<\/span><\/h2>\n\n\n\nThe Y-axis on this chart indicates the difference between the top and bottom temperature points.\u00a0 I to refer to this as the delta-T.\u00a0 Mathematically it is: top temperature \u2013 bottom temperature<\/p>\n\n\n\n
The X-Axis lists each of the 30 racks (labeled 1 through 30).<\/p>\n\n\n\n
The bubbles indicate the delta-T at each rack. There are multiple bubbles for each rack because the delta-T changes over the course of the 7-day measurement period.<\/p>\n\n\n\n
The size of the bubble indicates the length of time each rack maintained that particular delta-T. Larger bubbles indicate more time and smaller bubbles indicate less.<\/p>\n\n\n\n
The colors of the bubbles are displayed differently side by side for visual acuity only.<\/p>\n\n\n\n
<\/figure>\n\n\n\n<\/span>Part3:How do I read this chart?<\/span><\/h2>\n\n\n\nStart with Rack 1 which displays the dark blue bubbles to the far left.\u00a0 During the 7-day date range, Rack 1 had a delta-T (top temperature \u2013 bottom temperature) that fluctuated between 4\u00b0F and 6\u00b0F.\u00a0 The majority of the time the delta-T for Rack 1 was 4\u00b0F, indicated by the larger-sized bubble.<\/p>\n\n\n\n
Now look at Rack 9. The delta-T for Rack 9 fluctuated between 4\u00b0F and 14\u00b0F, but it spent most of the time at 12\u00b0F.<\/p>\n\n\n\n
Rack 12 spent the entire time (i.e. all 7 days) time at -2\u00b0F.<\/p>\n\n\n\n
<\/span>Part4:What does this data tell us?<\/span><\/h2>\n\n\n\nIn most data centers, we expect the temperature at the top of the rack to be hotter than the temperature at the bottom of the rack. This is simply because hot air rises.\u00a0 This will even be the case in an overhead air distribution environment (but perhaps not as much so as an underfloor distribution).\u00a0 However, since we are comparing the data to the rest of the temperatures in the facility, a marker of good analytics, this analysis will provide the same insights in any type of environment.<\/p>\n\n\n\n
This particular facility had underfloor air distribution. We would expect the delta-T to be somewhere between 0\u00b0F \u2013 8\u00b0F.\u00a0 Said another way, we wouldn\u2019t expect the top of the rack to be more than 8\u00b0F hotter than the bottom of the rack.\u00a0 And certainly, we wouldn\u2019t expect the bottom to be hotter than the top, which would be indicated by negative values.<\/p>\n\n\n\n
The green region indicates the expected range of delta-Ts. The red region indicates delta-Ts outside of this expected range.<\/p>\n\n\n\n
<\/figure>\n\n\n\nIt\u2019s clear that over the course of the 7-day period there were some delta-Ts significantly outside of the expected range.<\/p>\n\n\n\n
The beauty of this analysis is that not only are those aberrations clear, but it also becomes simple to identify the specific culprits by highlighting which racks have big bubbles in the red region.<\/p>\n\n\n\n
<\/figure>\n\n\n\nThe racks highlighted in gray on the X-Axis have delta-Ts that lie outside our expected range.<\/p>\n\n\n\n
<\/span>Part5:Let\u2019s start taking action on the data<\/span><\/h2>\n\n\n\nWhy are the delta-Ts for these racks outside our expected range?<\/p>\n\n\n\n
There could be a number of reasons, however it\u2019s usually an indication of mixing. Some reasons might include:<\/p>\n\n\n\n