Icebergs in the Clouds: the Other Risks of Cloud Computing

TL;DR

This paper highlights the hidden and complex risks of cloud computing, including security vulnerabilities, system instabilities, failure correlations, and preservation challenges, urging proactive study before widespread reliance occurs.

Contribution

It identifies and discusses underrecognized risks of cloud computing, emphasizing the need for further research to understand and mitigate these potential threats.

Findings

Cloud services share deep resource dependencies leading to catastrophic failure risks.

Unpredictable interactions in resource pooling can cause system meltdowns.

Digital preservation of cloud artifacts is significantly more difficult than traditional data archiving.

Abstract

Cloud computing is appealing from management and efficiency perspectives, but brings risks both known and unknown. Well-known and hotly-debated information security risks, due to software vulnerabilities, insider attacks, and side-channels for example, may be only the "tip of the iceberg." As diverse, independently developed cloud services share ever more fluidly and aggressively multiplexed hardware resource pools, unpredictable interactions between load-balancing and other reactive mechanisms could lead to dynamic instabilities or "meltdowns." Non-transparent layering structures, where alternative cloud services may appear independent but share deep, hidden resource dependencies, may create unexpected and potentially catastrophic failure correlations, reminiscent of financial industry crashes. Finally, cloud computing exacerbates already-difficult digital preservation challenges, because only the provider of a cloud-based application or service can archive a "live," functional copy of a cloud artifact and its data for long-term cultural preservation. This paper explores these largely unrecognized risks, making the case that we should study them before our socioeconomic fabric becomes inextricably dependent on a convenient but potentially unstable computing model.