GB2550348A

GB2550348A - Smart sensor based hotspot detection system

Info

Publication number: GB2550348A
Application number: GB1608511.0A
Authority: GB
Inventors: Hasan Ali Saeed Almusaad Jaafar
Original assignee: Individual
Current assignee: Individual
Priority date: 2016-05-15
Filing date: 2016-05-15
Publication date: 2017-11-22
Also published as: GB201608511D0

Abstract

The system comprises sensors 1, 2, 3, 4, 5, 6 located at the server rack cabinets to measure environmental conditions such as temperature, airflow, and humidity. These sensors wireless transmit data to a receiving station, using means such as Wi-Fi or Bluetooth (RTM). The sensors maybe located on the back and front of the cabinet. The data collected maybe stored in databases and used to produce a graphical representation of the environment, it may also be used to set off an alarm if a hotspot develops. Various analysis may be performed on the data, such as correlation and regression analysis, which can be used to identify factors contributing to hotspot development and also produce additional data on the environment. The invention can improve the management of a data center by allowing for an increase in operating temperature and a reduction in the energy consumed by cooling infrastructure. It can also avail heat-related throttling and stability issues.

Description

Smart sensor based hotspot detection system

This invention relates to a smart sensor based hotspot detection system.

It is a common practice to operate IT datacentres at low temperatures, typically between 15°C and 18°C. The widespread notion is that IT appliances are designed to operate at such low temperatures for reliability reasons. This is, however, not true. The real reason for datacentres to operate at low temperatures is to account for hotspots. Hotspots are areas within a datacentre where temperature is considerably higher. Hotspots are dynamic and very difficult to detect, and to deal with. For instance, a server is typically overloaded during peak period, thus it may generate an excessive amount of heat. However, the same server might be idle during off-peak hours and may not generate as much heat.

The energy required to cool down a datacentre ranges between 30% and 50% of the total energy consumed to run the datacentre. Excessive heat can cause other problems and inefficiencies. Modern Central Processing Units (CPUs) tend to throttle to keep temperature below a certain threshold known as Thermal Design Power, which is typically between 85°C and 100°C. Throttling can impact performance considerably, especially when it is critically needed.

Many datacentres today continue to adopt inefficient cooling strategies due to the lack of intelligent hotspot detection solutions. A smart sensor based hotspot detection system presents a solution to this problem by utilising specialized smart sensors distributed throughout the datacentre to detect the presence of hotspots in real time. Doing so makes it possible to optimise the datacentre cooling more efficiently, thus minimizing or eliminating the development of hotspots.

The system can also be used to measure the efficiency of the cooling infrastructure and air distribution in similar environments. This can be accomplished by placing the smart sensors in areas of interest to measure cold air temperature, hot air temperature, airflow and humidity.

Once hotspots are minimized or eliminated, it becomes safe to raise the cold air temperature of the datacentre, resulting in substantial energy saving. For instance, running a datacentre comprising 280 servers at 25°C instead of 15°C can save as much as 600 kWh (Kilowatt per hour). This translates into saving the environment 300kg of Carbon Dioxide per hour, or about 2,600 Tons per year, depending on the energy source.

Figure 1 illustrates a fully populated standard server rack cabinet equipped with six smart sensors. Three sensors are fitted at the front of the rack and the remaining three are fitted at the back. Sensors are paired. For instance, sensors 1 and 2 comprise a single measurement unit.

Figure 2 maps sensor temperatures to colours ranging from 15°C to 45°C. The greenish colour gradient indicates a cold temperature, whereas the reddish colour gradient indicates a high temperature, aka potential hotspot.

The sensors measure temperature and airflow and feed the data using either Bluetooth or Wi-Fi technologies to a receiving workstation. Bluetooth consumes less energy but covers a smaller area.

Bluetooth is appropriate for small datacentres or for appliances within approximately 15 metres from the receiving station, Wi-Fi can be used for appliances that are farther away.

The sensors can be attached to rack cabinets. Alternatively, they can be integrated into the chassis of appliances such as servers and/or network switches.

The receiving station receives data-feed from all sensors and pairs the data-feed as indicated above. The difference in temperature between the cold air entering the rack cabinet or the appliance at the front and the hot air leaving the rack cabinet or the appliance at the back is to be calculated. It is used to set the degree of colour as indicated in Figure 2 in real time on a 3-Dimensional map. The frequency of update can be adjusted to suit the particular needs of any datacentre.

The system can be set to trigger an alarm when a hotspot exceeds a predefined threshold.

The data is also stored in a database for detailed examination. Example of such examination is correlation analysis against appliance utilization, outside temperature and humidity and so on. The analysis helps identify reasons contributing to hotspot development. Regression analysis can be applied to estimate the temperature in areas where no sensors are installed.

The application software to carry out the functions indicated above does not currently exist and will be developed separately. The same applies to the relational database and the algorithms to be used for the analysis.

The smart sensors are produced by third parties; however, the software can be configured to work with different sensors from different providers simultaneously.

Claims

1. A smart sensor based hotspot detection system comprising specialised sensors placed at the front and back of a rack cabinet or an equipment to measure temperature, airflow and humidity in realtime and sends the data wirelessly to a receiving station.

2. A smart sensor based hotspot detection system, according to claim 1, in which an application software reads the data of sensors and translates the data into a graphical representation and triggers an alarm if hotspots are developed.

3. A smart sensor based hotspot detection system, according to claim 1, in which the data of sensors is stored in a database.

4. A smart sensor based hotspot detection system, according to claim 3, in which a data mining application reads the data from the database and applies correlation and regression analysis to help identify factors contributing to hotspot development.