Feeding a steady stream of electrical juice into a data centre has never been a favourite task. Completing such a project typically involves turmoil with building-maintenance staff, outside electricians, and sales engineers. What starts out as a fairly straightforward process usually winds up requiring several weeks of extra time and a box of NoDoze to complete.
Recently, however, we had a chance to test a new breed of data centre uninterruptible power supply (UPS) solutions. American Power Conversion’s (APC) InfraStruXure attempts to address the loose ends of a typical data centre UPS installation. The result is a line of modular UPS systems that emphasises not only reliability and monitoring, but resource centralisation and noticeable cost savings as well.
APC shipped its mid-sized InfraStruXure set-up to the Advanced Network Computing Lab at the University of Hawaii along with a soon-to-be sunburned sales engineer — the same treatment given to every InfraStruXure customer.
InfraStruXure systems are designed for use in 10,000 to 40,000 volt-amp environments and are generally configured in N, N+1, or 2N installations. We dubbed our system “the Big Black Rack” because it consisted of three interlocked APC rack systems.
The unexpected space savings was clear from the start — systems similar to the one we tested are usually deployed in at least four racks with much of the power distribution functionality left to the building’s infrastructure rather than the UPS itself. This system dropped all that functionality into a svelte two-rack system.
The InfraStruXure’s power distribution unit (PDU) underscores the benefits of system consolidation. Power-distribution intelligence and a step-down transformer are usually located next to the building’s switch box (typically taking up three or more metres of data centre floor space). InfraStruXure, however, drops everything into a dedicated rack complete with central redundant feed capability, central shut-off, remote monitoring, and an intelligent distribution unit.
We found some pleasant surprises in the InfraStruXure design when we connected the system to some very power-hungry 10GbE switch chassis. Power is fed to all InfraStruXure devices with three-phase hot leads and a ground; no neutral is required because of the built-in ferro-resonant transformer, which also prevents a wide variety of transient spikes from entering the system. This adds serious value in terms of stability, especially coupled with InfraStruXure’s overall ease of use. InfraStruXure can also handle feeds at 208 volts, 480 volts, or 600 volts via its built-in step-down transformer, saving the cost of an external step-down transformer and allowing you to take power directly from the building buss bars.
UPS installation is often a delicate dance based on what a customer wants to spend versus what they think they will need in the future. Overbuilding is considered a best practice in this situation, but the trick is knowing when to quit. InfraStruXure’s extreme modularity takes care of the problem, with each rack acting as a series of interconnected power modules connected to a centrally managed power “backplane” for scalability.
Each power module is basically a dedicated 2U UPS corresponding to a separate battery set. All devices are hot-swappable, making configuration extremely flexible with regard to overall power consumption and UPS redundancy. We hot-swapped a number of power modules and batteries, a truly painless process due largely to InfraStruXure’s logical re-entry into the system. Instead of introducing sudden current to or from the hot-swapped device, the PDU regulates power to new devices, allowing them to gradually realise full capacity. All this is done automatically, so once an InfraStruXure system is installed, it’s entirely likely that your IT staff can manage the device by themselves without service contracts from resellers. That’s a positive impact on your bottom line.
Standard InfraStruXure systems use an Information Controller (IC) for management along with a dedicated 24-port Ethernet hub for communication both within the system and to outside monitoring stations. The IC is a rack-mount Web server dedicated to browsing management information, event logging, alerting, and similar monitoring functions.
Our test setup, however, came equipped with an optional InfraStruXure Manager. This is another 1U, server-based appliance that provides Web browser-based remote management of any APC device on the network via a single IP address. It can handle more proactive management tasks than an IC, such as addressing overload situations or issuing environmentally based alerts. With InfraStruXure Manager, IT administrators can do some PowerStruXure reconfiguration tasks via their Web browsers — unlike the IC, which requires that reconfigurations happen at the device’s serial console interface.
Despite this flexibility, the InfraStruXure Web management interface could use a boost in direct reconfiguration capabilities as well as security.
At log-in, we noticed that all user names and passwords are sent to and fro via clear text; without SSL encryption support, hacking into your data centre power management system becomes a simple task of using any network sniffer. APC has promised to correct this problem.
Minor quibbles aside, the APC InfraStruXure system lives up to its promises. We continue to be impressed with its day-to-day performance.
You’ll be impressed by its price tag, which is considerably less than comparable UPS solutions even though it provides more features, more flexibility, and a smaller footprint. APC’s solid performance and price lead us to give InfraStruXure a score of 10 for value — a score neither of us have ever awarded before. Check it out, and you’ll see why InfraStruXure gets high marks.
Our InfraStruXure test unit was an N+1 configuration consisting of a 208/120-volt PDU, three 10kW power modules with corresponding batteries, a nifty air-distribution unit at the bottom of the rack, and an environmental-monitoring unit, all locked up in two NetShelter VX 42U rack enclosures with a third NetShelter thrown in for our convenience.
Unfortunately, this system was simply too much of a behemoth for us to manage any sort of power-saturation testing. Instead, we concentrated on long-term usage and management-based testing. In four months with the system, we went through seven reconfigurations and variations on three power-loss scenarios. These scenarios revolve around the data centre losing all power, losing power from one of our two primary feeds, and losing power while one or more of the UPS batteries are in hot-swap mode.
In all cases, the InfraStruXure system performed flawlessly. This may be the best data centre UPS with which we’ve ever worked.