In a very short span of time, interest has surged (ahem) in measuring the power consumption of many IT infrastructure...
devices. Widespread adoption of Voice over IP (VoIP) systems has made VoIP an area of particular interest.
VoIP systems are typically always-on devices, so differences in power consumption across various vendor solutions are magnified as the devices run 24/7, year round. The specific attributes of the telephony components deployed will affect power consumption. The presence or absence of specific phone models (differing displays) will have an impact on consumption. Notably, the presence or absence of options such as audio response, emergency services and/or high-availability configurations will also have a significant effect on overall power consumption.
Table of Contents: Testing VoIP system power consumption
Benchmarking VoIP System Power Consumption
- LAN/WAN communication infrastructure
- VoIP system scope and components
- System configuration
- System states
- Steady-state power draw
- Impact of power supply
- System measurements
- Power factor
The Tolly Group -- which has released a series of best-practice guides for network device testing -- has attempted to address this problem by producing a playbook for testing VoIP system power consumption. The test sets benchmarks and parameters for measuring VoIP system power consumption.LAN/WAN communication infrastructure
Almost all components of the VoIP system, wired and wireless, will connect to a switched LAN infrastructure as the basic communication pathway. Thus, the power consumption of the underlying switched infrastructure will certainly have an impact on system consumption. Most wired VoIP handsets are powered using the Power over Ethernet (PoE) that is provided by the LAN switch. While outside the direct scope of this document, it is important to note that the capabilities of the switch with respect to power delivery will have an impact on overall consumption.Power measurement
The biggest challenge is that benchmarking power consumption in general is an entirely new endeavor to equipment vendors and users alike. Where many benchmarking practices evolved over many years, power consumption tests, to date, have been ad hoc affairs. (This Common Test Plan is an attempt to remedy this situation.) Tests to date have been similar but not identical and results might not be comparable. This is a significant challenge. The tester will probably encounter new and unfamiliar terminology. The industry encountered this situation when networking specialists began to become involved with voice technology. Now, users will need to understand new terms and acronyms related to power delivery. Having a way to measure the various attributes of power consumption is, of course, a mandatory element of benchmarking. Fortunately, there is a wide range of measurement devices available.VoIP system scope and components
Once you have assured accurate measurements, the rest of your test will hinge on the configuration that you choose to measure.
At the macro level, one needs to understand the general characteristics of the environment to be modeled and benchmarked. In general terms, this can be referred to as, say, HQ, regional office, or branch office. Unfortunately, there are no generally accepted definitions of what each of these sites should consist of.
While it is typically safe to assume that all sites will be provided with unified messaging (i.e., voicemail), other functions like E911 or high-availability deployments may not be a part of the configuration for a small branch office.
Certainly, if any of these locations is used as a call center, additional components would need to be added to the configuration to support such services.
The type and number of handsets chosen for a given configuration will also affect your benchmarking results. More sophisticated handsets not only cost more but are typically outfitted with color displays that will consume more power. While an individual phone display may not consume much more power than a phone without a display, any difference can be magnified greatly when the phones number in the hundreds or thousands.
As noted elsewhere in this document, the system elements and detailed configuration will have a significant impact on results. It is necessary to be as specific as possible about all of the elements tested. Not only do specific model numbers need to be noted for handsets and system modules, but where services run on standard servers, it is important to note the key physical aspects of those servers (e.g., type and number of CPU cores) and, naturally, to measure those systems running the required software services.
All VoIP systems will connect to a LAN switch infrastructure -- wired phones and wired server components -- but that switch infrastructure may or may not be included in any particular benchmarking project.
While it is not common to find VoIP server functionality bundled into a LAN switch, it is common to find vendor organizations that can sell both to a customer under its own brand (e.g., Cisco, Avaya, Alcatel-Lucent). Such vendors are likely to include the underlying LAN infrastructure, especially if they have designed elements of that LAN switch, particularly PoE, that interacts with the VoIP system and further optimizes the system.
Vendors that sell only VoIP solutions, and thus are likely to be vendor neutral when it comes to LAN infrastructure, will typically not include the LAN infrastructure in the energy consumption measurements and subsequent calculations.
With remote office implementations, some vendors integrate telephony functions into routers that are used to establish data connectivity with the outside world via WAN ports and sometimes contain integrated LAN switches that can be used to support VoIP phones or ordinary LAN-connected workstations and/or servers. In such cases, it is often impossible to measure only the power of the telephony-related components.
The state or condition of the system is important to understand when evaluating power consumption. Back-end components like call servers are assumed to be always on. Telephone handsets, however, are typically measured both when idle and when off hook (i.e., someone is using the phone). Thus, it is important to note the state of the handset. Except with specialized applications like call centers, the steady state for most handsets is on hook. A handset dedicated to a single knowledge worker is likely to be on hook for those hours of each workday and weekend when the worker is not present in the office. Given that this downtime represents roughly 75% of the hours in a week and that it is not likely that a user is on the phone constantly while in the office, it is reasonable to use idle measurements in calculations. (According to VoIP system vendor Mitel, in typical business environments, a phone is idle as much as 96% of the time.) Experience to date has shown the power consumption of basic phones off hook to be about 10% to 15% more than the idle state.Steady-state power draw
When we use this term in the context of performance testing, we mean that the system is running in a state that it can maintain indefinitely. From a power consumption standpoint, this is not as clear cut.
With telephony systems, the primary variable for power consumption is related to whether a handset is being used in conversation or is on hook, idle and ready to receive or make calls.
Furthermore, one needs to determine whether steady-state power draw is based on a device with active but idle ports or a device processing traffic.
Regardless of the definition used, it is important to note that energy cost calculations based solely on steady-state power draw are not likely to be accurate because over the course of time the switch will not remain in a single state with respect to traffic load and other aspects of operation that can influence power consumption. Please see the section below on cost calculation for additional information.Impact of power supply
Testers note that for systems that provide multiple power-supply options, power supplies typically operate at their most efficient state when the load on them is between 50% and 90%. Therefore, power supply selection is important. For modular systems that are lightly loaded, selecting the highest power capacity supply for the system will result in inefficient operation of the power supply and higher power consumption than utilizing a lower capacity power supply operating closer to the mid-range of its capabilities.System measurements
A single test often provides multiple measurements -- effectively different ways of looking at the same data.Watts
When one measures power consumption, that measurement is made in watts. As the definition includes a unit of time (one joule of energy per second), the energy consumption is expressed simply as watts. Network specialists are used to understanding performance per second, but they need to remember that to discuss watts per second is fundamentally incorrect.Power factor
Perhaps as important as the absolute measure of power consumed is quantifying the efficiency with which the power is being used by the device under test (DUT). Power factor is just that measurement. According to Wikipedia, it is "the ratio of the real power flowing to the load to the apparent power."
Inefficient use of power means that the device consumes more energy than it is actually able to use, so long-term power consumption costs are higher than they need to be. Power factor (PF) is a number between zero and one -- with one representing maximum or 100% efficiency. Test tools, such as Watts Up's electricity meter, will calculate this value automatically. This measurement occurs only when dealing with AC power sources and can be disregarded for DC systems.
The apparent power consumed by a system is the product of the root-mean-scale (RMS) values of the voltage and current across the device, assuming that the waveforms are in phase. This value is used by power suppliers to assess the total energy used. The problem is that, more often than not, the voltage and current waveforms will not be in phase, owing to the complex series of networks within the device. For additional details, please see Wikipedia's power factor page.
To see specific test permutations and configurations for VoIP system power consumption testing, see the full report: The Tolly Group RPP 1082: VoIP System Power Consumption.