Share your experience using BitRefinery

Microsoft Azure is a ticketing system that provides support services to Microsoft partners based on their partnership tier. Partners can raise support tickets, particularly for production-related issues. When we had issues with our Cosmos DB instance in production, we would submit a support ticket with screenshots detailing the error. We usually receive a call from a support engineer the same day.

Once a support engineer was assigned, the process was well-organized. Each engineer worked shifts of about eight to nine hours, and before their shift ended, they would email us to let us know who would be taking over, for example, "John Doe." They would also provide a summary of the issue, progress made, and next steps, ensuring a smooth handover without the need to re-explain the problem to the next engineer. It was a very efficient process.

Microsoft Fabric is a new service that integrates various functionalities into one platform, such as data engineering, data science, and data visualization. While promising, some complexities and limitations become apparent only after you use the service.

Fabric is a comprehensive and expensive service with tier-based pricing. For someone new to the platform, like those coming from Snowflake or Databricks, it can be difficult to grasp the cost implications and potential limitations without extensive exploration. The documentation outlines these limitations.

Pricing is very expensive.

I have been using Microsoft Azure for five years.

Cosmos DB offered better scalability. In Azure SQL, we had to carefully manage the storage tier, switching back and forth based on needs. But when we first implemented Cosmos DB, we had a good understanding of the required data volume over the past three months. This helped us select the appropriate tier because we had those figures.

If we had done it the other way around, it would have been much harder to estimate the data accurately. While you can get a rough idea, whether it works out in practice is more of a hit-and-miss approach, which we have already experienced with Cosmos DB. We applied the knowledge gained from that exercise to build a similar solution over Azure SQL.

On the DevOps side, our team consisted of five members, ranging from senior to junior roles. We would raise requests for specific services needed in the development environment. The process typically takes three to five business days, including all necessary approvals, from submitting the request to receiving the resources and verifying access.

Sometimes, even with the documentation available, understanding how it applies to our specific issues requires contacting a system engineer and creating a support ticket. The support engineer often provided additional details that were not included in the official documentation. This highlights the gap between the provided information and practical, real-world use cases.

Positive

Deployment was relatively straightforward. We didn’t encounter any issues whether we were using the GUI mode or working with our DevOps team, who utilized Terraform to provision resources as needed.

Most people use Azure Synapse for data onboarding and integration. Synapse is a big data analytics platform where data engineers and scientists can collaborate on a unified platform. Another service called API Management serves data outside Azure to downstream consumers or third parties. This is useful for cases where third parties can't be onboarded onto your system due to compliance or regulatory issues or need a specific slice of your data in near real-time. When we used Cosmos DB, we hosted near real-time data with a latency of about 15 minutes and served it via API M. With API M, authentication tokens were shared with the end users, allowing us to know precisely who was consuming data outside of our Azure environment. This setup was unique because it required providing near real-time information with minimal latency, even as the number of concurrent users increased. Cosmos DB worked well in this scenario due to its horizontal scalability. We also hosted similar data on Azure SQL, as Cosmos DB can be more expensive for cases where a smaller amount of data needed to be served, such as just the last one to three months of data. Azure SQL was a good alternative. The same API M interface was used to serve data from Cosmos DB and Azure SQL. Under the hood, one API would fetch data from Cosmos DB, and another would pull from Azure SQL. This architecture was tested and worked well for our client.

It was easy to maintain. Overall, I rate the solution an eight out of ten,

My usual use cases for Google Compute Engine involve deploying the backends and APIs only.

What I like the most about Google Compute Engine is SSH access and the different price categories. I have used custom machine types in Google Compute Engine, but not for large servers, only for smaller ones. While the deployments are sometimes much faster compared to the normal or static servers, the velocity is notable.

Currently, I have not found much value in the features or capabilities of Google Compute Engine because I only use SSH access, as I have not used many of their services. It is already set up by a different team, and I have been using them as it is a normal case, similar to a normal Linux server.

I have noticed a positive impact since the deployment of Google Compute Engine; it is simpler. However, for those who are not very much familiar with Google Cloud, it can, at times, be difficult. Also, the freebies are not much from Google's side, referring to the credits, so that users can become more familiar since there is a limited amount or it is case-dependent.

Without trying Google Compute Engine totally, I cannot tell you much about it. From my personal perspective, I suggest that while setting up Google Compute Engine, the process is hard compared to AWS. The complexity increases, especially if we want to access the things over the API; we need to explicitly set up things and can only access the CLI environment of GCP, which is sometimes a hectic task.

I have been working with Google Compute Engine for not more than seven or eight months.

The stability of Google Compute Engine is on par with other servers, akin to using instances from AWS, as no one needs any downtime. Everyone is facing the competition and needs reliable uptime. For instance, I remember when the US East 2 region experienced issues a couple of months ago; it was concerning, but there were no complaints about it. My engagement with Google Compute Engine has only been a small primary work for a few days, as I do not use it daily or anything else regularly.

I would say scalability is how deep my experience goes; I have not worked more than eight months, with about two months for learning and the remaining six months on work that got finalized and handed over to the client. It is entirely dependent on how other teams want to execute their future programs. If it comes to us, we will handle it; otherwise, we will pursue other options.

I do not usually communicate with Google's technical support because I have not had a chance, and from my side, I am comfortable with GCP. I can start with things on my own, and if I do not have any requirements, there is no need to reach out.

For me, the initial setup and deployment of Google Compute Engine were straightforward, but I would not know what others have faced. At times, when I started with GCP during my college days, it was not much easier. Since starting to work with GCP, it is sometimes easier, sometimes hard, and sometimes complex as well; it totally depends on the criteria.

Regarding the pricing and licensing of Google Compute Engine, when I refer to licensing, it implies the price or a credit-based system. While doing my personal projects, I prefer AWS over GCP and usually suggest the same to others because I have tried everything in AWS. It is difficult to recommend GCP due to its complex architecture.

The effectiveness of Google Compute Engine's global network for meeting my applications' low latency requirements totally depends on the use cases; it is not comparing the services of each other. It is essential for the permanent use cases. My experience so far mostly involves GCP, GKE, and GCE, and I have not worked extensively on many other aspects.

I have not experienced the importance of Google Compute Engine's live migration feature for maintaining service availability during maintenance, so I cannot comment on it.

I have not integrated Google Compute Engine with any other Google Cloud services such as BigQuery; I just installed the services, and everything was done by another team.

I would rate this review a seven overall.