system reliability and security.
Above is a sample CloudWatch dashboard. What we need is a more intelligent self-reclaiming approach to right sizing and capacity management.

The Case Study: DeepCloud AI Platform

DeepCloud AI is a promising Cloud-AI Platform that is set to revolutionize the decentralized cloud ecosystem. The goal of DeepCloud AI is to enable a light-weight Cloud-AI based development platform & marketplace that both resource providers and consumers can use, to quickly come together and develop, test & deploy distributed applications (dApps). DeepCloud AI leverages a trust based Blockchain network in the backend for preparing data for use in machine learning algorithms and managing service orchestration. Large amounts of data can be moved into Cloud Storage and Smart Contracts to be used for complex rule-based processing.

DeepCloud AI is:

  • Unlike other providers, a lightweight easy to use Cloud-AI platform and marketplace with trusted backbone based on Blockchain technology.
  • Unlike other providers, DeepCloud AI is based on micro-services architecture that makes it easy to deploy, manage and optimize.
  • Customers can use the platform to deploy & manage their own applications while leveraging the plethora of AI services hosted here.

When resource providers lend their compute power to the DeepCloud AI platform and marketplace, they will be provided the ability to define configurations for their environments. The DeepCloud AI machine learning capability allows matching these configurations with the resource requestor to provide the best fit solution to host given distributed applications and services. The DeepCloud AI platform also includes capabilities to understand events that lead to performance bottlenecks and provide suggestions to remediate.

It is increasingly difficult to understand and define hardware and software specifications needed to provision complex, multi-tiered distributed applications. Further complexities exist when matching data and I/O requirements directly with the physical layer such as the number of CPU cores, RAM, storage and the network capacity one may need. The DeepCloud AI predictive analytics and AI Algorithms are tuned to study historical data and service usage patterns to closely predict capacity requirements. Users may use these recommendations to set up their environments by default or adjust it at real-time post deployment.

The DeepCloud AI right-sizing and capacity planning features can be used to determine compute, network and storage requirements to intelligently auto-configure workloads to according the resources. Through intelligent VM recommendation, the AI algorithms will reduce the burden on IT professionals and help them identify the right configurations and specifications of their physical and virtual infrastructure with optimized configurations.


As we have seen DeepCloud’s AI offering promises to provide a solution to automatically solve one of the most significant problems faced by Cloud service providers and consumers. Having the ability to automatically right size and plan capacity will remove the responsibility from service providers and consumers, to monitor systems and make informed judgements on utilization. This will also significantly reduce infrastructure costs and allow users to fully leverage the advantages of the Cloud ecosystem.

Join DeepCloud AI’s growing Community:

Official Website:
Telegram ANN:

DeepCloud AI: Decentralized Cloud and Edge Computing

When we say “Next Generation Cloud Computing” we do not take that claim lightly. Learn more about our platform and see how you can join in on the revolution!

DeepCloud AI

Written by

DeepCloud AI is an AI-driven Cloud Computing project built on blockchain.

DeepCloud AI: Decentralized Cloud and Edge Computing

When we say “Next Generation Cloud Computing” we do not take that claim lightly. Learn more about our platform and see how you can join in on the revolution!