There are significant challenges when dealing with external applications, and vendors, or pursuing a hybrid cloud strategy because AWS Lake Formation is specific to AWS and does not integrate seamlessly with third-party services or other cloud providers. Another limitation is that Lake Formation might not offer out-of-the-box automated reporting, so you may need to create your solutions for tracking and reporting on permissions granted to users. It falls short when it comes to more granular access control, such as cell-level or row-level entitlements which is a significant drawback for organizations that require precise control over who can access specific rows of data.

The solution could make improvements around orchestration and doing some automation stuff on AWS front automation. It would be useful if we could use automation to build images and use hardened images which are CIS compliant. 

For AWS, currently, we are facing the challenge when previewing. On AWS, we have set up a scanner instance with a Nexus scanner. For some reason, the scanner instance has been set up on AWS so that if it's in one region or maybe if it is deployed on one subnet, the other subnet is not reachable within the same region. The scanner is not reachable to the different subnets. For example, if there is a subnet A where the scanner is deployed, and if we want to scan the subnet B within the same region, the scanner is not reachable due to the fact that it's in Subnet A. That's a problem we are currently facing and trying to troubleshoot that.

AWS Lake Formation's pricing could be cheaper.

We occasionally work with third-party libraries and they fail to integrate with this product effectively.

For the end-users, it's not as user-friendly as it could be. For example, if I have a Lake Formation return, once the code is in place, then working Lake Formation, or deploying Lake Formation code, still requires an engineer as opposed to being automated to the extent where end-users can actually deploy the whole infrastructure. 

The other issue is the debugging on the solution is not so straightforward. This should be simplified.

In our experience what could be improved are not the support, performance or monitoring, but at a managerial level, the very expensive professional services of AWS. This could be an area of improvement for them. It's too expensive to acquire their support.

Generally, when you adopt a new technology, you always have to acquire a basket of professional services from the vendor. It's not only for system integrators, but also for clients because they always need some support. In this case, AWS is very expensive.

Something else they could improve is covering the end-to-end of the data platform with AWS products. They have to evolve. They have to improve the tools of data visualization a lot with QuickSight and ETS tool Glue. If they want to provide an end-to-end service with all the technologies to build up a complete data platform they need to improve these two technologies, these two solutions.