diff --git a/use-case-submissions/dow-msft-Graph-DB.md b/use-case-submissions/dow-msft-Graph-DB.md index 83c1a83..dad2b31 100644 --- a/use-case-submissions/dow-msft-Graph-DB.md +++ b/use-case-submissions/dow-msft-Graph-DB.md @@ -37,16 +37,19 @@ For this case study, we have considered R to be per API call with 32,000 calls o Energy was used by measuring CPU and memory utilization and using a model of Thermal Design Power (TDP) of the processors, memory and storage. The equation used to model the energy consumption is: -P[kwH] = (Power consumed by CPU or Pc Number of cores + Power consumed by Memory or Pr + Power consumed by GPU or Pg Number of GPUs)/1000 - • CPU Utilization doesn’t scale linearly with power consumption, we will use the power curve as described in the [article](https://medium.com/teads-engineering/building-an-aws-ec2-carbon-emissions-dataset-3f0fd76c98ac) - • TDP of server used in Azure is [Intel Xeon Platinum 8380 Processor](https://www.intel.com/content/www/us/en/products/sku/212287/intel-xeon-platinum-8380-processor-60m-cache-2-30-ghz/specifications.html?wapkw=intel%20xeon%20platinum%208370c) for Graph DBs and [Intel Xeon Platinum 8270 Processor](https://www.intel.com/content/www/us/en/products/sku/192482/intel-xeon-platinum-8270-processor-35-75m-cache-2-70-ghz/specifications.html) for the APIs - • [From this article](https://www.crucial.com/support/articles-faq-memory/how-much-power-does-memory-use), it shows that RAM consumes 0.38kwh/GB. So a 32G of RAM allocated for each VM, would equate to 32 * 0.38 = 12.16 kwh + +`P[kwH] = (Power consumed by CPU or Pc Number of cores + Power consumed by Memory or Pr + Power consumed by GPU or Pg Number of GPUs)/1000` + +• CPU Utilization doesn’t scale linearly with power consumption, we will use the power curve as described in the [article](https://medium.com/teads-engineering/building-an-aws-ec2-carbon-emissions-dataset-3f0fd76c98ac) +• TDP of server used in Azure is [Intel Xeon Platinum 8380 Processor](https://www.intel.com/content/www/us/en/products/sku/212287/intel-xeon-platinum-8380-processor-60m-cache-2-30-ghz/specifications.html?wapkw=intel%20xeon%20platinum%208370c) for Graph DBs and [Intel Xeon Platinum 8270 Processor](https://www.intel.com/content/www/us/en/products/sku/192482/intel-xeon-platinum-8270-processor-35-75m-cache-2-70-ghz/specifications.html) for the APIs +• [From this article](https://www.crucial.com/support/articles-faq-memory/how-much-power-does-memory-use), it shows that RAM consumes 0.38kwh/GB. So a 32G of RAM allocated for each VM, would equate to 32 * 0.38 = 12.16 kwh #### Intensity was used by utilizing the Carbon Aware SDK and retrieving current emissions for each Azure region. #### Embodied Carbon (M) -The equation to calculate M = TE * (TR/EL) * (RR/TR) +The equation to calculate M: `M = TE * (TR/EL) * (RR/TR)` + Where: • TE = Total Embodied Emissions, the sum of LCA emissions for all hardware components associated with the application server. • TR = Time Reserved, the length of time the hardware is reserved for use by the software. @@ -57,20 +60,30 @@ Where: ### (Quantify) SCI Value Calculation #### TigerGraph DB + Memory Allocation = 32G vCPU Allocation = 8 CPU Utilization = 17.12 Memory Utilization = 6.2 -E-CPU = ((270*.4) ) x 1) = 108 w / 1000 = 0.108 kwh (TDP * coefficient, converted to kwh) -E-Mem = ((0.38 * 6.2) x 1) = 2.36 w / 1000 = 0.054 kwh (power consumption per gig * gig memory utilized, converted to kwh) -Total E = 0.110 kwh 0.108 + .0.054 +`E-CPU = ((270*.4) x 1) = 108 w / 1000 = 0.108 kwh` (TDP * coefficient, converted to kwh) + +`E-Mem = ((0.38 * 6.2) x 1) = 2.36 w / 1000 = 0.054 kwh` (power consumption per gig * gig memory utilized, converted to kwh) + +`Total E = 0.110 kwh 0.108 + .0.054` -I - 554g (East US & North Central) - Collected on 10/27/2022 +I = 554g (East US & North Central) - Collected on 10/27/2022 -M=1533.12 * (1/35040) * (2/64)= 0.0014kg, ~1.4 gCO2e ((total embodied coefficient * (1 hour/3 years of use) * (instance memory / platform Memory) ) -M=1533.12 * (1/35040) * (1/8) = 0.0054kg, 5.47gCO2e ((Total Embodied Coefficient * (1 hour/3 years of use) * (instance CPU/ platform CPU) ) -Total M = 6.87 1.4 + 5.47 +`M = 1533.12 * (1/35040) * (2/64)= 0.0014kg, ~1.4 gCO2e` + +(total embodied coefficient * (1 hour/3 years of use) * (instance memory / platform Memory) ) + +`M=1533.12 * (1/35040) * (1/8) = 0.0054kg, 5.47gCO2e` + +((Total Embodied Coefficient * (1 hour/4 years of use) * (instance CPU/ platform CPU) ) +Note that 4 years of use = 35040 hours (4 * 365 * 24 = 35040) + +`Total M = 6.87 1.4 + 5.47` > SCI Total for TigerGraph DB = (0.110 kwh * 554 gCO2e/KwH) + 6.87 gCO2e = 67.81 kgCO2e @@ -80,20 +93,25 @@ Memory allocation = 1.75 CPU Utilization = 25.0 Memory Utilization = 70% -1.75G-1.22 = 0.53 (Memory allocated - Memory used) = what is available - -E-CPU = (205*.4) ) x 1 = 82 w / 1000 = 0.082 kwh (converted number to kilowatts (TDP * coefficient) -E-Mem = (.38 * 1.22) x 1 = 0.4636 w / 1000 = .00046 kwh (power consumption * memory utilized) -Total E = 0.0824 kwh 0.082 + .00046 +`1.75G-1.22 = 0.53` (Memory allocated - Memory used) = what is available + +`E-CPU = (205*.4) ) x 1 = 82 w / 1000 = 0.082 kwh` (converted number to kilowatts (TDP * coefficient)) + +`E-Mem = (.38 * 1.22) x 1 = 0.4636 w / 1000 = .00046 kwh` (power consumption * memory utilized) + +`Total E = 0.082 + .00046 = 0.0824 kwh` I - 554g (East US & North Central) - Collected on 10/27/2022 -M=1216.62 * (1/35040) * (2/64)= 0.0012kg, 1.12 gCO2e ((total embodied coefficient * (1 hour/3 years of use) * (instance memory / platform memory) ) -M=1216.62 * (1/35040) * (1/8) = 0.0043 ~4.34 gCO2e ((total embodied coefficient * (1 hour/3 years of use) * (Instance CPU/ platform CPU) ) -Total M =5.46 gCO2e CPU + Memory 1.12 + 4.34 +`M = 1216.62 * (1/35040) * (2/64)= 0.0012kg, 1.12 gCO2e` +(total embodied coefficient * (1 hour/4 years of use) * (instance memory / platform memory)) -> SCI Total TigerGraph API - (0.0824 kwh * 554 gCO2e/KwH) + 5.46 gCO2e = 51.10 gCO2e +`M=1216.62 * (1/35040) * (1/8) = 0.0043 ~4.34 gCO2e` +((total embodied coefficient * (1 hour/4 years of use) * (Instance CPU/ platform CPU) ) + +`Total M = 1.12 + 4.34 = 5.46 gCO2e` +> SCI Total TigerGraph API - (0.0824 kwh * 554 gCO2e/KwH) + 5.46 gCO2e = 51.10 gCO2e ### Total Tiger DB + API = 118.91 (67.81 + 51.10) @@ -106,16 +124,19 @@ vCPU Utilized = 28.05 Memory Utilized= 32G - 25.63 (Min Available Mem) = 6.3 -E-CPU = (270*.5) x 1) = 135 w / 1000 = 0.135 kwh (converted number to kilowatts (TDP * Coefficient) -E-Mem = (.38 * 6.3) x 1 = 2.40 w / 1000 =0.024kwh (power consumption * memory utilized) -Total E=0.159 kwh 0.135 +0.024 +`E-CPU = (270*.5) x 1) = 135 w / 1000 = 0.135 kwh` (converted number to kilowatts (TDP * Coefficient) +`E-Mem = (.38 * 6.3) x 1 = 2.40 w / 1000 =0.024kwh` (power consumption * memory utilized) +`Total E=0.159 kwh 0.135 +0.024` I - 554g (East US & North Central) -M=1533.12 * (1/35040) * (2/64)= Memory 0.0014, ~1.40 gCO2e ((total embodied coefficient * (1 hour/3 years of use) * (instance memory / platform memory) ) -M=1533.12 * (1/35040) * (1/8) = CPU 0.0054kg, ~5.47 gCO2e ((total embodied coefficient * (1 hour/3 years of use) * (Instance CPU/ platform CPU) ) +`M = 1533.12 * (1/35040) * (2/64)= Memory 0.0014, ~1.40 gCO2e` +(total embodied coefficient * (1 hour/4 years of use) * (instance memory / platform memory) ) -Total M 6.87 gCO2e CPU + memory 1.12 + 4.34 +`M = 1533.12 * (1/35040) * (1/8) = CPU 0.0054kg, ~5.47 gCO2e` +((total embodied coefficient * (1 hour/4 years of use) * (Instance CPU/ platform CPU) ) + +`Total M = 1.12 + 4.34 = 6.87 gCO2e` > SCI = Total Neo4J DB = (0.159 kwh * 554 gCO2e/kwh) + 6.87 gCO2e = 94.95 gCO2e @@ -128,15 +149,19 @@ Memory Utilization = 65% 1.75 * 0.38 = 0.665 watts (Memory allocated * kwh Consumed) 1.75G-1.13 = 0.62 (Memory allocated - Memory used) = what is available -E-CPU = (205*.8) )x 1 = 164 w / 1000 = 0.164 kwh (converted number to kilowatts (TDP * coefficient) -E-Mem = (0.38* 1.13) x 1= 43 kwh / 1000 = 1000 .00043 kwh (power consumption * memory utilized) -Total E = 0.1644 kgCO2e/kwh 0.164 + .00043 +`E-CPU = (205*.8) )x 1 = 164 w / 1000 = 0.164 kwh` (converted number to kilowatts (TDP * coefficient) +`E-Mem = (0.38* 1.13) x 1= 43 kwh / 1000 = 1000 .00043 kwh` (power consumption * memory utilized) +`Total E = 0.164 + .00043 = 0.1644 kgCO2e/kwh` I - 554 gCO2e (East US & North Central) -M=1216.62 * (1/35040) * (2/64)= Memory 0.0011kg, ~1.12g gCO2e ((total embodied coefficient * (1 hour/3 years of use) * (Instance CPU/ platform CPU) ) -M=1216.62 * (1/35040) * (1/8) = CPU 0.0043kg ~4.34 gCO2e ((total embodied coefficient * (1 hour/3 years of use) * (Instance CPU/ platform CPU) ) -Total M 5.46 gCO2e CPU + memory +`M = 1216.62 * (1/35040) * (2/64)= Memory 0.0011kg, ~1.12g gCO2e` +(total embodied coefficient * (1 hour/4 years of use) * (Instance CPU/ platform CPU) ) + +`M = 1216.62 * (1/35040) * (1/8) = CPU 0.0043kg ~4.34 gCO2e` +(total embodied coefficient * (1 hour/4 years of use) * (Instance CPU/ platform CPU) ) + +`Total M = 5.46 gCO2e` > SCI Total Neo4J API = (0.1644 kwh * 554 gCO2e/kwh) + 5.46 gCO2e = 96.5 gCO2e