This invention relates to a method of managing a system comprising a plurality of power sources and a plurality of power consumers, the method being configured for execution in a computing device, the computing device being assigned to the system, and to a system comprising a plurality of power sources and a plurality of power consumers, wherein such a managing method is applied to the system.

As systems grow in complexity with increasing numbers of consumers and sources, the management of such systems, for example keeping track of which consumers are supplied by which sources, becomes more difficult. Further, as the practice of remotely monitoring systems increases, keeping track of the physical location of consumers and sources becomes more difficult. Moreover, consumers and consumers may be continuously added, removed, temporarily disconnected, relocated, and reconnected. Thus, the locations of consumers and a topology of the connections between consumers and sources is a live, dynamic environment.

One example where topology information is useful is in data center applications. As the costs of energy and, in particular, electricity increases, having up-to-date information regarding power topologies and specifically which devices, like servers, storages, switches etc. are connected to which power sources, circuit branches or power points becomes increasingly important. This information enables data center managers to optimize power consumption via adequate planning, right-sizing of energy supply to data center equipment, power balancing, consumer shedding, overconsumer protection, efficient problem diagnosis, and so on. Additionally, in some data center applications, managers may be required to maintain up-to- date power topology information in order to monitor reliability and reduce troubleshooting time in case of failure.

Power Line Identification (PLI) is a method that analyses measurements of power consumption or computing activity of server and measurements of power output of power outlets. Such a method is known, for example, from US 2011/0307111 Al . The algorithm produces pairs of power sources and power sinks. In a data center with a very large number of servers and power outlets, applying such a method becomes an issue, as it requires a huge amount of time and huge amount of computing resources to directly treat a large number of servers and outlets, like, for example, 1000 of each. The resources and time needed is a nonlinear, approximately quadratic, function of the number of devices.

One objective of the invention is, to provide a method which allows retrieving the topology of a system with a large number of devices by matching of consumers and sources more efficiently and less intrusive for the system.

The objective is accomplished by a method and by a respective system according to the independent claims. In the dependent claims, advantageous embodiments are described.

The method of managing a system comprising a plurality of power sources and a plurality of power consumers, according to the invention, is configured for execution in a computing device, which is assigned to the system. The system, in the sense of the invention, is a composition of associated devices, which are either power consumers or power sources. Generally, the method, or parts thereof, may as well be executed manually. The plurality of power sources and consumers is meant to comprise at least two of each. Preferably, in a data center, as a realization of the system, the number of devices, which represent the sources and consumers, is considerably greater.

According to the invention, the sources are monitored for source events of changes in power consumption from the sources, wherein power consumption from the sources means that at least one of the consumers is supplied with electric energy by the source. Further, the consumers are monitored for consumer events of changes in availability of the consumers, wherein availability is meant in the sense of availability for at least one of the other consumers of the system. Furthermore, a temporal sequence of the source events and the consumer events is analyzed in order to detect temporal correlations between source events and consumer events.

The method according to the invention advantageously allows matching of power sources to power consumers, when the source events and consumer events of which reveal a temporal correlation. Finding such temporal correlations is advantageously simple and can be achieved by analyzing the sequence of events, which events are easily monitored without the necessity of sophisticated measurements of power consumption or detecting algorithms to discover details about activities of the power consuming devices.

When a pair comprising one source of the plurality of sources and one consumer of the plurality of consumers, for which a temporal correlation of the monitored source events and consumer events is detected, the pair is preferably linked together in an infrastructure map of the system, which is generally known as pairing of power sources and power consumers. Power supply links consist of pairs, even if one power consuming device is supplied by two or more power sources, which is actually standard for many devices. In that case, the consumer is partner in a plurality of pairs.

According to a preferred embodiment, a temporal correlation is given between a source event and consumer event, if the source event and the consumer event occur in a predetermined period of time, which period is particularly chosen short enough to avoid false correlations.

A further preferred requirement for a temporal correlation between a source event and consumer event, is, that the source event and the consumer event share the same direction of change. For example, a rising power consumption from a power sources correlates to a power consumer, which has become available in the system, approximately at the same time, or, if a consumer is switched off the respective power consumption will drop.

According to a further preferred embodiment, a source event is given, if the power consumption changes from no consumption to a detectable consumption or vice versa. Preferably, the monitoring of the power sources allows a resolution down to each power outlet, wherein one power outlet is only connected to one power consumer. This is favourable, as only two conditions have to be detected in that case, in particular, either current flows or current does not flow. Furthermore preferred, in this regard, is a detection of the power consumption using high-pass filtering, which avoids detection errors caused by intrinsic consumption. According to a furthermore preferred embodiment, the power consumption changes of the power sources are monitored using a power management tool of the system. In particular, data center management tools usually comprise a power management tool, which can advantageously be used in the context of the invention.

According to a furthermore preferred embodiment, a consumer event is given, if the availability of the respective power consumer changes from not accessible to accessible or vice versa. The availability of the power consumers is preferably monitored via at least one connection between the consumers and the computing device. In particular, the consumer is accessible if it is responding over the connection, and the consumer is not accessible if it is not responding over the connection. A power consuming device, for example in a data center, is thus available if the device can be accessed, for example over the data center network, i.e. if data can be read from a storage device of a server or if network data of the consumer can be retrieved by Internet Protocol (IP) operations, including User Datagram Protocol (UDP), Internet Control Message Protocol (ICMP) etc. It is an advantage of so-called host discovery or ping scanning operations, which are regularly part of network scanning tools, giving basic evidence about the availability of a remote consumer, whereas the function of the power consumer, or the data stored therein, need not be accessed. Thus, according to a furthermore preferred embodiment, the system comprises a network, wherein the connection is provided by the network. Alternatively or additionally, the connection is provided preferably by a point-to-point communication, like serial bus or USB connection.

Another object of the invention is a system comprising a plurality of power sources and a plurality of power consumers, wherein a managing method as described herein before is applied to the system, which managing method is adapted to update an infrastructure map of the system.

The invention is now further explicated with respect to the attached drawings. The information refers to the method according to the invention as well as to the system. The described embodiments are exemplary and do not limit the scope of the invention. In the Figures,

Figure 1 illustrates schematically an embodiment of a system according to the invention; Figure 2 illustrates an exemplary log of a temporal sequence of events recorded during an execution of the method according to the invention in a system according to Figure 1 ;

Figure 3 illustrates a detection of temporal correlations in another example of a temporal sequence of events recorded during an execution of the method according to the invention in a system according to Figure 1 ;

Figure 1 illustrates a schematic representation of an embodiment of a system 110 according to the invention. The system 110 includes n power consumers CI - Cn and m power sources SI - Sm. The consumers CI - Cn are fed from the sources SI - Sm and therefore each consumer is linked to at least one source. However, it is not clear, which consumers are linked to which sources, as indicated by the ambiguous cloud representing the undiscovered links 1. The system 110, such as in a data center, grows in complexity with increasing numbers of consumers and sources, and thus, keeping track of which consumers are linked to which sources becomes more difficult. According to the invention, a method of managing the system 110 comprising the plurality of power sources SI - Sm and the plurality of power consumers CI - Cn is configured for execution in a computing device 300, the computing device 300 being assigned to the system 110. In the depicted embodiment, the computing device 300 is operably connected to the consumers CI - Cn and to the sources SI - Sm, which means the computing device 300 is adapted to access, for example, a consumer data network 2, at least for monitoring purposes, and a source data connection 3 to a power supply system of the system 110, which is illustrated only schematically by the box encompassing the sources SI - Sm. The system 110 comprises one or a plurality of consumer data connections 2 from the computing device 300 to the consumers CI - Cn and further comprises the source data connection 3 from the computing device 300 to the power sources SI - Sm. The connections 2, 3 to the consumers CI - Cn and to the power supply system with the sources SI - Sm can be provided by a common network, for example, an Ethernet network, but may as well be independent of each other, for example, as direct point-to-point connections, like USB or serial bus connections or any combination. The computing device 300 may be one of the consumers CI - Cn, which is supplied by at least one of the sources SI - Sm, but not necessarily. The computing device 300 monitors the sources SI - Sm for source events of changes in a power consumption from the sources SI - Sm and the consumers CI - Cn for consumer events of changes in availability of the consumers CI - Cn. A temporal sequence of the source events and the consumer events, which is preferably logged, will be illustrated with respect to Figure 2.

In Figure 2, an exemplary log of a temporal sequence of the source events and the consumer events is shown, which, according to the invention, is recorded during an execution of the method in a system according to Figure 1. The monitoring logs for three exemplary power sources SI , S2, S3 and for three exemplary power consumers CI, C2, C3 are shown, wherein the logs are denoted with the respective references of the sources and consumers. The source events are changes in power consumption from the respective power source. The power consumption is shown on the ordinates axes PI, P2, P3 in discreet values Low and High. The consumer events are changes in availability of the respective consumer. The availability is shown on the ordinates axes Al, A2, A3 in discreet values Off (Low) and On (High). The monitored logs are recorded against time on the abscissas axes T. A power off event or high to low change 10 and a subsequent power on event or low to high change 11 of the power source S2 are exemplarily marked.

According to the invention, the temporal sequence is analyzed in order to detect temporal correlations between the source events and the consumer events. These temporal correlations are marked with the double-headed arrows 20. A requirement for the temporal correlations is a coincidence of a source event and a consumer event. A further requirement is, that both events share the same direction, i.e. from low to high or from high to low.

With respect to Figure 3, the detection of the temporal correlations is explicated in more detail. Another example of a temporal sequence of events recorded during an execution of the method according to the invention in a system according to Figure 1 is shown, wherein the logged sequences of the power sources SI, S2, S3 and of power consumer CI are illustrated in the same way as in Figure 2. The power off consumer event or high to low change 30 of the power consumer CI has been monitored, and by analysis of the three monitored power off source events or high to low changes 31, 32, 33 of the power sources SI, S2, S3, it is determined, which one of the source events, if any, is temporally correlated to the consumer event 30. The source events 31, 33 of the power sources SI and S3 show a too high latency with respect to the consumer event 30, which is indicated by two longer double-headed arrows 40, 41. The latency between the source event 32 of the power source S2 and the consumer event 30 is short enough to conclude, that a temporal correlation exists, which is indicated by the shorter double-headed arrows 42.

Reference Numerals

1 Cloud representing the undiscovered inks

2 Consumer data connection

3 Source data connection

110 System

300 Computing device

CI - Cn Power consumers

SI - Sm Power sources

P1, P2, P3 Axes of ordinates

Al , A2, A3 Axes of ordinates

T Axes of abscissae

10, 11 Source events

20 Arrows

30 Consumer event

31, 32, 33 Source events

40, 41, 42 Arrows