I see this post - Benchmark for dgraph
But would like know if there is a latest comparison/performance metrices
Our requirement :
Need performance metrices to substantiate the decision to have a Dgraph.
thanks in advance
Hi Antony, we’ve got some internal benchmarking that we’ve been doing along with the vector research. Let me see what we can clean up for public consumption.
Is there a specific type of benchmark you’re looking for? thru-put? data volume? type of query performance?
Thank you very much @KVG for a such a generous reply.
We’re primarily interested in understanding Dgraph’s performance under high write loads. Specifically, we’re processing around 300 million journal entries daily. We’re keen on knowing how Dgraph handles such a volume while maintaining good p99 latency.
Additionally, we’re curious about the performance implications of query types (e.g., high depth queries on huge data set and so on).
Finally, operational overhead is also a factor. We’re considering both self-managed and cloud-based options. Any insights into the management complexity and resource requirements would be valuable.
In short. :
- Throughput: Handling 300 million journal entries daily
- Latency: Achieving good p99 performance
- Complexity: Evaluating the operational overhead of running Dgraph in-house vs. cloud
- Cost-Benefit Analysis: Justifying the decision to adopt Dgraph
Hey @antony …had a similar uee-case earlier. Would love to hear what kind of reads do you guys do on top of this data?
hey, sorry for the late response. ours is sort of money transaction data.
Sorry for the delay-- here’s the script we use for benchmarking writes: GitHub - dgraph-io/dgraph-locust: locust testing framework for Dgraph, without initial data set to save space
We do some stuff for our own internal testing as well, e.g., modifying LSM options (numLevels
etc.) but this should at least get you started.
Here’s a sample of what the load test results look like:
thank you very much @KVG
This topic was automatically closed 24 hours after the last reply. New replies are no longer allowed.