The benefits of data center temperature monitoring
Ş. Albert, M.R.C. Truşcă, M.L. Soran
National Institute for Research and Development of Isotopic and Molecular Technologies, 67-103 Donat, 400293
Cluj-Napoca, Romania; e-mail: [email protected]
INCDTIM
INCDTIM Datacenter • Grid site RO-14-ITIM
– 400 CPU Core, Storage 100 TB, Technology 1U servers + Blade system (IBM & HP)
• HPC Cluster (IBM iDataPlex unit) – 512 CPU Core (16 Core, 96 GB RAM, 450 GB
HDD / Computing unit), Storage 15 TB • Network infrastructure of the Institute
– Applications servers (e-mail, web, ftp, databases)
– Cisco Core switching system
2
INCDTIM
Network capabilities
• Layer 3 Core Switch - Cisco 6509E • 10 Gbps link to RoEduNet starting from 1 Feb. 2011 • Nexus switching system
– 40 Gbps inside the Grid site – 20 Gbps between Grid Site and Core Switch
3
INCDTIM
Why temperature monitoring?
the temperature is an important parameter for the equipment functioning.
the computer systems are designed to work efficiently when the ambient temperature is in the range 20 – 23°C.
constant temperature air conditioning systems power consumption
to minimize energy consumption
INCDTIM
Practical implementation
The followed parameters:
• voltage power supply of the cooling units;
• temperature in different points of data center;
• humidity in different points of data center;
• state of functioning of the air conditioning unit fans.
INCDTIM
1. digital power meter type UPT-210 (built by Algodue Elettronica, Italy) that provide information about power line parameters.
Hardware components INCDTIM
Electrical values measured are: The system voltage and the phase
voltages Phase currents, and current absorbed by
the system Power factor, Frequency Active, Reactive and the Apparent power
2. microcontroller-based module (designed and built in INCDTIM) that measured temperatures, humidity and monitor the state of the AC units using different type of sensors.
INCDTIM
The temperature sensors, LM35 (National Semiconductor), allow to measure temperatures in the range -40 ... +150°C with 0.5°C accuracy without the need for calibration.
The humidity sensors, SYH-2R, allow determining the humidity in the range 10 ... 95% with an acceptable accuracy (3 to 5%) for usual temperatures (0-70°C)
Electronic schema for adaptation board and sensors
INCDTIM
NEW: Humidity sensors
Block diagram of the monitoring system
Graphic interface for cooling system, temperature and humidity monitoring
Software components INCDTIM
Both software applications have a graphic interface and they are designed to: - read the watched parameters values and display them
online; - register all the values over the time; - generates an alarm when values are out of established
range; - implements 2 mechanisms: trigger an alarm by
authorized users and send an abort command.
INCDTIM
R1 – R2 – network distribution racks
R3 – R8 – computing servers racks
R4 – HPC rack
R7 – GRID rack
- position of the temperature sensors
- position of the humidity sensors
- ventilation grids
Schema of Data Center INCDTIM
Variation of different parameters
0
10
20
30
40
50
60
9:18
:07
AM9:
54:2
3 AM
10:2
6:05
AM
11:0
2:50
AM
11:3
1:11
AM
11:5
9:11
AM
12:2
4:35
PM
12:5
5:28
PM
1:24
:16
PM1:
55:5
8 PM
2:23
:31
PM2:
51:5
6 PM
3:17
:46
PM3:
42:4
8 PM
4:10
:48
PM4:
51:1
4 PM
5:30
:02
PM6:
27:3
7 PM
8:43
:39
PM10
:27:
58 P
M1:
31:1
8 AM
3:22
:57
AM5:
11:2
7 AM
7:03
:41
AM7:
58:2
2 AM
8:28
:24
AM8:
57:1
3 AM
9:27
:07
AM9:
54:4
0 AM
10:2
3:05
AM
10:5
0:36
AM
11:2
6:28
AM
11:5
4:00
AM
12:2
4:28
PM
12:5
3:15
PM
1:27
:03
PM1:
57:3
1 PM
2:33
:52
PM3:
18:2
8 PM
3:51
:37
PM4:
23:4
5 PM
5:23
:50
PM7:
38:3
7 PM
9:05
:24
PM10
:50:
08 P
M1:
02:0
0 AM
3:32
:13
AM5:
22:4
7 AM
6:22
:52
AM7:
20:2
9 AM
8:00
:31
AM8:
29:4
3 AM
8:57
:41
AM9:
24:4
8 AM
9:50
:15
AM10
:20:
43 A
M10
:51:
36 A
M11
:24:
59 A
M12
:02:
57 P
M12
:43:
00 P
M1:
14:1
8 PM
T(0C)
Hour (hh:mm:ss)
12-14.02.2014
T1
T2
T3
T4
0
5
10
15
20
25
30
35
40
45
9:17
:24
AM10
:10:
45 A
M11
:04:
07 A
M11
:57:
28 A
M12
:50:
50 P
M1:
44:1
1 PM
2:37
:32
PM3:
30:5
4 PM
4:24
:15
PM5:
17:3
7 PM
6:10
:58
PM7:
04:1
9 PM
7:57
:41
PM8:
51:0
2 PM
9:44
:24
PM10
:37:
45 P
M11
:31:
06 P
M12
:24:
28 A
M1:
17:4
9 AM
2:11
:11
AM3:
03:3
1 AM
3:56
:53
AM4:
50:1
4 AM
5:43
:35
AM6:
36:5
7 AM
7:30
:18
AM8:
23:4
0 AM
9:20
:01
AM10
:13:
22 A
M11
:06:
43 A
M12
:00:
05 P
M12
:53:
26 P
M1:
46:4
8 PM
2:40
:09
PM3:
33:3
0 PM
4:26
:52
PM5:
20:1
3 PM
6:13
:35
PM7:
06:5
6 PM
8:00
:18
PM8:
53:3
9 PM
9:47
:00
PM10
:40:
22 P
M11
:33:
43 P
M12
:27:
05 A
M1:
20:2
6 AM
2:13
:47
AM3:
07:0
9 AM
4:00
:30
AM4:
53:5
2 AM
5:47
:13
AM6:
40:3
5 AM
7:33
:56
AM8:
27:1
7 AM
9:20
:39
AM10
:14:
00 A
M11
:07:
22 A
M12
:00:
43 P
M12
:54:
04 P
M
P(kW)
Hour (hh:mm:ss)
12-14.02.2014
P
P1
P2
P3
a)
b)
a) registered temperatures in the fixed points; b) power consumption in the same period
T1 - rack site GRID T2 - HPC cluster rack T3 - temperature of the heat agent of AC unit T4 - outside ambient temperature
12
Test periods for reducing energy consumption
Simulation of data center operating using experimental data registered
Hour 07.02.2015 12.02.2015 14.02.2015
tDC(°C) text (°C)
no. of AC units tDC(°C) text
(°C) no. of
AC units tDC(°C) text (°C)
no. of AC units
1 - 4 21,35 9 3 21,37 8 3 20,31 9 1 5 21,24 7 3 21,22 7 3 20,42 9 1 6 21,21 6 3 21,20 7 3 20,45 9 1 7 21,18 6 3 21,17 7 3 20,47 8 1 8 21,17 7 3 21,15 8 3 20,42 9 1 9 21,15 8 3 21,14 9 3 20,53 9 1
10 21,13 9 3 21,15 10 3 20,57 9,5 1 11 21,37 11 3 21,13 11 3 20,61 9,5 1 12 21,09 12 3 21,11 15 3 20,64 9,5 2 13 21,08 14 3 21,02 20 3 20,64 10 2 14 21,07 14 3 21,17 21 3 20,65 12 2
15-18 21,07 14 3 21,12 23 3 20,68 13 2 19 20,98 15 2 21,02 16 3 20,69 9 2
20-21 20,91 10 2 20,99 15 2 20,71 9 2 22-23 20,89 9 2 21,00 12 3 20,73 9 2
Conclusions INCDTIM
• The device allows to register some parameters in different points of Datacenter and monitoring of these parameters can be achieved in real-time.
• The variation of the parameters values can be followed by data interpretation on different period of time and season.
• Advantages of the temperature monitoring: - simulation of the Datacenter operation based on recorded data for
long periods of time; - setting a schedule for operation of the air conditioning system; - maintaining Datacenter environmental conditions in the range in
which computer systems work efficiently; - reducing energy consumption;
• It can be determined and implemented a system of functioning of the entire system, in order to improve resource utilization.
The authors would like to express appreciation for the support of the Romanian National Authority for Scientific Research, CNCS – UEFISCDI, through the projects 8EU/2012 and IUCN-DUBNA.
ACKNOWLEDGEMENTS
Thank you for your attention !
INCDTIM