Energy consumption of Database Management - Florina Jonuzi.pdf

Energy consumption of Database Management - Florina Jonuzi - slides from meetup Green Software Development Manifesto, Jun6 2024

1. INTERNAL
Energy consumptionofDatabase Management
Systems
Florina Jonuzi

2. INTERNAL
Agenda
1.Relevanceoftheresearch
2.Research questions
3.Methodology
4.Results
5.Conclusionand outlook
people.code.commitment.

3. INTERNAL
Problem description
Currentchallenge
energyconsumptionin data
centresisincreasing
DBMSs are a core
component of data
centres
Lack of methods for
determining the energy
requirements of DBMS
Relevanceoftheresearchtopic

4. INTERNAL
Objectives
Measurement of energy consumption and energy efficiency of
different DBMS under different load scenarios
Decisionsupportforenergy-efficientdatabasesolutions
Promotion of development strategies for optimisation
Relevance of the research topic

5. INTERNAL
Research questions
Which methods and tools are suitable for measuring the energy requirements of database
systems?
Are there significant differences between relational database management systems and
non-relational database management systems in terms of their energy requirements?
How do different load scenarios influence the energy requirements of different database
systems?
Can the energy requirements of database systems be reduced or improved through
approaches such as indexing, caching and optimised queries?

6. INTERNAL
6
YCSB-Benchmark Extended Method
Figure 1: YCSB-Method
Figure 2: extendedMethod
Methodology

7. INTERNAL
Methodology – experimental setup
Figure 13: Experimental setup for measuring the YCSB (based on the measurement
setup at Trier University of Applied Sciences) [1]
Figure 14: trial

8. INTERNAL
Methodology – experimental design
Table 1: experimental design -YCSB
Note
Onlya datasetof100.000 was usedtomeasurethecomplexqueries

9. INTERNAL
Results – YCSB 100.000
Figure 5: YCSB 100.000 datasetMariaDB Figure 6: YCSB 100.000 datasetMongoDB

10. INTERNAL
Results – YCSB 1.000.000
Figure 7: YCSB 1.000.000 datasetMariaDB Figure 8: YCSB 1.000.000 datasetMongoDB

11. INTERNAL
Workload A, B, C, D in detail
Table 2: Workload A evaluation
Workload A: 50% read, 50% update Workload B: 95 % read, 5% update
Table 3: Workload B evaluation
Workload C: 100% read Workload D: 95% read, 5% create
Table 4: Workload C evaluation
Table 5: Workload D evaluation

12. INTERNAL
Workload E, F in detail
Table 6: Workload E evaluation
Workload E: 95% scan, 5% read Workload F: 50% read, 50% read-modify
Table 7: Workload F evaluation

13. INTERNAL
Results – complex queries
Figure 9: complexqueriesMariaDB
Figure 10: complexqueriesMongoDB

14. INTERNAL
Aggregation, Delete, Create, Read
Table 8: aggregationevaluation
Aggregation
Delete
Create Read
Table 9: deleteevaluation
Table 10: createevaluation Table 11: readevaluation

15. INTERNAL
Results – index optimization
Figure 11: Index optimizationMariaDB
Figure 12: Index optimizationMongoDB
Table 13: withoutIndex Table 14: withIndex

16. INTERNAL
Answering the research questions
1.Whichmethodsandtoolsaresuitableformeasuringtheenergy
requirementsofdatabasesystems?
→Various benchmark tools (YCSB, TPC, Sysbench, ...) that analyse the
performance of database systems under various load scenarios
→several methods to measure energy consumption
→different measurement methods to measure consumption (e.g.
using external measuring devices such as Shelly Plug, software-
based methods such as Powertop)
2. Are there significant differences between relational database
management systems and non-relational database management
systems in terms of their energy consumption?
→ Significant discrepancies between SQL and NoSQL

17. INTERNAL
Answering the research questions
3.How do different load scenarios (e.g. predominantly read access,
predominantly write access, etc.) influence the energy requirements of
different database systems?
→ Different load scenarios have different influences on energy
requirements for both database systems
4.Can the energy requirements of database systems be reduced or
improved through approaches such as indexing, caching and
optimised queries?
→ Energy efficiency is significantly promoted by optimizing queries
using an index
→ Analyses have shown that both MariaDB and MongoDB save
considerable energy with an index.

18. INTERNAL
Conclusion and outlook
•Energy consumption
varies
•MongoDB moreenergy
efficientwhenreading
& creatingdata
•MariaDBmoreenergy-
efficientforcomplex
queries
Findings
•Approach to
selecting
benchmark
•Shelly plugreal time
data
•Hardware factor:
HDD harddrive
•Noconsiderationof
multi-user access
•Selectionofa
benchmark
•Executionon a
different OS
•Use ofcooling
systemsorfans
Challenges Outlook