To make sure that the power system at a colocation facility is robust enough to handle both power and equipment failures, two words should be remembered: capacity and redundancy.
The primary factor that determines uptime for servers in a colocation facility is power. Power outages will knock a network offline and even damage hardware such as motherboards, memory, and hard drives. Despite how intrinsic power is to keeping businesses connected to their networks, only 2% to 3% of colocation facilities have the right power systems in place. The other 97% of facilities most commonly lack redundancy, multiple units carry the energy load even if one unit fails, or have units that are running above capacity, so a unit failure will cause the other units to overload and fail. Every part of the power system – uninterruptible power supplies (UPS), transfer switches or circuit breakers, generators, and power distribution units (PDU) – should be redundant and running below capacity.
Problem 1: Non-redundant Power Grids
Multiple PDUs connected to separate power grids and multiple UPSs should be designed into the colocation facility to offset a power grid failure. Colocation facilities with redundant power grids can connect customer servers to different grids at the same time, so that even if one goes offline, the other will work, keeping the network running without interruption.
Problem 2: Non-redundant UPSs
The UPSs supply power during an outage until the generator can come online; if the UPSs do not turn on immediately at the time of failure, then the network will go down. Even with high quality UPSs, failures are common, so it is critical for there to be multiple redundant UPS units in an “n+1” configuration – all of the necessary UPSs, plus an extra. Functionally, this means that each UPS runs sufficiently below capacity to handle a unit failure without the other units overloading. If there are two UPSs, then each unit must run below 50%, so that if one fails, the other can continue without overloading. If there are three units, each must run below 66%; four units, below 75%. The current load is shown on the display on the front of the UPS.
Problem 3: Transfer Switch Failures
Most colocation facilities use mechanical transfer switches, which are not as dependable as circuit breakers, to switch power from the electric utility to the generator. These switches are one of the most common places the power system fails. Without redundant switches to transfer power at the same point, a transfer switch failure will mean that a network goes down.
Problem 4: Insufficient Generator Capacity
Generators supply power during an outage. To run without overloading, the generator must have capacity to run 1.5 times the total building load. Ideally, a colocation facility should have a redundant backup generator in case the primary generator fails, and the facility should have a process in place for switching power between generators. Having multiple generators is not the same as having redundant generators. One of the most common generator problems with colocation facilities is that the facility started out with a small generator and added generators as it grew. This creates multiple points where power has to be transferred during an outage, increasing the likelihood that a network will go down. As a practical consideration, the generators must be well-maintained, tested monthly, and fully supplied with fuel.
Points to Consider
Fewer than one in twenty colocation facilities have the best power systems in place despite the fact that power systems have the most impact on network uptime. Without well-maintained and redundant components running below capacity at every part of the system, network performance as well as server performance and equipment lifetime will suffer. To make sure that the power system at a colocation facility is robust enough to handle power and equipment failures, two words should be remembered: capacity and redundancy.
Tampilkan postingan dengan label colocation service provider. Tampilkan semua postingan
Tampilkan postingan dengan label colocation service provider. Tampilkan semua postingan
Senin, 25 Juli 2011
Selasa, 21 Juni 2011
Colocation Facility Cooling System Considerations
It is important to look for a colocation facility that has Internet provider options and backup power systems. Equally important is a facility with the right design, chillers, and CRAC units to cool the server room.
Most companies know to find a colocation facility that has good Internet service and Internet provider and carrier options. However, cooling systems are often overlooked. Most servers now are made to operate at normal office conditions, which mean that two environmental factors should be met:
If the cooling system goes down, it can cause the equipment to overheat, causing the network to go down and even damaging equipment. Colocation facility cooling systems include three areas: the facility layout, chillers, and computer room air conditioning (CRAC) units.
Facility Layout
Servers generate heat as they run. This is generally blown out the back of the server by fans; cool air is taken in through vents at the front. Server racks and rows should be laid out to manage hot air coming from the servers and conditioned air going to their intakes by designating “hot rows,” which face the server fans, and “cold rows,” which face the intakes. Designated hot and cold rows keep air circulating in the best directions to keep servers from overheating.
Facility Design
There are two kinds of building design for colocation facilities: raised floor and solid floor. Raised floor was the preferred structure when servers were towers because the mesh flooring allowed air conditioned air to blow from the bottom directly across their intakes. However, since servers are stacked in racks, cold air must reach the tops of the racks to cool machines. Solid floor facilities are a better option. The facility should take into account the following:
Chillers pipe water or water/glycol coolants through the CRAC units. Chiller systems include pumps, pipes, and the chillers themselves. The following points should be considered to make sure that the system is adequate to cool the facility:
CRAC units are massive, dedicated air conditioners which manage temperature and humidity. They have many parts and are complicated – meaning failures can be common. The CRAC units should be well maintained, but the most important thing is redundancy. Redundancy is determined by the overall capacity of the CRAC units; there should be enough units that even with a failure, the facility will still be adequately cooled. As with chillers, capacity for CRAC units is determined by tons per square foot.
Verify Capacity
Capacity is determined by the ratio of tons per square foot. To determine whether a system has sufficient capacity, divide the total tonnage of the units by the square footage of the building. If there are four 30-ton CRAC units in a 4,000 square foot facility, then the CRAC capacity is .030 tons per square foot (120 tons / 4,000 square feet). The ideal ratio of tons to square feet for the CRAC system is .030 ton/square foot or higher; the ratio should not be below .025 tons/square foot. This same formula is used for chiller systems, with the same ideal ratios.
It is important to look for a colocation facility that has Internet provider options and backup power systems. Equally important is a facility with the right design, chillers, and CRAC units to cool the server room. Keeping equipment from overheating will keep networks up and makes equipment last longer. Like Internet connections, network uptime, and power backups, cooling systems signify a quality colocation facility.
Most companies know to find a colocation facility that has good Internet service and Internet provider and carrier options. However, cooling systems are often overlooked. Most servers now are made to operate at normal office conditions, which mean that two environmental factors should be met:
- The temperature should stay between 72° F and 76° F
- The humidity should be between 45% and 60%, with 45% ideal
- Cooling units must be on continuously, so servers will not overheat (this precludes standard HVAC units).
If the cooling system goes down, it can cause the equipment to overheat, causing the network to go down and even damaging equipment. Colocation facility cooling systems include three areas: the facility layout, chillers, and computer room air conditioning (CRAC) units.
Facility Layout
Servers generate heat as they run. This is generally blown out the back of the server by fans; cool air is taken in through vents at the front. Server racks and rows should be laid out to manage hot air coming from the servers and conditioned air going to their intakes by designating “hot rows,” which face the server fans, and “cold rows,” which face the intakes. Designated hot and cold rows keep air circulating in the best directions to keep servers from overheating.
Facility Design
There are two kinds of building design for colocation facilities: raised floor and solid floor. Raised floor was the preferred structure when servers were towers because the mesh flooring allowed air conditioned air to blow from the bottom directly across their intakes. However, since servers are stacked in racks, cold air must reach the tops of the racks to cool machines. Solid floor facilities are a better option. The facility should take into account the following:
- Solid floor designs must have mesh front and rear doors for th server cabinets to allow hot air to escape.
- For raised floor designs, cold air must have enough air pressure to force air to the top of the server cabinets. Ideally, special duct work will be in place to force the airflow upward.
- For raised floors, server cabinets should not have vents in the doors that would let cool air escape.
Chillers pipe water or water/glycol coolants through the CRAC units. Chiller systems include pumps, pipes, and the chillers themselves. The following points should be considered to make sure that the system is adequate to cool the facility:
- Sufficient capacity
- Backup units for the pumps and chillers which switch over automatically
- Regular maintenance
CRAC units are massive, dedicated air conditioners which manage temperature and humidity. They have many parts and are complicated – meaning failures can be common. The CRAC units should be well maintained, but the most important thing is redundancy. Redundancy is determined by the overall capacity of the CRAC units; there should be enough units that even with a failure, the facility will still be adequately cooled. As with chillers, capacity for CRAC units is determined by tons per square foot.
Verify Capacity
Capacity is determined by the ratio of tons per square foot. To determine whether a system has sufficient capacity, divide the total tonnage of the units by the square footage of the building. If there are four 30-ton CRAC units in a 4,000 square foot facility, then the CRAC capacity is .030 tons per square foot (120 tons / 4,000 square feet). The ideal ratio of tons to square feet for the CRAC system is .030 ton/square foot or higher; the ratio should not be below .025 tons/square foot. This same formula is used for chiller systems, with the same ideal ratios.
It is important to look for a colocation facility that has Internet provider options and backup power systems. Equally important is a facility with the right design, chillers, and CRAC units to cool the server room. Keeping equipment from overheating will keep networks up and makes equipment last longer. Like Internet connections, network uptime, and power backups, cooling systems signify a quality colocation facility.
