WEBVTT 00:00:00.040 --> 00:00:04.530 The heating ventilation and air conditioning as an element of design 00:00:04.540 --> 00:00:09.240 resilience. Research out of Toronto University illuminated that the 00:00:09.240 --> 00:00:13.610 world's data centers are estimated to consume power equivalent to about 00:00:13.610 --> 00:00:20.350 17 1000‑megawatt power plants, equaling more than 1% of the total 00:00:20.350 --> 00:00:22.090 world's electricity consumption. 00:00:22.100 --> 00:00:25.290 Now, that was a study that was done some 4 years ago, 00:00:25.300 --> 00:00:28.110 so it wouldn't be a shock to discover that that 00:00:28.110 --> 00:00:30.750 footprint has become heavier and heavier. 00:00:30.820 --> 00:00:34.880 Most of the cost of the data center is actually for cooling. 00:00:34.880 --> 00:00:37.780 That same study showed that the yearly cost of running the 00:00:37.780 --> 00:00:41.940 cooling infrastructure for a typical data center can reach 00:00:41.950 --> 00:00:45.080 anywhere from $4‑$18 million. 00:00:45.090 --> 00:00:48.630 Google states that their data centers use much less 00:00:48.640 --> 00:00:50.450 energy than the typical data center. 00:00:50.460 --> 00:00:56.070 They raise their temperature to 80 degrees Fahrenheit, use outside air for 00:00:56.070 --> 00:01:02.020 cooling, and build custom servers, and they have detailed performance data to 00:01:02.020 --> 00:01:07.050 help move the entire industry forward, showing that they get better efficacy 00:01:07.060 --> 00:01:11.210 from running at those temperatures anyway. The standard recommendation from 00:01:11.220 --> 00:01:14.550 ASHRAE, American Society of Heating, Refrigeration and Air‑Conditioning 00:01:14.550 --> 00:01:19.750 Engineers issued its first thermal guide for data centers way back in 2004. The 00:01:19.750 --> 00:01:26.660 original ASHRAE temperatures recommended this range of 18‑27 degrees Celsius, 00:01:26.660 --> 00:01:35.680 roughly translates to 68‑77 degrees Fahrenheit; also, a 5‑15 degrees Celsius dew 00:01:35.680 --> 00:01:36.090 point. 00:01:36.100 --> 00:01:38.740 Now, the dew point, if it is too low, 00:01:38.750 --> 00:01:43.230 obviously, can induce an environment full of static electricity, and if 00:01:43.230 --> 00:01:47.860 it's too high, can promote corrosion. Sixty percent should be the relative 00:01:47.860 --> 00:01:52.190 humidity that is the larger measure of moisture. 00:01:52.340 --> 00:01:56.550 There are two types of cooling, latent cooling, which is the ability for 00:01:56.550 --> 00:02:00.920 the air‑conditioning system to remove moisture, and sensible cooling, 00:02:00.920 --> 00:02:04.000 which is the ability of the air‑conditioning system to remove heat that 00:02:04.000 --> 00:02:06.170 can be measured by a thermometer. 00:02:06.180 --> 00:02:08.960 Those are the two types of cooling that take place. 00:02:08.970 --> 00:02:13.580 Imagine a very humid place where you live and a very dry, hot place 00:02:13.580 --> 00:02:17.630 where you live. Both of those would need a type of cooling to bring 00:02:17.630 --> 00:02:22.150 the environment down so that it is livable. Inside the data center, 00:02:22.150 --> 00:02:26.490 imagine that you had a bird's eye view and you're above the server 00:02:26.490 --> 00:02:27.660 racks looking down. 00:02:27.670 --> 00:02:32.760 Imagine then that you are arranging your servers in rows. So here we have a row 00:02:32.770 --> 00:02:37.770 of three servers. F stands for front, B stands for back. The next row of 00:02:37.770 --> 00:02:43.100 servers, we would actually put the back of those racks so that they are 00:02:43.100 --> 00:02:48.660 back‑to‑back with the rack before, and therefore, the front of that second rack 00:02:48.670 --> 00:02:50.880 with the third rack's front, and so on. 00:02:50.890 --> 00:02:55.690 Now, this would allow us to have server inlet temperatures, 00:02:55.700 --> 00:02:59.370 server outlet temperatures, in a more controlled fashion. 00:02:59.380 --> 00:03:03.990 Imagine if we actually created the environment so that this 00:03:04.000 --> 00:03:08.600 environment said that I'm going to make it so that I contain all the 00:03:08.600 --> 00:03:12.380 hot air. Containing all the hot air by means of some type of 00:03:12.380 --> 00:03:16.790 structure to prevent the hot air from leaving that aisle would be 00:03:16.790 --> 00:03:18.360 what's called hot aisle containment. 00:03:18.370 --> 00:03:20.960 Now imagine how the rest of the data center would feel. 00:03:20.970 --> 00:03:23.180 It would most likely feel pretty cool. 00:03:23.190 --> 00:03:26.740 But if you did something called cold aisle containment, 00:03:26.750 --> 00:03:31.340 where you now contained the air from flowing freely in the 00:03:31.340 --> 00:03:33.230 front‑to‑front configuration, 00:03:33.240 --> 00:03:38.710 then the rest of the data center may actually feel inhospitable to humans. 00:03:38.720 --> 00:03:42.020 It would really depend on what is it that you're 00:03:42.020 --> 00:03:44.760 trying to derive from the environment. 00:03:44.770 --> 00:03:48.700 What efficiencies are you bringing about? As stated before, many of 00:03:48.700 --> 00:03:52.600 the major cloud service providers have actually moved beyond what 00:03:52.610 --> 00:03:54.970 they would consider to be rudimentary, 00:03:54.980 --> 00:03:58.930 but these are what the standards have been for a number of years. 00:03:58.940 --> 00:04:03.790 Next, let's get into design resilience from an availability class perspective.