Highly reliable data centers using managed PDUs

by Emerson Network Power on 10/8/15 9:09 AM

Ronny Mees | Emerson Network Power

24

Today’s most innovative data centers are generally equipped with managed PDUs since their switching capabilities improve reliability. However, simply installing managed PDUs is not enough – an “unmanaged” managed PDU will actually reduce reliability.

So how do managed PDUs work? These advanced units offer a series of configurations which – if properly implemented – improve the availability of important services. The main features are Software Over Temperature Protection (SWOTP) and Software Over Current Protection (SWOCP), which are well described in the blog post “Considerations for a Highly Available Intelligent Rack PDU”.

It is also well-known, that managed PDUs can support commissioning or repairing workflows in data centers. The combination of well designed workflows and managed PDUs pushes the operational reliability to a higher level.

In high performance data centers, using clusters, another important point comes into play: clusters are complex hierarchical structures  of server farms, which are able to run high performance virtual machines and fully automated workflows.

As described here or here, such clusters are managed by centralized software together with server hardware.

Over the last couple of years cluster solutions have been developed following strong and challenging availability goals, in order to avoid any situation, which make physical servers struggle within the cluster. However, there would still be the risk of applications and processes generating  faults and errors and screwing-up the complete cluster, unless there was an automated control process – the good news is: there is.

rack-blog-post-709x532

The process which controls those worst case scenarios is called fencing. Fencing automatically kicks out of the cluster any not working nodes or services in order to maintain the availability of the others.

Fencing has different levels, which are hopefully wisely managed. In a smooth scenario fencing will stop disturbing services, or re-organize storage access (Fibre channel switch fencing) to let the cluster proceed with its tasks.

Another power fencing option is also called “STONITH” (Shoot The Other Node In The Head) and allows the software to initiate an immediate shutdown (internal power fencing) of a node and/or a hard switch off (external power fencing).

The internal power fencing method uses IPMI and other service processer protocols, while the external power fencing uses any supported network protocol to switch of a PDU outlet.  It is recommended to use secured protocols only, such as SNMPv3. So managed PDUs as MPH2 or MPX do not only support a nice power balance, monitor power consumptions or support datacenter operations workflows – they also allow the fence software to react quickly for higher cluster reliability. So it’s not a secret that cluster solutions manufacturers – e.g. Red Hat with RHEL 6.7 and newer – openly support such managed rack PDUs.

For More Emerson Network Power Blogs, Click Here

Read More

Topics: Data Center, PUE, robust data center, Containment, efficient data center, DVL, electrical distribution, energy, Battery, Thermal Management, energy efficiency, 7x24, PDU

The Server Closet Must Die...

by Mike Rinaldi on 6/20/13 12:46 PM

They are small, poorly managed, wasteful and high risk environments for a business’s IT systems and they’re lurking in a building near you. The server closet, it’s a danger to us all and it must be eradicated.

The server closet is a room set aside for IT gear, but not designed as a data center. Not even as a bad data center. They vary in size, but are generally between 100 SF and 2500 SF. The rooms usually have some extra AC assigned to them and maybe additional jury rigged power, but beyond that they are a large closet that IT gear has been force fit into. There are 10s of thousands of these server closets out in the wild. 

 

 

Read More

Topics: Emerson Network Power, data center infrastructure, Data Center, data center design, data center infrastructure management, robust data center, Containment, efficient data center

Half of Data Centers Will Run Out of Power/Cooling Before Space

by Mike Rinaldi on 5/17/13 10:28 AM

 

data center users group

Read the full data center report at www.emersonnetworkpower/dcugreport

Read More

Topics: Emerson Network Power, data center infrastructure, reduce cost, Data Center, Green IT, data center design, data center energy, data center infrastructure management, robust data center, efficient data center, reduce downtime, Green Technology

Does DC Have a Chance in the Data Center?

by Mike Rinaldi on 5/14/13 7:38 AM

As data centers take an ever larger bite of worldwide energy production, efficiency becomes an increasingly urgent topic. Much of the focus in recent years has been on areas like virtualization, improved cooling practices (like free cooling and bumping up the thermostat), consolidation of idle equipment and so on—and rightly so. In the background, however, has been a growing discussion about a fundamental topic:alternating-current (AC) versus direct-current (DC) power. AC power won the day when electrical infrastructure was first being deployed, but could DC be poised for a return from the dead? 

[The Data Center Journal]

> Read More

 

plugs

Read More

Topics: data center infrastructure, reduce cost, data center design, robust data center, efficient data center, DVL, electrical distribution, data center outages, DC Power

How NFPA Electrical Safety Requirements Affect Power Users?

by Mike Rinaldi on 4/17/13 10:36 AM

The 2012 edition of NFPA 70E: Standard for Electrical Safety in the Workplace® became effective Aug. 31, 2011, and requires facility managers to take action in regards to safety-related work practices.

 

electrical safety

 


Read More

Topics: Emerson Network Power, data center infrastructure, reduce cost, Data Center, data center design, data center infrastructure management, robust data center, electrical distribution, reduce downtime, data center outages, DC Power, sccr, short circuit current ratings

Subscribe to Our Blog

Recent Posts

Posts by Tag

see all