Quantifying the potential of hot standby PSUs in ISP networks
This project is the continuation of Reducing power conversion losses in modern PSUs, in which we established that using more efficient, better dimensioned, or even single power supply units in network routers could yield sizable energy savings in ISP networks.
In this follow-up project, we will focus on the last lever. Specifically, we will study more in depth if and how we can apply hot standby mode to power supply units (PSUs), how effective it is, and how much savings this could result in if applied globally in the network of ETH’s Internet Service Provider. To do so, the project will combine two investigation directions:
Effectiveness and prevalence of hot standby mode in modern hardware
Setting redundant power supply units in hot standby mode is well known. This project will first seek to clarify:
- How to use or configure PSU hot standby
- How commonly it is supported by operating systems and PSUs
- How efficient it is; that is, are there power differences from running a secondary PSU in hot standby versus turned off
- How reliable it is; that is, do we experience transient disturbances when a hot standby PSU takes over.
Realities of modern PSU efficiency
Hot standby modes may result in energy savings because it would minimize the pwoer conversion losses in lowly loaded PSUs. To quantify this effect, we need a better understanding and modeling of the power conversion efficiency of modern PSU hardware. Thus, this project will explore:
- How to properly measure the efficiency of PSUs? What is the procedure required by the 80Plus standard?
- Is the power conversion efficiency measured by PSUs? How commonly? How accurately?
- How to validate the accurancy of the reported conversion efficiency with external measurements?
Equipped with this extended knowledge and models about PSU efficiencies and hot standby modes, the thesis will aim to estimate the potential energy savings from universally deploying hot standby mode in actual networks.
