Browse Prior Art Database

Method to reprovision, reallocate, and rebalance IT infrastructure power to multiple servers in a building location.

IP.com Disclosure Number: IPCOM000032790D
Original Publication Date: 2004-Nov-12
Included in the Prior Art Database: 2004-Nov-12
Document File: 2 page(s) / 42K

Publishing Venue

IBM

Abstract

Disclosed is a method for hardware on each blade chassis to communicate wirelessly to other chassis the information of actual power usage of the blades that each chassis currently has installed. This way, a more accurate assessment of the actual power consumption of the entire IT room can be made. This assessment can then be compared to the existing power capacity of the building and therefore more blades can be installed than the maximum power rating per blade would allow. This disclosure also includes a method for prioritizing the servers within an IT infrastructure to determine which individual blade(s) should be throttled down first.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 52% of the total text.

Page 1 of 2

Method to reprovision, reallocate, and rebalance IT infrastructure power to multiple servers in a building location.

    With the increasing power needs for next generation servers, blade solutions in particular, the limiting factor for IT infrastructures is shifting from space constraints to total power consumption. Even now with current blade products, IT infrastructures can exceed building power thresholds while using only a relatively small amount of space. In the future, the demand for larger numbers of faster blades will continue to grow and these blades will emerge still denser and consume more power per blade than current models. This obviously will present major power issues for blade consumers hoping to take advantage of the space saving benefits of blade servers in existing buildings. Most buildings simply were not built to handle such a huge power load. Currently the only solution to this problem is to redesign the building's power infrastructure, an extremely unlikely option. A new infrastructure would incur far too great a cost for design and implementation, as well as interfering with business operations during the upgrade. These factors make a redesign an unfeasible option.

    The other problem with power limitations of IT infrastructures involves the efficiency of power allocation to each individual server in the room. Right now, each server is allocated the maximum power consumption for that particular model. However, it is highly unlikely that each individual server is making full use of the power allocated to it. Of course, any single server could be using all of its power at a given time, but most likely not all of the servers all of the time. Therefore, when an IT room is being fully populated, the maximum number of blades possible is determined by the maximum power consumption of each blade and the building's power capacity. However, the actual power consumption of that number of blades in a real-world environment will most likely never come close to overloading the building's total available power. The problem, therefore, is how to more efficiently maximize the number of blades being run in an IT room without overloading a building's power infrastructure.

    Each blade chassis will include a means for wireless communication in order to implement this idea. A group of chassis will combine to create a small local network separate from the main network specifically for this purpose. Through this wireless network, each chassis will keep track of the information provided by each other chassis, as seen in the Figure below. This information includes: the number of blades in the chassis, what type of blades they are (reveals the maximum power of that particular blade), and the actual power usage of each blade. The actual power usage will be determined over time by software. This software will ta...