Amazon Neptune Reviews

We managed traffic at Zetta, and traffic would be moving between multiple services in our microservice architecture. Because of this setup, we were using Amazon Neptune to understand how many requests were going from one service to another service, how many were successful (2XX), and how many were failing (5XX).

We had an architecture where requests would go via multiple ingress objects, and we needed to identify which cluster was being hit the highest, which service was being hit the lowest, and which service was most unhealthy in terms of providing more 5XX responses. To get visibility on all our services present across Zetta, we used Amazon Neptune, where we represented every service as a node and represented edges as requests and responses.

The onboarding part and documentation where we could ideally use Amazon Neptune is excellent. Amazon Neptune as a product by AWS is exceptional because it supports multiple graph models such as RDF and property graph. It also has support for multiple querying languages such as Gremlin, SparkQL, and OpenCypher. It is very comprehensive in supporting every requirement we had at Zetta.

Amazon Neptune's best features include its multiple servers, each supporting different languages such as OpenCypher, SparkQL, and RDF. For the same RDF graph or property graph, we could use multiple languages to query on different servers. This is exceptional because we have one graph DB with two endpoints exposed where we could interact with different languages on the same graph. Additionally, Amazon has a Sagemaker Jupyter Notebook which interacts with the Amazon Neptune database itself, providing a clean UI for representing nodes since the Jupyter Notebook has predefined graph representation capabilities through queries.

The main issue was the limited community of Amazon Neptune users, which meant everything needed to be explored independently. Although this was adventurous, it required more time investment in the product.

We had a strict time constraint, and it took many sleepless nights to find information in the documentation. There was even a section within the documentation that contained incorrect information.

Apart from the community, documentation aspect, and technical support, everything in Amazon Neptune is already excellent. As Amazon consistently provides great services, these three areas are the only ones where I could see room for improvement.

I worked with Amazon Neptune for almost eight months, though I am not currently using it.

We did not experience any issues with Amazon Neptune. The service was never down during our POC, so we didn't need to test its self-healing capacity.

I attempted to contact support but did not receive a response.

Negative

We weren't using any graph DB at the beginning. We later adopted Amazon Neptune.

We assessed Neo4j alongside Amazon Neptune. We chose Amazon Neptune primarily because all our infrastructure is maintained by AWS, so we preferred not to manage any additional services ourselves. The second reason was that Neo4j only supports property graphs, whereas Amazon Neptune also works with RDF.

The initial setup for Amazon Neptune was very straightforward.

The cost aspects were managed by our SRE team who provisioned the instances. The pricing structure is similar to how EC2 instance pricing varies. I was informed that it was somewhat expensive, though I'm not certain about the specific reasons or potential optimizations.

One difference we found was that Neo4j has the in-place feature to override edge values between multiple nodes, which is not available within Amazon Neptune. Additionally, transitive relations were not definable on Amazon Neptune.

We used Amazon Neptune for making POCs for one of our projects, utilizing its real-time capabilities. There were approximately five to six different relational lists that were managed efficiently using Amazon Neptune.

Amazon Neptune helped us by providing meaningful insights in one place and in one iteration. Previously, it required three to four iterations and interaction between teams to get the same level of observability that Amazon Neptune provides immediately.

It's an excellent product, but potential users should ensure they have enough time to explore all the capabilities of Amazon Neptune before adoption. The time investment and community aspects should be kept in mind when considering Amazon Neptune as a solution.

On a scale of 1-10, this solution rates an 8.

In my organization, we need to maintain hierarchies of employees and customers. We also have to track the history of these hierarchies, which is why we are using a graph database. Our earlier choice was Neo4j, but due to cost issues, we switched to Neptune.

Unlike Neo4j, which has visual representations of the graph, I haven't seen that in Neptune. I think Neptune should introduce that.

I would like to see a visual representation of the graphs. By graphs, I refer to the data structure in the graph database. We should be able to visually see how the nodes we insert are connected. 

For example, on Facebook, people have multiple friends who have their own friends. We should be able to visually see the connections between people and how they are related when we run a query, instead of just getting results in JSON. This visual representation is helpful for analysis and other purposes.

I have recently used it. It has been six months. 

So far, the stability has been fine.

There are around two end users. 

The customer service and support are good. 

Positive

We were on Neo4j earlier, and the primary reason for switching was cost.

The initial setup is actually simple. We don't have much trouble with it.

In my scenario, the integration wasn't easy because we're using the Quarkus framework in Java. I had to figure out some other methods, some basic raw methods, to run the query. But they do have built-in SDKs that can be used for plain Java or Spring Boot applications. So, I would rate the integration difficulty as above average.

It is not easy for a beginner to learn to use Neptune. You would need prior knowledge of graph data or similar concepts. It's about having a basic understanding.

Opting for this solution totally depends on your parameters. One parameter is definitely the cost. For a single instance, the monthly cost of Neo4j is $65 without the visual representation and over a hundred with it. In terms of AWS, it's around $10 to $20  a month. So, from a cost perspective, you should go with Neptune. It gives a similar performance to Neo4j. 

Also, consider if your developers have a good understanding of the graph data structure they are going to implement in the graph database. If so, I think Neptune is a fine tool.

Overall, I would rate the solution a nine out of ten.

So, I'm not using it currently, but when I was working with my last organization, which was a gaming company, we used Amazon Neptune for real-time fraud detection. We maintained user data, mobile device data, and other information as graph data. These vertices and the connections between them (edges) were stored along with properties. A real-time pipeline was used to input data into the Neptune database, and several data science jobs triggered every minute or so for real-time fraud detection.

It allowed us to maintain complex user data relationships and carry out real-time fraud detection efficiently. The structured data in graphs made debugging easier using Jupyter Notebooks which added significant value.

The integration with Jupyter Notebooks was highly valuable. It was very easy to debug using Jupyter Notebooks. Also, it was the first graph database I used, and the concept of maintaining IDS files was easy to manage.

Some improvements could be made by introducing a UI that would clearly show the vertices and connections, which would make it easier to visualize and debug. Additionally, the need for downtime during scalability is a concern. Also, many new features were not backward compatible, requiring core system upgrades.

I used Amazon Neptune for two years.

I did not experience any issues or outages with Amazon Neptune while working at my last organization. It was smooth and reliable.

We did not try scaling, but it seemed to require downtime, which is different from our experience with RDS that did not need downtime for scaling.

The customer service was great. Every time we had an issue, they were approachable and supportive. They assisted completely with setting up and designing our data within Neptune, even though many new features required upgrading due to lack of backward compatibility.

Positive

We used RDS, which did not require downtime for scaling unlike Neptune.

The initial setup was straightforward, thanks to AWS taking care of most aspects. Deploying a new database was just a click away.

When taken as part of the whole AWS contract, it was cheaper, but as a single service, Neptune was more expensive compared to Neo4j.

We compared Neptune with Neo4j, which was cheaper as a single service.

I recommend Amazon Neptune and hope it allows querying languages other than Gremlin. This will enhance its integration capabilities.

The Amazon Neptune is a graph database. In the graph database, instead of relational tables, you have nodes and edges. For example, in a conversation between two people, the two people would be a node, and they are connecting because they are talking and that would be an edge.

If you can imagine the model spreading out, this person knows this person, this person works with this person. It's more of a LinkedIn or Facebook-type relationship, which has its benefits.

Relational databases are never good at identifying patterns in graphs or other similar relationships, whereas Amazon Neptune is. 

Amazon Neptune is good for identifying fraud and many banks are using it. For example, know that I go to certain places all the time that are connected with nodes and edges, such as gas stations, shops, and Amazon. Since people are creatures of habit, the banks can use those nodes and edges to make a model of where I like to shop and spend money. If they see something pop up that's out of character for me, then they know immediately that it is possible fraud. They can determine this because a purchase I made was somewhere I never go or because of geography.

Amazon Neptune could improve by spreading more awareness for others to have an understanding of the solution because the technology is fairly new. The developer community and larger community do not understand it yet.

I have used Amazon Neptune within the last 12 months. 

The stability of Amazon Neptune is very good. Many companies, such as Amazon and Facebook, couldn't make things work with a relational database. Relational databases are too restrictive to do the things they wanted to. This is how these new types of databases came to be, they needed a better solution.

Amazon Neptune was built to be a scalable solution.

The implementation of Amazon Neptune is straightforward. With my Amazon AWS account and VPC, I can create a Neptune database. You then can start entering data.

Microsoft Azure and Amazon AWS are on par for pricing and Google has been raising its prices.

I have evaluated Microsoft Azure and Google Cloud.

My advice to others wanting to start using Amazon Neptune would be to start small and it is more geared toward the developer that's creating the website. For example, if they were creating a LinkedIn website, they are going to use that type of database, there's is not much more to do. These databases are geared more towards developers where DBAs don't have much to do with the Neptune type database.

I rate Amazon Neptune a nine out of ten.