Cassandra Reviews

We have some big logs, and we need to do some AI analysis. Currently, we are storing data on Cassandra, and we are checking which database has fast retrieval. We have done a couple of load tests and have found good results with Cassandra, so we are using it.

Its retrieval is similar to an RDBMS, so our team finds it easy to adapt. Our team is involved with the database mainly for retrieving data. We found a good capability in Cassandra, and we found this read replication somewhat easy compared to other options. For this reason, we want to use managed services. To forecast costs, we created our own PR custom launching to avoid some cloud costs.

We developed a Product Information Management system tailored for e-commerce. This homegrown system has evolved into a Software-as-a-Service offering available on the marketplace. The primary challenge addressed by this solution was managing vast amounts of data associated with products.

In this context, major e-commerce players like Amazon handle billions of images, regularly updating metadata and product attributes. The details visible on platforms such as Amazon, Alibaba, or Walmart, encompassing product specifications like laptop configurations, processors, RAM, etc., involve handling an enormous volume of information. 

For this use case, a key requirement was a reliable system with high uptime that did not adhere to the traditional master-slave or leader-follower architecture. Initially opting for Cassandra to meet these criteria, we encountered challenges related to the availability of skilled developers and Database Administrators. Consequently, we decided to transition to MongoDB to address these technical skill constraints.

The use of Cassandra in real-time data analytics has been pivotal for our e-commerce platform. As our platform operates 24/7, providing services to sellers and customers alike, the need for real-time updates is paramount.

For instance, when a customer leaves comments or feedback on an image, they anticipate an immediate reflection of these changes on the portal. Similarly, sellers altering product attributes or updating images expect instant visibility of these modifications.

Handling large data volumes with Cassandra has been an excellent experience. Despite challenges related to the influx, these were not attributed to Cassandra itself but rather to middle-layer issues. Generally, it demonstrated scalability with workloads, thanks to its horizontal scaling capabilities. We could easily add new nodes to the system as needed, ensuring the platform coped well with increasing loads.

The tool's most beneficial feature for scalability is its entire architecture. The absence of a single point of failure or a leader within the ecosystem contributes to its robust scalability. This key aspect influenced our decision to opt for the Cassandra ecosystem.

In terms of performance, it demonstrated the ability to handle approximately 1.6 billion requests per day. This was achieved on AWS using EC2 instances, and it was during a period about four to five years ago.

We use Cassandra mainly for key certificate and key management. Additionally, we use it for handling large-scale data management, particularly in the telecommunication nature of our applications.

Cassandra handles large-scale data management adequately without major issues, even during peak times. Its scalability fulfills our basic requirement for managing large data volumes.

This solution was a component of our messaging system. We used it as a high-capacity memory-based database.

The most valuable features of this solution are its speed and distributed nature. 

We use the tool to get unstructured data into a database so I can do data analytics. It's all internal; we do not sell it outside of the company. 

I gobble up all the information that is available to me. I grab emails sent to us, PDF files, and Word files and read the contents. I try to get labeling for those unstructured data and do analytics on it with Python.

Since I haven't had years of experience with it, it's still new to me. One valuable feature is its distribution, so I can run it partly in the cloud and part on-prem. That's a feature I'd like to use but haven't yet because we're trying to move to Azure. I don't know if or when that will happen. Ideally, we'd have it distributed over the cloud and on-prem simultaneously, so if something happens to our on-prem, we can keep going in the cloud, like a pay-as-you-go model with Azure.

We used the product because we wanted a database to handle heavy write workloads. Our events had time series data, so we found the product useful.

Cassandra is available most of the time. It is quite good for what it is made for. We write tons of time series data for the DB. We can add almost one million columns to the solution. It is awesome.