Overview
PuppyGraph is a comprehensive graph analytics engine that integrates seamlessly with your existing lakehouse, such as Apache Iceberg, Apache Hudi and Delta Lake. PuppyGraph's query engine allows you to query the same dataset as a graph without additional ETL processes or duplicating data storage and processing.
Highlights
- No ETL Query the tables as graphs on your existing data lakes such as Apache Iceberg, Apache Hudi and Delta Lakes.
- Auto Scaling Scalability is no longer an issue for graphs with us. The data are auto-sharded and the compute and storage are separated.
- High Performance Low latency for complex queries such as 10-hop neighbors.
Details
Features and programs
Financing for AWS Marketplace purchases
Pricing
Free trial
Dimension | Cost/hour |
|---|---|
r6i.8xlarge Recommended | $3.60 |
c6i.16xlarge | $3.60 |
r6i.2xlarge | $0.90 |
c6i.32xlarge | $7.20 |
r6i.4xlarge | $1.80 |
c7i.24xlarge | $5.40 |
c7i.16xlarge | $3.60 |
Dimensions summary
Top-of-mind questions for buyers
Vendor refund policy
No Refunds
How can we make this page better?
Legal
Vendor terms and conditions
Content disclaimer
Delivery details
64-bit (x86) Amazon Machine Image (AMI)
Amazon Machine Image (AMI)
An AMI is a virtual image that provides the information required to launch an instance. Amazon EC2 (Elastic Compute Cloud) instances are virtual servers on which you can run your applications and workloads, offering varying combinations of CPU, memory, storage, and networking resources. You can launch as many instances from as many different AMIs as you need.
Version release notes
Additional details
Usage instructions
For more details, please find the latest document in https://docs.puppygraph.com/user-manual/ami-version
Resources
Vendor resources
Support
Vendor support
Please contact us at contact@puppygraph.com for more details.
AWS infrastructure support
AWS Support is a one-on-one, fast-response support channel that is staffed 24x7x365 with experienced and technical support engineers. The service helps customers of all sizes and technical abilities to successfully utilize the products and features provided by Amazon Web Services.

Standard contract
Customer reviews
Unified graph querying has driven deep multi-hop reasoning and transformed data governance
What is our primary use case?
Our main use case for PuppyGraph is as our unified zero-ETL graph query engine to power our next-generation GraphRAG and Enterprise Knowledge Path architecture. Our enterprise GenAI agents require deep multi-hop relational reasoning across structured, transactional data, consumer interactions, and tabular documents.
PuppyGraph helps us with multi-hop relational reasoning for our GenAI agents by assisting in querying our relational data warehouses and lakehouses, which in our case are Snowflake and PostgreSQL. It enables us to use OpenCypher and Gremlin without copying, moving, or duplicating any data, and it also helps us execute multi-hop relationship traversals of up to eight to 10 hops across 3.2 billion edges and 450 million nodes in real time.
How has it helped my organization?
PuppyGraph has positively impacted our organization by driving transformative capital efficiency across our pipeline infrastructure, our storage footprint, and our product delivery. The net financial ROI has been an audited 443% within 10 months of implementation, which corresponds to around 4.5X the amount invested being recovered within that time frame. This is a great achievement.
Additionally, we have seen savings in pipeline infrastructure and compute, as the need for dedicated ETL pipelines was eliminated, as was the immediate sync required for microservices to mirror warehouse data and standalone graph stores. This has reduced operational spend by around $530,000 annually.
Furthermore, we have eliminated storage duplication by querying data in place across existing data lakes. This avoids duplicating approximately 38 terabytes of transactional data sets into proprietary graph storage, saving us around $70,000 a year in redundant storage and cross-AZ egress costs.
The time to value has improved engineering velocity, as the time to production for RAG and GraphRAG features has come down from 14 weeks to five days. This is a 93.88% reduction in delivery lead time and saves approximately 3,300 engineering hours per year across our data and AI platform teams. The overall program delivery timelines have improved by 78%, enhancing project delivery by 83.22%.
The engineering teams have responded positively to these changes, investing the time saved into other priority tasks instead of manual work.
What is most valuable?
There are multiple features of PuppyGraph that I personally love. One of them is zero-ETL graph querying, which helps us connect with our data warehouses easily and link via dynamic JSON graph schemas. This enables us to explore relational tables, nodes, and edges in under seven minutes, which previously took three and a half hours, so it is a great reduction.
I also love the massive distributed graph scalability. The distributed vectorized compute engine helps us auto-shard queries across nodes and consistently execute 10-hop traversals across multi-billion edge data sets in under two seconds.
Additionally, I appreciate the unified single source of truth. It helps us commit to the underlying tables, and those commits are immediately queryable in the graph layer within zero seconds of ETL replication lag. This ensures that our AI agents never reason over stale data and always access the latest data without any lag or changes left.
The governance and security of PuppyGraph are very good, and I would rate it 10 out of 10 in that aspect. PuppyGraph's governance and security features have helped our organization, as it queries lakehouse tables directly where they originate. This delivers an exceptional governance posture for our enterprise applications with the help of inherent source-level governance and ABAC, which has supported the existing row-level security while helping us implement RBAC and column masking policies. It stabilizes operations inside Snowflake in our data warehouses and has achieved 100% policy consistency with zero security drift.
The accuracy and reliability of PuppyGraph's output are really great.
What needs improvement?
I have not experienced any pain points regarding PuppyGraph. Everything is all good on that front.
For how long have I used the solution?
We have been using PuppyGraph for the last two and a half years.
What do I think about the stability of the solution?
PuppyGraph is a 100% stable solution, and we have not encountered any challenges or problems while using it.
What do I think about the scalability of the solution?
PuppyGraph's scalability is very good. We can easily scale up and down with the queries at a time, and I do not feel that we have had a problem with that.
How are customer service and support?
We spoke with customer support six months ago, and they resolved our query within two hours.
Which solution did I use previously and why did I switch?
We did not use a different solution earlier.
How was the initial setup?
Deploying PuppyGraph in our environment was very easy, and it was completed within five working days from the date of purchase.
The configuration process was straightforward and was completed within three business days.
The integrations with the other AWS services we use are very straightforward, and I truly appreciate the fact that PuppyGraph has helped us.
What's my experience with pricing, setup cost, and licensing?
The procurement process was very easy, and we did not face any difficulty. The metering and billing experience is straightforward, and we did not encounter any challenges while managing it.
Everything regarding pricing, setup cost, and licensing is very good, and there were no surprises at all.
Which other solutions did I evaluate?
Before choosing PuppyGraph, we evaluated other options such as Neo4j Enterprise, Neptune, and warehouse joins.
What other advice do I have?
On a scale of one to five, I rate PuppyGraph a five.
I give it a five because of the exceptional ROI it has delivered.
If public cloud, private cloud, or hybrid cloud, which cloud provider do you use?
Virtual graph modeling has accelerated feature selection and now needs a path to a persistent graph
What is our primary use case?
The main use case for PuppyGraph is establishing a graph data model layer on top of our existing systems. We do not want to bring in a new product which is completely graph-oriented, mainly because we are not sure about what attributes have to be brought into the graph. PuppyGraph provides an easy virtualization layer wherein we can keep the data wherever it is, simply connect PuppyGraph to the sources and establish the graph model very easily to see whether we are able to get the data model correctly. We can check all the attributes and check the relationships. When it comes to graph modeling, checking whether one-to-one relationships exist is very challenging. However, with PuppyGraph, we are able to decide what kind of nodes and what attributes they should have. This way, PuppyGraph makes it much easier, particularly when you have a large number of attributes and you do not even know which attributes are needed for establishing a proper graph data model.
PuppyGraph helps manage large numbers of attributes and relationships by addressing the question of what kind of edges we need when we have a large number of many-to-many transactions and relationships. If we model all the many-to-many relationships at a transaction level, the number of edges will be too high in any graph. To avoid that, we need to figure out the optimal way by which we can have the minimum number of edges, minimum number of nodes, and minimum number of attributes and still be able to find and identify customer churn or fraud detection very easily. These are two complex algorithms where graph data modelers, even though they are technically well-equipped, will end up with a large number of edges and nodes. A tool PuppyGraph, which is not a real graph database but sits on the existing sources, helps us a lot. It helps us get the clear, proper, specific graph data model for our attributes and system properly and very fast.
PuppyGraph fits into our workflow mainly because we do not want to spend time on importing and exporting. Every time we add or remove an attribute, it takes a lot of effort to bring the data into a proper database such as Neo4j, TigerGraph, or Memgraph. PuppyGraph simply always connects to the source. It is a matter of simply specifying the attribute name. We are not doing a full import into the graph data model. We keep the data wherever it is but simply specify the list of attributes and list of graph pages, and we are able to see the revised graph data model in action immediately and easily. The traditional solutions help a lot only if you have a proper graph data model and you have established an import-export mechanism. If there is a huge amount of uncertainty and you are not sure about what nodes you have, what attributes you need, and what edges you need, a virtual graph database PuppyGraph helps us a lot. It is very easy to arrive at the proper data model with PuppyGraph than with any other solutions.
PuppyGraph offers the best features by being a virtual graph data model which sits on the other data sources that we already have. It makes it very easy for us to model the graph that we exclusively need, that graph model that gives us the best results. We are able to get there very fast, mainly because it is so easy to make corrections with PuppyGraph.
What is most valuable?
The flexibility and speed of PuppyGraph help my team in case we have 300 attributes and we are not sure about the importance of every attribute. We will not load a lot of data into Neo4j, TigerGraph, or Memgraph without even knowing which attributes will be helpful. However, with PuppyGraph, we are able to design the graph with all 300 fields and then gradually remove what we do not want, so easily, that we are able to trim the 300 features to the real 40 features very fast. We are not doing a real import into a real graph. We are simply keeping the attributes. PuppyGraph simply points to the source all the time. This way, it saves a lot of time to come down to the real 40 fields you need than when you start with 300 fields. PuppyGraph helps you to evolve very fast when there is uncertainty, within a matter of four or five days, to arrive at the best graph model.
PuppyGraph has positively impacted our organization because we have a huge warehouse with a huge number of fields every day. It is easy for us to bring the elemental view of the data very fast. We can add an attribute easily or remove an attribute easily, mainly because we only specify. We do not import the data. The traditional mechanism of taking the data, importing it into Neo4j, and adding a new feature disappears with a solution like PuppyGraph. PuppyGraph is a virtual graph, not a real graph database. This is the best case when you are not sure about what graph model you need and you have too many features. PuppyGraph makes it very easy, mainly because you do not spend a lot of time trying to import and export the data. You profile the data very fast in a graph and get the elemental view of any data as a graph very fast. PuppyGraph is the best answer.
What needs improvement?
PuppyGraph is a virtual graph and not a real graph data model. However, at some point in time, once we freeze this as the graph and the graph data model, we would like to extend to a proper graph, mainly because some of the attributes in the source will change over a period of time. We would like to maintain the historical data. Therefore, once we build a virtual layer, we need a real concrete graph layer. PuppyGraph has not provided the concrete graph layer or a mechanism by which we can build our own concrete layer on top of the virtual layer. This functionality will be needed if we have to keep the investment on PuppyGraph after we discover what attributes we need. If we need to continue with PuppyGraph, we need a compelling reason. It has to be extended with a real graph database solution.
PuppyGraph could be improved because at some point in time, it always connects to the source system, which means that the source system will always be overloaded because it is a virtual graph. At some point in time, we should be able to decide to bring in the data, but PuppyGraph does not have the support for it. It always depends on the source. The source systems are overloaded once we finalize the graph. PuppyGraph does not have a real native engine for a graph. It has to be extended with a native engine for a graph in the future.
For how long have I used the solution?
I used PuppyGraph for about four years.
What do I think about the stability of the solution?
PuppyGraph is very much stable.
What do I think about the scalability of the solution?
The scalability of PuppyGraph is really good in terms of the volume of data. However, they cannot support many users. PuppyGraph essentially does not have a full-fledged native engine. The load is still on the source system. PuppyGraph can scale a lot. However, the load is transferred to the source system. This is the reason why it can be used only by a limited number of users because the actual source transfers the load to the source systems.
How are customer service and support?
The customer support is very satisfactory.
Which solution did I use previously and why did I switch?
We have not used a different solution before PuppyGraph. We considered Neo4j, Memgraph, and TigerGraph as we were using PuppyGraph side by side. PuppyGraph is not a long-term solution as it is a virtual graph. Eventually, after we get the model right, we needed to move to Memgraph, another graph database.
What other advice do I have?
There are no concrete recommendations. The important thing is that if they have to choose PuppyGraph, they need to have the clarity that PuppyGraph is not a long-term solution. Only if they have a huge collection of attributes and they do not want to spend time importing every attribute into Neo4j, Memgraph, or TigerGraph, then only they can consider PuppyGraph. PuppyGraph is a short-term solution for you to arrive at the correct list of attributes and how they should be modeled. It is not a long-term solution.
I have no specific comments on PuppyGraph because we went with the clear awareness that PuppyGraph is not a full solution for a graph database. The overall review rating for PuppyGraph is 7 out of 10.