3 Ağustos 2016 Çarşamba

Overload Protection in a Dual-Corded Data Center Environment

Executive summary:

In a dual-corded environment, the loss of power on onepath will cause the load to transfer to the other path,which can create an overload condition on that path.This can lead to a situation where the failure of one pathleads to the failure of both paths. This paper explainsthe problem and how to solve it, and provides a set ofrules to ensure that a dual-path environment providesthe expected fault tolerance.

Introduction:
The key purpose of a dual-corded or dual-path electrical architecture is to ensure continuity of IT operations during a failure of part of the power distribution system, by providing an alternative power path. By dual-corded, we mean IT devices that have two separate power inputs, and are   continue to operate when one cord loses power. By dual-path, we mean a power distribution system that supplies a dual-corded IT environment with two separate power paths, such that each IT device receives power from the two separate paths. The two paths may join at some upstream point, which might be at the distribution panel, the UPS output, the switchgear, or the utility mains connection. In some cases, data centers have been constructed where the two paths extend into the mains distribution system and are provided by separate substations or even separate high voltage lines. Most data centers that have a dual-path system extend it up to the facility switchgear, and a standby generator is used to achieve redundancy when there is only a single utility mains feed to the building.
When there is a failure in the distribution system or even in an IT device power supply, a system with dual-corded IT and a dual-path power distribution system is designed to maintain the operation of the IT load. While this is simple in concept, some rules and monitoring must be established to ensure the system operates correctly. This paper first explains how the IT devices behave in this environment, then explains the conditions that must be met to ensure that the expected availability is achieved, and finally provides strategies about how to manage a dual corded environment.
A correctly implemented and verified dual-path power system provides fault tolerance and allows for concurrent maintenance of any point in the power system. This is true even if there are no cross connects between the power paths, and even if one of the paths has no UPS. Many users implement dual-path architecture but do not trust it to work when needed, as evidenced by the use of static transfer switches and cross ties for maintenance. It is a common design practice to ensure power to both paths during many types of failures, even during maintenance. Yet this extra insurance should not be necessary if the dual-path system works correctly to begin with. If a dual-path system is correctly implemented and verified, the system will tolerate the loss of a path without incident, which allows for data center designs that are simpler and less costly.

Behavior of the dual-corded IT device:
A dual-corded IT device is assumed to be able to operate correctly when powered from either cord. However, this assumption about the behavior of the devices is not always correct. In a dual-corded IT environment, it is first necessary to establish if the devices truly meet the dual-corded assumption. Over 95% of all dual-corded IT devices typically encountered wil correctly operate from a single cord, but this is not 100%. The reasons that there are some devices that do not correctly operate according to the dual-corded assumption include:
• The device specifically implemented dual cords not for redundancy but rather as a means to get more power to the IT device by using multiple power supplies. There are a number of reasons why an IT device might use multiple supplies instead of one larger supply, which include: A) the device was designed to be expandable over time including the ability to add more power, B) the device designers did not want to require a special large power plug, and by using two supplies with conventional plugs they avoided this issue.
• The device has three power cords, and requires two of them to operate correctly. There is no way to plug three cords into two power paths so that the device will survivethe failure of either path (it might survive the failure of the path with one cord attached, but would not survive the failure of the path with the two cords plugged in).
• The device has implemented dual cords for redundancy under normal configurations, but under full internal IT device configuration the power load is greater than a single supply can provide, so that the dual-cord assumption is only good for lightly configured devices. While this seems to be a design flaw, it has occurred in some networking equipment where some newer plug-in cards did not exist yet at the time of the chassis design. Most vendors in this situation later issued updated power supplies of higher rating, but the burden is on the user to ensure that the configuration meets the dualcorded assumption.
• The device was designed as a dual-corded device, but one power supply has failed in service and that condition has not been noticed or corrected. The device is now behaving as a single-cord device and will drop when the power is lost to that remaining path.
• The device is a dual-corded device but both cords have inadvertently been plugged into the same power path. The device will act normally but will drop when the path feding the two cords drops. This is a common occurrence, particularly in data centers where there are many different people with rights to access and change equipment.
• The device is not a dual-corded device but is a single-corded device, and has been deployed in a dual-path environment. If dual-corded behavior is needed for this device, it can be achieved by installing a small rack-mount transfer switch for one or a few devices, or, if the single-corded load is large, dual-corded behavior can be obtained by installing a large static switch feeding a special third path to the single-corded racks or zone. For more information see, Çağlar Arlı Research, Powering Single-Corded Equipment in a Dual Path Environment.

Hiç yorum yok:

Yorum Gönder