Cloud platforms remain underutilized despite multiple proposals to improve their utilization (e.g., disaggregation, harvesting, and oversubscription). Our characterization of the resource utilization of virtual machines (VMs) in Azure reveals that, while CPU is the main underutilized resource, we need to provide a solution to manage all resources holistically. We also observe that many VMs exhibit complementary temporal patterns, which can be leveraged to improve the oversubscription of underutilized resources.

To mitigate climate change, we must reduce carbon emissions from hyperscale cloud computing. We find that cloud compute servers cause the majority of emissions in a general-purpose cloud. Thus, we motivate designing carbon-efficient compute server SKUs, or GreenSKUs, using recently-available low-carbon server components. To this end, we design and build three GreenSKUs using low-carbon components, such as energy-efficient CPUs, reused old DRAM via CXL, and reused old SSDs.We detail several challenges that limit GreenSKUs, carbon savings at scale and may prevent their adoption by cloud providers. To address these challenges, we develop a novel methodology and associated framework, GSF (GreenSKU Framework), that enables a cloud provider to systematically evaluate a GreenSKU’s carbon savings at scale. We implement GSF within Microsoft Azure’s production constraints to evaluate our three GreenSKUs’ carbon savings. Using GSF, we show that our most carbon-efficient GreenSKU reduces emissions per core by $28 \%$ compared to currently-deployed cloud servers. When designing GreenSKUs to meet applications’ performance requirements, we reduce emissions by $15 \%$. When incorporating overall data center overheads, our GreenSKU reduces Azure’s net cloud emissions by $8 \%$.

If a trader could predict price changes in the stock market better than other traders, she would make a fortune. Similarly in the electricity market, a trader that could predict changes in the electricity load, and thus electricity prices, would be able to make large profits. Predicting price changes in the electricity market better than other market participants is hard, but in this paper, we show that attackers can manipulate the electricity prices in small but predictable ways, giving them a competitive advantage in the market.

Historically, robotics systems have not been built with an emphasis on security. Their main purpose has been to complete a specific objective, such as to deliver the correct dosage of a drug to a patient, perform a swarm algorithm, or safely and autonomously drive humans from point A to point B. As more and more robotic systems become remotely accessible through networks, such as the Internet, they are more vulnerable to various attackers than ever before. To investigate remote attacks on networked robotic systems we have leveraged HoneyPhy, a physics-aware honeypot framework, to create the HoneyBot. The HoneyBot is the first software hybrid interaction honeypot specifically designed for networked robotic systems. By simulating unsafe actions and physically performing safe actions on the HoneyBot we seek to fool attackers into believing their exploits are successful, while logging all the communication to be used for attacker attribution and threat model creation. In this paper, we present the HoneyBot and discuss our proof of concept implementation. Our HoneyBot prototype swaps between physical actuation and using prebuilt models of sensor behavior for simulation at runtime given user input commands.