Is Public Cloud or On-Premises Right for Your IT Infrastructure?

Do you feel like you are in a constant tug-of-war between vendors trying to convince you that their public cloud is the best solution or vice versa, that on-premises is the way to go? Truth be told, there is no one right answer, the better platform is dependent on a few factors that can have significant impact on your bottom line business success. Given that our Matrix software runs equally well in the cloud and on-premises we get to help customers maneuver this very murky decision-making process. Here are some key criteria we have seen that influence the decision.

• The predictability of the workloads
• The amount of data under management
• The level of application performance required
• The amount and type of compute resources needed

In this blog series I will share real-life customer stories to help highlight the benefits of both platforms and help illuminate the decision process to go in-house or to the cloud.

Why the Cloud is Great for Project-Based Workloads

TRE ALTAMIRA, headquartered in Milan Italy, is a leader in satellite radar-based ground measurement and structural movement for oil and gas, mining, civil engineering and geo-hazard industries. The company uses satellite imaging to analyze earth surface movements on areas that interest their customers. It provides value added services to many companies in the mineral exploration industry as well as many space exploration agencies, such as NASA and the European Space Agency. Its business is project driven and jobs can range from small imaging and analytics workloads to whole-country geo-spatial analytics.

The amount of available computer processing and storage performance have a strong impact on TRE ALTAMIRA’s business since they determine the company’s production capacity and time-to-market. Up until 2017, TRE ALTAMIRA managed all its projects in an on-premises datacenter so production capacity, time to market and innovation capabilities were limited by the number and the power of its compute resources and by the performance and the capacity of the underlying storage. Production capacity was limited to the available on-premises compute cluster, which in turn limited their ability to win new orders as delivery time is a critical factor. These limitations stifled innovation, and even led to engineers to discard new promising algorithms due to the lack of high-performance computing resources.

TRE ALTAMIRA pivoted to the AWS public cloud to run its complex workflows. After experimenting with several solutions including Lustre and NFS based file sharing, they opted for WekaIO Matrix that offered them the best performance and stability. Data is ingested from S3 to a high performance compute cluster, sized and optimized for the workload and deadline. Each project has a dedicated HPC cluster and compute resources can be dynamically scaled up or down depending on the project needs.
The benefits from a cloud investment include

• It removed any production capacity limits. Previous on-premises analyses would have exceeded two weeks and consumed all available resources. By running WekaIO in AWS, analyses times were reduced from two weeks to less than two days, a 7x reduction, and only required the precise amount of resources to complete each analysis.
• Costs scale linearly with the workload and can be directly applied to the project. If there are no running analyses, there is no cost. With WekaIO’s Matrix costs were reduced by 3x compared to their previous solution, delivering significant bottom line saving and enhanced profitability.
• There is no dependency between time-to-market and the number of concurrent analyses. Each run is independent making the sales process faster, simplifying the product process and ensuring that they can execute on as much work as is available.
• It has encouraged innovation. Products and services that had been discarded in the past due to lack of resources are now feasible with AWS and WekaIO Matrix.

The journey to the public cloud transformed TRE ALTAMIRA’s business by delivering an infinitely scalable HPC environment for project-based workloads. It sped up time to results by a factor of 7 and it provided granular cost insights by project. Most importantly it mapped directly to their business model of project based demand with high variability in project size, making it easy to directly attribute cost to the project.

Stay tuned for my next blog which will share a company decision to stay on-premises.