Support of variable object size for single xk6 run #161

New issue

Open

opened 2024-08-12 08:07:39 +00:00 by a-baranov · 3 comments

a-baranov commented

2024-08-12 08:07:39 +00:00

Customer who has known payload (e.g. dump of object to upload) asks to estimate upload of objects of variable object size.

sample task sounds like:

Test upload of the dataset:
upload objects in size range 2k - 8k, 700 streams
upload objects in size range 8k - 64k, 500 streams
upload objects in size range 64k - 512k, 240 streams
upload objects in size range 512k - 4000k, 24 streams

To support this request it is required to add variable object size (randomized in specific range or array of values).

Customer who has known payload (e.g. dump of object to upload) asks to estimate upload of objects of variable object size. sample task sounds like: **Test upload of the dataset:** upload objects in size range 2k - 8k, 700 streams upload objects in size range 8k - 64k, 500 streams upload objects in size range 64k - 512k, 240 streams upload objects in size range 512k - 4000k, 24 streams To support this request it is required to add variable object size (randomized in specific range or array of values).

fyrchik commented

2024-08-12 08:23:48 +00:00

Owner

WRITE_OBJ_SIZE should accept 2000..8000 or 2000-8000 syntax.
The distribution should be uniform, we can customize it later, if there will be a necessity.

`WRITE_OBJ_SIZE` should accept `2000..8000` or `2000-8000` syntax. The distribution should be uniform, we can customize it later, if there will be a necessity.

fyrchik commented

2024-08-27 08:51:05 +00:00

Owner

The implementation should reuse the existing datagen buffering (probably with the max size) and only use variable at read stages.

The implementation should reuse the existing datagen buffering (probably with the max size) and only use variable at `read` stages.