Share your experience using Oracle Compute Cloud Service

We process data mostly by scraping from retail chains on behalf of our customers, and then we homologate and clean that data for our customers to use in analytics displays and decision-making tools. The original architecture was very basic as we used applications such as Pentaho for all ETL workloads running on just fixed VMs from EC2, and then we moved these workloads into AWS Fargate. We didn't use any sophisticated Kubernetes clustering integration, so it was very basic in that case.

Currently, we are deploying workloads because this company has been going through a lot of merging and acquisition strategies. The company I am speaking of was a Mexican company that had 100% of the infrastructure over AWS, but as we integrated to manage the other companies we bought, we had to manage a different set of AWS services, such as Redshift and so on.

In our case, the workloads were not oriented to micro-services with AWS Fargate, just servers. The ETL flows we moved into Fargate using Pentaho weren't designed for micro-service architecture. The front end was very monolithic rather than micro-services oriented, and we had some simple API services, but not a formal micro-service architecture, so we didn't use AWS Fargate for that.

One of the best features of AWS Fargate is that it was useful for us because we didn't require to run container workloads and we didn't need to deal with the management of a Kubernetes cluster directly, and the ability to run those workloads just in a scheduled manner is also a great feature. Additionally, we reduced costs because the distribution over time of our workloads was not the whole day, so we had fixed VMs and some capacity that we weren't using.

AWS Fargate has definitely impacted our organization positively because before that, we couldn't accommodate the workload requirements as flexibly as we can with Fargate. We were able to process all the workload in a shorter time as we could increase the number of containers processing data by schedule. We provided operations because, in some cases, they required special workloads processed outside the schedule, and we offered some automations for them to just go to an internal application and configure their schedule and the amount of horsepower they require to increase in AWS Fargate, and it worked very easily. I understand that this could be accomplished with the auto scaling capabilities of Fargate, but as mentioned, the DevOps were just learning to use AWS Fargate and chose a simpler approach they considered less risky.

Currently, I think that the program is great the way it is, and maybe we use less than 50% of the current features of the platform. For example, we have been evaluating Dask with Python to work with distributed processing, and I don't know if AWS Fargate could be used to build Dask clusters, which would be great because there are many people on our team working with Python workloads. Having something that can manage distributed processing using Dask in a containerized deployment could be valuable. For a company that does not require complexity or managing Kubernetes clusters, AWS Fargate is a great way to go with the use of containers in a simple way, providing the power of containers and scalability without the complexity of going into Kubernetes.

I started to use AWS Fargate about two to three years ago because we had some loads on fixed virtual machines, and we needed to scale based on a schedule.

I would rate AWS support more than perfect because when I started working with the Mexican company, which was 100% AWS, we weren't that big so we didn't qualify as high-value customers, but even so, the AWS team in Mexico gave us special attention that went beyond our service level agreement. Even though we didn't contract support, every two weeks I had a 30-minute meeting with a cloud architect from AWS to help our team use different products of AWS, especially with SageMaker for a forecasting algorithm we were developing.

Positive

The pay-as-you-go pricing model of AWS Fargate was one of the major drivers for us to move there because we reduced costs while increasing the quality of the processing services by about 30%.

My team hasn't used AWS Fargate's capability for automatic scaling in our applications, even when they knew about it; they were very new to using these features.

AWS Fargate's security features weren't the most important criteria or value we got from AWS Fargate.

I would recommend AWS Fargate to others; I would use it myself if I were in a startup or a medium-sized team because of the simplicity it offers without having to manage Kubernetes clusters. As we started using Kubernetes, I found it complex and only justified when you have a larger team that can manage certified cloud architects and cloud engineers. For a small team, I would go blindly using AWS Fargate if I want to take advantage of containers in a scalable way.

I rate AWS Fargate overall as 10 out of 10. I don't have any regrets using it. We have never had any problem using it, even when we used it with very basic knowledge of the cloud.

Currently, I am in operations. We have clients that we support with their infrastructure and multi-cloud operations. We use AWS, Azure, and some parts of Google.

Mainly we are using EC2, RDS, S3, VPC Network components.

That depends on our role as we are supporting infrastructure. When clients are using applications, we are not involved with their applications, but we support them with infrastructure.

We use it for network fundamentals.

For security and ACLs, Routing Tables, route tables, subnet, and subnetting, these are very useful functions.

For public and private subnetting, we need subnets.

The good part is what I mentioned regarding security reasons and subnetting. I cannot think of any bad aspects of Amazon Virtual Private Cloud.

There is always a chance to improve.

I have been using it for around 7 years. I started working with AWS and Amazon Virtual Private Cloud in 2018.

When we use business support, the availability of the engineers is very good.

Positive

It is very simple compared to my previous experience. I used to work in a data center with virtualization where we handled VLAN and other network virtualization and storage virtualization. This has been simplified to the core on cloud.

The only alternative I can think of is Microsoft Azure.

They have developed a product that works for both large-scale and small-scale industries. The UI is very user-friendly. The services are very well defined in AWS, and newcomers will not get confused, unlike in Azure.

I have started exploring AI, and I am using cloud and ChatGPT. I can say I am using prompt engineering.

I rate Amazon Virtual Private Cloud a nine out of ten.