One of the biggest ‘loose-ends’ when Enterprises design, deploy and manage power distribution within a datacenter is the last mile component: the “Power Strip” or the ‘iPDU’. The power strip is actually a placeholder bit of terminology that usually refers to all of the power distribution hardware units which are installed WITHIN the rack itself, which typically take a single or dual power FEED from the building, and then splits them into individual outlets usable by each of IT devices housed inside the rack. Most common racks have positions for up to 42 or 47 1U tall devices, so in many cases the power strip requirement is handled by several hardware power appliances which yield the appropriate number of individual outlets. Remembering that in a redundant scenario, the number of outlets is doubled to 84 or 94, this can create a fairly large requirement for power outlets.

These hardware power distribution devices are available in horizontal or vertical form-factors, with the horizontal units offering from 8 to 16 outlets, and the vertical styles offering 21 to 24 outlets. These form-factors are available in a wide range of capabilities, from non-intelligent versions which are fundamentally sheet-metal, outlets and wire all the way up to exotic power management units which have embedded processors (typically a small Linux computer) running power monitoring and/or control capabilities. It boils down to cost to produce which translates to manufacturer’s list price. The lowest price at perhaps $25 per outlet is the ‘dumb’ strip, the mid-range at about $50-$75 per outlet and is referred to the ‘monitored’ strip and at the high end is the ‘monitored and switched’ type which commands over $100 per outlet!

In years past, it was very common for datacenter architects to specify higher end power strips with the ability to both monitor and power-cycle each device using these ‘monitored and switched’ power strips. In many cases, this switching feature was used as the preferred method for remediation when a Windows Server encountered a Blue-Screen condition (a.k.a. BSOD) and was hence standard best-practice for configurations in most ‘Windows Shops’.

Many shops have this style of Monitored and Switched power strip today. Curiously, in many Enterprises, even though fully featured Monitored and Switched power strips were chosen and deployed, the datacenter operators quickly embraced the switching features, but rarely took advantage of the monitoring features. Power was cheap and un-interesting from an ongoing monitoring standpoint.

With the maturity and reliability of both the server hardware manufacturers’ platforms and the robustness of the server operating systems, a shift has occurred over the last several years which significantly limits the needs or concern with ‘power cycling’, and instead focuses on the monitoring capabilities of the given power distribution solutions. Switched power strips are currently considered mostly overkill and too expensive for modern operating environments which are designed for a high degree of fault-tolerance and redundancy. Monitoring, NOT switched control is key!

Today, all customers are trying to establish baselines across their power-chains, with monitoring in real-time being the highest priority. It is quite rare today for datacenter managers to look to power-cycling as a strategic remediation tool, nor do they rely on power-cycling as a best practice. As such software management solutions for power-chains are less often required to control the switched availability of power at each individual outlet, in favor of highly granular monitored power across all points the power-chain.