OK, so you are running a simple benchmark and do not yet have any real data?
About the benchmarking script; are you actually creating a new instance of the client for each bulk request? That is quite inefficient. You should generally create the client at the start and then reuse it, so that you can benefit from persistent connections.
If you just want to benchmark, and make sure you do so efficiently, you can create a large data file with a single document in JSON format per line and then use e.g. [our benchmarking tool Rally](https://github.com/elastic/rally
) to run it against the cluster. It may be even more interesting to use what you will be using in production, e.g. Filebeat and/or Logstash.