PDPS-Bench TPCTC2024@VLDB
Published:
⚠️ 𝗢𝗻𝗲 𝗼𝗳 𝘁𝗵𝗲 𝗸𝗲𝘆 𝗰𝗵𝗮𝗹𝗹𝗲𝗻𝗴𝗲𝘀 𝗶𝗻 𝗺𝗮𝗰𝗵𝗶𝗻𝗲 𝗹𝗲𝗮𝗿𝗻𝗶𝗻𝗴 𝗲𝘀𝗽𝗲𝗰𝗶𝗮𝗹𝗹𝘆 𝗳𝗼𝗿 𝗼𝗽𝘁𝗶𝗺𝗶𝘇𝗮𝘁𝗶𝗼𝗻 𝗽𝗿𝗼𝗯𝗹𝗲𝗺𝘀, 𝗶𝘀 𝗴𝗲𝗻𝗲𝗿𝗮𝘁𝗶𝗻𝗴 𝗵𝗶𝗴𝗵-𝗾𝘂𝗮𝗹𝗶𝘁𝘆 𝗱𝗮𝘁𝗮. This data directly impacts the accuracy of models. In scenarios like parallel and distributed stream processing, the need for high-quality training and evaluation data becomes even more crucial.
💡 To tackle this, we introduced 𝗣𝗗𝗦𝗣-𝗕𝗲𝗻𝗰𝗵, a performance benchmarking system designed to evaluate parallel and distributed stream processing (DSP) in heterogeneous environments. 𝗣𝗗𝗦𝗣-𝗕𝗲𝗻𝗰𝗵 𝗮𝗶𝗺𝘀 𝘁𝗼 𝗽𝗿𝗼𝘃𝗶𝗱𝗲 𝗵𝗶𝗴𝗵-𝗾𝘂𝗮𝗹𝗶𝘁𝘆 𝗽𝗲𝗿𝗳𝗼𝗿𝗺𝗮𝗻𝗰𝗲 𝗯𝗲𝗻𝗰𝗵𝗺𝗮𝗿𝗸 𝗱𝗮𝘁𝗮𝘀𝗲𝘁𝘀 to ensure that DSP optimization mechanisms using machine learning can be effectively trained and fine-tuned in diverse and dynamic environments.
⭐ I had the privilege of presenting our paper 𝗣𝗗𝗦𝗣-𝗕𝗲𝗻𝗰𝗵 at 𝗧𝗣𝗖𝗧𝗖 𝟮𝟬𝟮𝟰, held as part of the 𝗩𝗟𝗗𝗕 𝗖𝗼𝗻𝗳𝗲𝗿𝗲𝗻𝗰𝗲 in China and sharing our work with the global data management community.
