The Community for Technology Leaders
2018 IEEE International Conference on Services Computing (SCC) (2018)
San Francisco, CA, USA
Jul 2, 2018 to Jul 7, 2018
ISSN: 2474-2473
ISBN: 978-1-5386-7250-1
pp: 129-136
Previously, Amazon EC2 Spot prices were always driven by short-term trends in supply and demand, requiring consumers to have an in-depth understanding of Spot markets and the bidding process in order to make "intelligent" time-vs-money-vs-value trade-offs. However, with the newly announced streamlined access model for Spot instances, Amazon states that the Spot prices will adjust more gradually based on long-term trends instead of reacting to short-term fluctuations in demand and supply. Therefore, consumers are no longer required to understand Spot markets and bidding, and yet can save up to 90% off the On-Demand prices. In this paper, we study the pricing patterns before and after the introduction of the modified model using standard statistical approaches including econometric inequality indices (the Gini coefficient and the Theil index), logistic regression, a hybrid forecasting technique based on Naïve, and Principal Component Analysis. Our findings confirm the announcements made by Amazon including less frequent Spot price changes, disappearance of sudden spikes, and smooth (but not necessarily gradual) adjustments in the Spot prices. Rather surprisingly, with the introduction of the new model, the median Spot prices have risen in the majority of the Spot markets. In addition, even in the changed access model Spot price forecasting can still yield valuable insights into the evolution and structure of a given Spot market, although there may no longer be a need for sophisticated bidding strategies.
econometrics, power markets, pricing, principal component analysis, regression analysis, supply and demand

M. Baruwal Chhetri, M. Lumpe, Q. B. Vo and R. Kowalczyk, "To Bid or Not to Bid in Streamlined EC2 Spot Markets," 2018 IEEE International Conference on Services Computing (SCC), San Francisco, CA, USA, 2018, pp. 129-136.
302 ms
(Ver 3.3 (11022016))