Closed mulugetam closed 1 month ago
@mulugetam thanks for raising the PR. Could you please share some performance numbers for these modes?
@mulugetam thanks for raising the PR. Could you please share some performance numbers for these modes?
Here are some performance numbers for indexing using stack overflow workload <html xmlns:v="urn:schemas-microsoft-com:vml" xmlns:o="urn:schemas-microsoft-com:office:office" xmlns:x="urn:schemas-microsoft-com:office:excel" xmlns="http://www.w3.org/TR/REC-html40">
| `qat_deflate` relative to `deflate` | `qat_lz4` relative to `default` -- | -- | -- Total time | -11.4% | -2.5% Mean Indexing throughput | 23.7% | 3.4% Store Size | 1.6% | -1.76%
Description
Adds hardware-accelerated DEFLATE and LZ4 compression codecs for stored fields. The hardware in focus here is Intel (R) QAT, which is an integrated, built-in accelerator on the latest 4th and 5th Gen Intel Xeon processors. The implementation relies on the Qat-Java library.
The PR adds two additional valid values for
index.codec
:qat_deflate
andqat_lz4
. It also introduces a new setting,index.codec.qatmode
, that specifies the mode of execution for QAT.Two values are supported for
index.codec.qatmode
:hardware
andauto
. Ahardware
execution mode uses only the QAT hardware, while anauto
execution mode may switch to software if hardware resources are not available.Closes
https://github.com/opensearch-project/custom-codecs/issues/130
By submitting this pull request, I confirm that my contribution is made under the terms of the Apache 2.0 license. For more information on following Developer Certificate of Origin and signing off your commits, please check here.