Proceedings Article10.1109/bigdata59044.2023.10386414
Colocation Datacenter Customer Power Usage Forecasting Using Synthetic Data and Integration of Macroeconomic Indicators
Neda Zarayeneh,Malarvizhi Sankaranarayanasamy,Pegah Mavaie,Omanshu Thapliyal,Prasun Singh,Ravigopal Vennelakanti +5 more
- 15 Dec 2023
pp 3453-3457
TL;DR: This work proposes an innovative power forecasting system that takes into account both internal and external demand signals, leading to more accurate and realistic predictions, and synthesized a comprehensive dataset tailored to the data center environment using the small dataset provided by S&P Global.
read more
Abstract: Colocation data centers play a pivotal role in the digital era, serving as the backbone for a diverse range of businesses, from startups to large enterprises. The global data center construction market, which reached an approximate value of US ${\$}$218.88 billion in 2021, is experiencing unprecedented growth. This surge can be attributed to the ever-increasing volume of data, propelled by economic advancements and population expansion. Consequently, long-term private equity firms and real estate investment trusts (REITs) are increasingly drawn to data center investments for their attributes of transparency and accountability. However, data centers face challenges in terms of project lead times and budget constraints, particularly during expansion phases. In response to these challenges, we propose an innovative power forecasting system that takes into account both internal and external demand signals, leading to more accurate and realistic predictions. These predictions serve as valuable guides for resource allocation and utilization. The system lever- ages a combination of data generation, deep learning (DL), and time-series analysis methods, and later, we use it in our future work to proactively address issues such as supply shortages, ensure the maintenance of Service Level Agreements (SLAs), and optimize resource usage. We synthesized a comprehensive dataset tailored to the data center environment using the small dataset provided by S&P Global. This dataset was enriched with macroeconomic data to capture external influences accurately. Subsequently, we conducted a rigorous evaluation, testing various machine learning models, including linear models, transformer-based models, and a multivariate LSTM model. Our experiments revealed that the PatchTST model outperformed the others, providing the most reliable and precise results. The implementation of advanced analytics further enhances energy efficiency, optimizes equipment utilization, and maximizes the effective utilization of floor space within data centers. Furthermore, efficient resource allocation, guided by the power forecasting system, ensures that customer demands are met promptly and effectively.
read more
Chat with Paper
AI Agents for this Paper
Find similar papers on Google Scholar, PubMed and Arxiv
Write a critical review of this paper
Analyze citations of this paper to find unaddressed research gaps
References
Long short-term memory
TL;DR: A novel, efficient, gradient based method called long short-term memory (LSTM) is introduced, which can learn to bridge minimal time lags in excess of 1000 discrete-time steps by enforcing constant error flow through constant error carousels within special units.
99K
•Proceedings Article
Attention is All you Need
Ashish Vaswani,Noam Shazeer,Niki Parmar,Jakob Uszkoreit,Llion Jones,Aidan N. Gomez,Lukasz Kaiser,Illia Polosukhin +7 more
- 12 Jun 2017
TL;DR: This paper proposed a simple network architecture based solely on an attention mechanism, dispensing with recurrence and convolutions entirely and achieved state-of-the-art performance on English-to-French translation.
Attention Is All You Need
Ashish Vaswani,Noam Shazeer,Niki Parmar,Jakob Uszkoreit,Llion Jones,Aidan N. Gomez,Łukasz Kaiser,Illia Polosukhin +7 more
- 01 Jan 2017
Abstract: The dominant sequence transduction models are based on complex recurrent or convolutional neural networks in an encoder-decoder configuration. The best performing models also connect the encoder and decoder through an attention mechanism. We propose a new simple network architecture, the Transformer, based solely on attention mechanisms, dispensing with recurrence and convolutions entirely. Experiments on two machine translation tasks show these models to be superior in quality while being more parallelizable and requiring significantly less time to train. Our model achieves 28.4 BLEU on the WMT 2014 English-to-German translation task, improving over the existing best results, including ensembles by over 2 BLEU. On the WMT 2014 English-to-French translation task, our model establishes a new single-model state-of-the-art BLEU score of 41.8 after training for 3.5 days on eight GPUs, a small fraction of the training costs of the best models from the literature. We show that the Transformer generalizes well to other tasks by applying it successfully to English constituency parsing both with large and limited training data.
51.8K
Greedy function approximation: A gradient boosting machine.
TL;DR: A general gradient descent boosting paradigm is developed for additive expansions based on any fitting criterion, and specific algorithms are presented for least-squares, least absolute deviation, and Huber-M loss functions for regression, and multiclass logistic likelihood for classification.
What is a support vector machine
TL;DR: Support vector machines are becoming popular in a wide variety of biological applications, but how do they work and what are their most promising applications in the life sciences?