I would say that Google Compute Engine is easy to use for skilled people and people with a tech background. For regular users who do not have experience with the cloud platform, it is not very user-friendly. If you wish to use Compute Engine, you must know a lot of technical terms such as availability, instance, and scalability. You also need to choose the proper instance for you. In my experience, there is a standard option, an advanced option, and a customized option for Google Compute Engine. You can choose the Compute Engine with high speed or with memory speed. All of this is too much for a person who is not tech-savvy. I am not sure whether they designed it like that on purpose or if they plan on improving it.

I cannot think of any features that need to be upgraded.

There have been instances when a customer has tried to deploy a certain number of VMs inside a project, and they come across quota issues. When they come across these quota issues, they have to reach out to the Google support team and ask them to increase their quota. Doing so typically slows down the deployment.

It would be better if there was an option to change the background. Like in Gmail, there's an option to change your theme. That's my recommendation.

Google Compute Engine needs to have multi-region support. It would also be nice to have a tracking mechanism. 

In terms of improvement, one is definitely the licensing piece. So there is a feature, the BYOL (Bring Your Own License) licensing piece, to bring your own license. It is not that straightforward. It requires some support from Google to get it sorted, access those licenses, and configure those licenses. It is not a very straightforward process. So, BYOL definitely needs some sort of improvement out there.

There can be many additional features in future releases. So GCE has been the core, pure module; from a feature set perspective, it has matured enough. But if there could be an auto-switching capability. Autoscaling is there, but it will differ or be restricted to certain types of machines.

So, for example, I use an N2 class machine or a memory-optimized M1 or M2. It will only scale in that particular configuration itself. The configuration might not change. Now, I should have that capability depending on the workload and what I deploy. So, depending on the type of workload, Google should recommend that this should be the recommended configuration based on which the instance can be set up. So, more of an automated configuration or a recommendation is something that I'll look for in future releases.

So, one feature I would like to see is auto-switching, or auto-scaling. The second feature is a recommendation based on the application.

I would like to improve the solution’s UI while deploying a container. It is sometimes hard to figure out the container’s details and format that you want to deploy. The tool does not give you a guide to finding out the error and why the container is not starting up which could be because you have configured it wrong. This is always a hit on the setup.

Google Compute Engine does not have many options at a lower tier level. If they had more options it will be better. For example, Amazon AWS or Microsoft Azure, have more options and different types of instances, of VMs we can select.

You can already run BigQuery SQL queries in Google Cloud Datalab but one thing I would like to see improved is the user experience when constructing these SQL queries, similar the tools that the Big Query web console offers. For example, I would like to see improved support for automatically formatting SQL queries. It would also be helpful to have a button that will check a query for SQL syntax errors before running the query which is also available in the BigQuery web console.

The solution could be more similar to something like ESXi. For example, in GCP, you don't get layer two connectivity between two VMs on the same subnet. So that could be great. Between two VMs on the same subnet in the same VPC.

Another area of improvement is the support. They did respond very fast, but they did not find a solution to our problem.

The biggest problem is that it's got a very arcane and complex security environment that has to be very carefully set up and maintained. Because GCP has is such a huge attack surface and is so prominent it any security lapses can quickly become fatal.

The problem with good security is that  it more difficult to use the GCP features. Faced with a difficult to use interface people will find ingenious ways to circumvent the security.

Because GCE is tied into the overall Google services security model I do not believe that it can be improved without making the user experience more difficult.

The GCE only has virtual machines with processors and video cards of the new generation in the North American region, the GCE has expanded in new areas, like Latin America, but it does not carry powerful hardware.