Achieving The Next Level in High Availability Systems

Increasing availability to five "9"s and beyond for computing requires several considerations beyond the computing hardware and software platforms. Although the actual computer hardware and software are critical components in achieving high availability, one of the most important areas that is often overlooked at the computer system level is a reliable and highly available power source. This requirement is either assumed, or it is requested that the end user ensure that a reliable power source is available 24/7. Because high availability computing is primarily found in the converging market where computer telephony and telecommunications have come together, the power infrastructure is typically robust and has been designed from both a high reliability and high availability perspective. It may seem redundant to mention both high reliability and high availability, however they are not mutually exclusive but mutually dependent.

You may have a very reliable system that seldom fails: however, when it does fail and takes your system down for hours, it can be devastating and significantly reduce overall availability to less than five "9"s. Conversely a highly available system that has redundant hot swap capability with no downtime during a failure does not mean it is a reliable system. A highly available system may not cause downtime when replacing hot swap redundant modules, however, if they are replaced continuously the chances that a failure of the redundant modules may occur during a replacement increases with each failure.

Merging the availability and reliability of the power system and the computing system is sometimes difficult because these systems are normally the responsibility of two different system engineers. IBUS has brought together a unique combination of both power and computing system engineers who combine their respective knowledge from their areas of expertise to define a new level of reliability and availability for the complete system.

Achieving five "9"s availability with any system is challenging enough, going beyond that requires an entirely different way of thinking. Developing an independent five "9"s computer system and powering it with an independently developed five "9"s power system will not necessarily accomplish the objective of going beyond that level of availability. An integrated understanding of both the power and computing systems will be required to achieve higher levels of availability.

A Variety of Power Sources
The convergence of computer telephony and telecommunications has created challenges for computing hardware, firmware and software. Much attention has been focused in these areas to ensure high availability fail over schemes and hot swap capabilities. Power is often an afterthought that someone else worries about and has been addressed by providing redundant power supplies within the computing system. This convergence has also created challenges for reliable available power for these computing systems and to achieve the next level of system availability, it requires going beyond supplying redundant power supplies and feeding them with a power source that someone else worries about.

The primary power source can be a highly available Dual DC source with a maintained and monitored DC power plant that will reliably provide backup power for 4 hours or more at a central office of a telecommunications service provider. The power source can be a Dual AC source comprised of two primary AC sources from a utility or a utility primary and a local AC source that can be a generator with a UPS bridge at a customer premise site. The UPS may be a centralised parallel redundant large UPS or a decentralised non-redundant small UPS. The power source can be a single feed AC source at a co-located ISP premise with or without a UPS. Each of these power sources provides different levels of reliability and availability.

Depending on the point of view one may be considered more reliable or available than the other. A centralised UPS advocate can provide data that shows a large UPS with redundant modules is more reliable and available than having a plethora of small decentralised UPS. While this may be true from a theoretical standpoint there are some considerations that may cause second-guessing. A centralised UPS is typically hundreds of feet from the actual critical load. While the wire, panel boards, PDUs, and circuit breakers that cover this distance are very reliable if they do fail, they typically are not a hot swap few minute repair. These interim power systems may or may not be redundant and if they are not, even if you continue power with a second source you have lost redundancy, and therefore availability. Additionally, the reliability of all of the loads can potentially be an issue if the failure of one can cause a fault clearing device (breaker, fuse) or protection circuit on the UPS to open to clear the fault. Also the repair and maintenance time required for large UPS is longer than smaller decentralised UPS. The system may be completely available during these maintenance times with redundancy if the service person doesn't make a mistake.

A decentralised UPS scheme gets the power protection closer to the critical load, removing some uncertainties. However, there would be more UPS, and the overall reliability when you add up all the systems may be less. Even if the reliability is less, the availability may be greater for several reasons. One would be redundancy at the lower level, which is a popular design today and is normally a much faster swap and repair than a larger UPS system. Also depending on the computer system scheme, a decentralised approach may only reduce availability to a smaller portion of the overall computer system that may be redundant as well. One of the drawbacks of decentralised UPS is the use of potential revenue generating rack space that could be filled with additional computing power.

Meeting The Challenge
In order to provide a next-generation high availability computer system, IBUS proposes to consider the power source availability as an integral part, and provide high availability choices for power without depending on external batteries in a box. In addition, it is important to get the protection as close to the source as possible without using valuable rack space.

There are several challenges to be overcome to service a market with a variety of power sources and a need for high availability, and there is not one single solution. In fact, five different solutions that can be provided in one computer server chassis (currently the nFUZION 8U systems) have been defined and are in development.

Two of the solutions feature dual cross feed redundant hot swap DC or AC modules. These modules fit in a 3U slot and directly interface with the IBUS cPCI back plane. They provide interconnection to a Dual DC source or a Dual AC source with cross feed redundancy to redundant DC or AC power supplies. These modules also provide fault isolation, loss detection and hot swap capability. This enables high reliability and availability at the computer power interface when there is a highly reliable and available power source.

The Dual AC can be used with a hot swap DC boost converter module and hot swap battery module that is contained in a special compartment that is in the airflow path of the IBUS patented hot swap multi-fan cooling module integrated into the 8U chassis. This multi-fan assembly provides the required cooling for both the high density SBCs and the batteries for an integrated UPS. This integrated UPS approach provides high available and reliable back-up close to the critical load without the use of valuable rack space.

When looking at a UPS on the motherboard IBUS's design may not be considered a UPS at all. Using Maxwell PowerCache solid-state energy storage technology in conjunction with a DC regulator at the bus voltage of the microprocessor enables back-up for a short period without batteries for the motherboard micro. This protection does not provide back-up to the entire computer system but does provide power back-up as close to the primary source as possible and at least provides a short-term back-up of the intelligence of a system without batteries.

The I-Bus approach to high availability system is not a traditional design. It is an integrated approach that takes advantage of understanding both the power and computing side of high availability and provides power protection options that can utilise external dual power sources or internal power back-up sources.