Temperature levels within racks, server room or datacentre environment are one of the most important environmental factors to regularly monitor. Sudden room temperature rises can indicate the failure of a critical infrastructure system such as an air conditioner and lead to server failures and worst case a fire starting. Hot-spots within server racks can literally ‘cook’ the electronics and batteries within in-rack UPS systems. Low temperatures can be a sign that the cooling system operating parameters are set too low, leading to excess energy usage and higher operating costs.
Server room environmental monitoring systems range in sophistication from simple data loggers to device that can be accessed via an internet connection. Essentially the system comprises of an environment monitor, a sensor (which may be built-in or plugged-in) and a monitoring software package. The principle features of an environmental monitoring system include:
In addition to monitoring and analysis, events can be also be used to trigger automatic actions using digital outputs (DO) to initiate actuators or other devices.
For small IT closets and server installations, the local IT manager may rely on temperature sensors built-into installed systems i.e. uninterruptible power supplies (UPS), air conditioners, servers and storage devices. Internal temperature and sometimes fan-failure monitoring may be built-into these devices with optional plug-in sensors for assemblies including external UPS battery packs.
Each of these critical items will likely generate an alarm if they see an internal temperature above a pre-set threshold. Temperature-related alarms can indicate the failure of a cooling fan in a UPS system or a pump failure within an air conditioning system. The alarm may be audible, visual, reported remotely using SNMP over an IP network or relay signal contact.
Other infrastructure systems including rack-level power distribution units (PDUs) may also offer additional monitoring sensor points. Some intelligent PDUs can be installed with plug-in temperature and access control sensors. The information they collate can include temperature readings within a server rack and ON/OFF (open or closed) states for rack doors. The information and state changes are reported to the manufacturer specific monitoring software package over an IP network or via SNMP to third-party platforms.
The main weakness in this approach to temperature monitoring is that it does not provide a single ‘pain of glass’ view but a disparate one that may be split across several monitoring programs. Whilst the use of a data centre infrastructure management (DCIM) package could overcome this, for most server room and small-to-medium sized datacentre operators, the cost of DCIM implementation can be prohibitive.
When it comes to IT closets, computer and server rooms and small-to-medium sized datacentres with a smaller number of server racks, a dedicated server room monitoring, and control system is the most widely adopted solution. These types of system centre around a combination of hardware-based environment monitors, plug-in sensors, detectors and converters and a master-station type monitoring package that may be hosted on a local server or Cloud-based facility.
There can be several drivers for monitoring environments. Within the food and pharmaceutical manufacturing industries, there may be legislative or insurance requirements that can only be met by a comprehensive environmental monitoring system.
In the IT industry, from small IT closets to server rooms and datacentres, the drivers may be more operational. The higher the power demand of a server, the greater the need to monitor ambient temperature and heat loads placed onto local cooling systems. When temperatures rise quickly, especially within a server rack, reliability can be compromised and there is a potential for thermal runaway in the rack, especially those with higher power demands (15-30kW). Energy efficiency and electricity usage are other metrics affected by room temperatures.
There are three ways to record and access environment-related data:
A basic data logger may only be capable of monitoring one or possible two environmental aspects i.e. temperature or temperature and humidity. Automatic and cloud-based environment monitors may be able to monitor a wider range of digital inputs (DI) and analog readings using additional plug-in sensors and detectors.
An environmental monitoring system consists out of three main elements: a base unit or hub, probes or sensors, and network connectivity and integration. The hubs may contain one or more built-in sensors, as well as ports for hooking up external probes. They also include an Ethernet port and have software for remote configuration and graphing. This software may also work with existing network management software, such as SNMP systems.
For complete server room monitoring and control a dedicated environment monitoring system should be installed. This will comprise an environment monitor, along with a set of sensors (temperature only, temperature and humidity, water and leaks etc, detectors (smoke and security), converters (to monitor generator fuel tank levels or air flow) and monitoring software.
When selecting a server room monitoring system one of the first things to identify is what needs to be monitored and where. It is important to consider both the room and rack levels and especially as the IT facility increases in size from a single rack to several racks or rows.
The recommended temperature for a server room is 18-27degC with a relative humidity (RH) of 40-60% and Dew Point temperature at 60% RH of 5.5degC. These are targets for the complete room but due to layout and equipment configuration there could be ‘hot-zones’ into which temperature sensors should be placed. ‘Hot-zones’ can be identified using a thermal imaging camera.
More information on temperature and humidity level recommendations in a server room:
Individual air conditioning units in the server room may or may not be monitored over a local IP/Ethernet connection, mobile application or building management system. Temperature sensors should be placed near to local air conditioners to get ‘failure’ detection as early as possible and provide an additional layer of security. This can be more important where there are multiple AC units installed, as the running units compensate for a failed unit.
Humidity monitoring is also important. If the relative humidity is too high, there can be a build-up of static electricity. Too humid and condensation can build-up leading to corrosion and the potential for a short-circuit condition.
In larger facilities it may be necessary to monitor the temperature and humidity levels within and outside hot/cold aisle containment areas.
ASHRAE is often referred to for guidance on server room temperature and humidity levels. The organisation recommends from six sensors per rack at the top, middle and bottom of a rack in pairs for the front and rear of the sections. Front and rear monitoring is important as cool air is drawn into the rack from 18degC and expelled to the rear at no less than 20degC.
The principle reason for rack level temperature monitoring is the power drawn by typical racks which can quickly lead to a ‘hot-spot’ and especially if a local air conditioner fails. High internal rack temperatures place greater workloads on server fan assemblies and can lead to CPU level data processing errors and potentially heat-induced stress failures.
The sensors do not have to use fixed wires. Wireless sensors can be an option for larger server room and datacentre facilities and are cheaper to install and reconfigure. The wireless approach also reduces the need for IP connections on the local network.
Other sensors to consider at a room level include smoke/fire, water/leakage, airflow/pressure and access control including motion sensing IP cameras.
Temperature and humidity monitoring will continue to grow in importance to organisations reliant on their on-premise service rooms and internet connectivity. The use of a server room environmental management system not only helps with capacity planning and energy efficiency. It can also help an IT or facility manager to respond to a sudden temperature rise and prevent a catastrophic system failure. In addition, climate change is also placing an extra cooling burden on many installed air conditioning systems and real-time monitoring can identify when to bring in additional cooling or upgrade an existing air conditioning system.
Rising cooling requirements are a common problem faced by many organisations as they adopt the latest server technologies, including high power computing. The higher the power draw of the latest servers the greater the need for air conditioning as their kW power draw adds an equal demand to the size of the air conditioning required. Edge Computing and micro data centres add another layer of complexity to the need for a dedicated remote environmental monitoring system.
Many computer room and server closets start out with a single server and quickly grow as more capacity is added. New servers and routers are typically added to an existing server racks or new rack cabinets brought into the facility to extend the capabilities. This can also happen within existing server rooms and datacentres. Sometimes the extra capacity was designed in from day-one or is engineered into the space available as part of an additional requirement rather than as part of a refurbishment program.