# How about a graph compute engine based on dgraph except for spark?

**URL:** <https://discuss.dgraph.io/t/how-about-a-graph-compute-engine-based-on-dgraph-except-for-spark/11465>\
**Category:** Dgraph\
**Tags:** kind:question\
**Created:** [November 16, 2020, 7:15am UTC](https://discuss.dgraph.io/t/how-about-a-graph-compute-engine-based-on-dgraph-except-for-spark/11465 "2020-11-16T07:15:42Z")\
**Posts on this page:** 5\
**Page:** 1

<div class="post-metadata">

**Author:** ![yilong2001](https://yyz1.discourse-cdn.com/flex007/user_avatar/discuss.dgraph.io/yilong2001/32/5823_2.png) [@yilong2001](https://discuss.dgraph.io/u/yilong2001)\
**Post date:** [November 16, 2020, 7:15am UTC](https://discuss.dgraph.io/t/how-about-a-graph-compute-engine-based-on-dgraph-except-for-spark/11465/1 "2020-11-16T07:15:42Z")

</div>

dgraph is a graph database. How about a graph compute engine?  
Does Spark go best with dgraph?  
Does anybody create a native graph compute engine based on dgraph, is it valuable?

---

<div class="post-metadata">

**Author:** ![RJKeevil](https://yyz1.discourse-cdn.com/flex007/user_avatar/discuss.dgraph.io/rjkeevil/32/2625_2.png) [@RJKeevil](https://discuss.dgraph.io/u/RJKeevil)\
**Post date:** [November 16, 2020, 8:54am UTC](https://discuss.dgraph.io/t/how-about-a-graph-compute-engine-based-on-dgraph-except-for-spark/11465/2 "2020-11-16T08:54:59Z")

</div>

Agree this would be a very handy feature. If you google e.g. “DGraph PageRank” you will find some attemps to do this, but being unnofficial, the repositories tend to go stale. Would be amazing to have something more official.

My company did indeed go down the spark route for this for now. You are probably aware already but there is a very useful spark DGraph connector here [GitHub - G-Research/spark-dgraph-connector: A connector for Apache Spark and PySpark to Dgraph databases.](https://github.com/G-Research/spark-dgraph-connector)

---

<div class="post-metadata">

**Author:** ![yilong2001](https://yyz1.discourse-cdn.com/flex007/user_avatar/discuss.dgraph.io/yilong2001/32/5823_2.png) [@yilong2001](https://discuss.dgraph.io/u/yilong2001)\
**Post date:** [November 18, 2020, 1:53am UTC](https://discuss.dgraph.io/t/how-about-a-graph-compute-engine-based-on-dgraph-except-for-spark/11465/3 "2020-11-18T01:53:04Z")

</div>

Thanks very much!

---

<div class="post-metadata">

**Author:** ![BlankRain](https://yyz1.discourse-cdn.com/flex007/user_avatar/discuss.dgraph.io/blankrain/32/1592_2.png) [@BlankRain](https://discuss.dgraph.io/u/BlankRain)\
**Post date:** [November 19, 2020, 7:22am UTC](https://discuss.dgraph.io/t/how-about-a-graph-compute-engine-based-on-dgraph-except-for-spark/11465/4 "2020-11-19T07:22:17Z")

</div>

you can read graph data from dgraph ,then load it to spark graphx as RDD,then do some computing jobs, then rewrite it back to dgraph.

In neo4j, it has a graph-loader to load data into memory,if you are able to write a graph-loader for dgraph, then you can use neo4j 's algorithm.

Also, there is a graph api in nvidia 's CUDA package. you can read your data from dgraph ,then call the API to compute .

---

<div class="post-metadata">

**Author:** ![yilong2001](https://yyz1.discourse-cdn.com/flex007/user_avatar/discuss.dgraph.io/yilong2001/32/5823_2.png) [@yilong2001](https://discuss.dgraph.io/u/yilong2001)\
**Post date:** [December 28, 2020, 7:57am UTC](https://discuss.dgraph.io/t/how-about-a-graph-compute-engine-based-on-dgraph-except-for-spark/11465/5 "2020-12-28T07:57:08Z")

</div>

thanks very much!