The document describes a configuration optimization tool that aims to automatically optimize the configuration of big data technologies. It does this by running experiments on data intensive applications, measuring performance under different configurations, and using this data to recommend optimal configurations. The tool implements two approaches for optimization - Bayesian optimization and transfer learning. It consists of several components, including an experimental suite to run tests, an optimization module, interfaces to various big data technologies, and a performance repository to store results. The goal is to help users like SMEs reduce the time and cost of testing and configuring big data applications between releases.

1. DICE
Horizon 2020 Research & Innovation Action
Grant Agreement no. 644869
http://www.dice-h2020.eu
Funded by the Horizon 2020
Framework Programme of the European Union
Configuration Optimization
Tool
Pooyan Jamshidi
Imperial College London

2. Big Data Technologies
Cloud (Priv/Pub)
`
DICE Framework
2©DICE
DICE IDE
Profile
Plugins
Sim Ver Opt
DPIM
DTSM
DDSM TOSCAMethodology
Deploy Config Test
M
o
n
Anomaly
Trace
Iter. Enh.
Data Intensive Application (DIA)
Cont.Int. Fault Inj.
WP4
WP3
WP2
WP5
WP1 WP6 - Demonstrators

3. Configuration Optimization Tool
The problem:
o Big Data technologies have
100s of tuneable parameters
o A knowledge gap is faced by
SMEs in configuring these
technologies
What does the tool do?
o Automatically runs
experiments on DIAs
o Returns recommended
configuration parameters for
Big Data technologies
3©DICE
Innovation:
o Automate DIA configuration
across release cycles
o Prior art focuses on manual
configuration
Impact & stakeholders:
o Reduce cost and time of
testing between releases
o Support operators of DIAs

4. CO Tool Architecture
4©DICE
Configuration
Optimisation Tool
performance
repository
Monitoring
Deployment Service
Data Preparation
configuration
parameters
values
configuration
parameters
values
Experimental Suite
Testbed
Doc
Data Broker
Tester
experiment time
polling interval
previous versions
configuration
parameters
GP model
Kafka
Vn
V1 V2
System Under Test
historical
data
Workload
Generator
Technology Interface
Storm
Cassandra
Spark
Two impementations of CO: BO4CO, TL4CO

5. Tool Input (Parameters and Options)
5©DICE
1- Information
about the
experiment: budget,
config file, duration
of each experiment
2- Information
about the
configuration
parameters and
their options that
testers determine

6. CO Tool Architecture
6©DICE
Configuration
Optimisation Tool
performance
repository
Monitoring
Deployment Service
Data Preparation
configuration
parameters
values
configuration
parameters
values
Experimental Suite
Testbed
Doc
Data Broker
Tester
experiment time
polling interval
previous versions
configuration
parameters
GP model
Kafka
Vn
V1 V2
System Under Test
historical
data
Workload
Generator
Technology Interface
Storm
Cassandra
Spark

7. Optimization Component (Matlab)
7©DICE
- This component select
the next configuration to
experiment considering
the current
measurements,
- This continues until
optimum configuration
located or experimental
budget finished.
The optimization overhead
is negligable comparing
with measurements
This componly relies on
rayality free MCR
component

8. CO Tool Architecture
8©DICE
Configuration
Optimisation Tool
performance
repository
Monitoring
Deployment Service
Data Preparation
configuration
parameters
values
configuration
parameters
values
Experimental Suite
Testbed
Doc
Data Broker
Tester
experiment time
polling interval
previous versions
configuration
parameters
GP model
Kafka
Vn
V1 V2
System Under Test
historical
data
Workload
Generator
Technology Interface
Storm
Cassandra
Spark

9. Experimental Suite
9©DICE
This component runs the experiments and measures the
performance of the system under test, the data are flushed to
csv file and communicated with the optimization component

10. CO Tool Architecture
10©DICE
Configuration
Optimisation Tool
performance
repository
Monitoring
Deployment Service
Data Preparation
configuration
parameters
values
configuration
parameters
values
Experimental Suite
Testbed
Doc
Data Broker
Tester
experiment time
polling interval
previous versions
configuration
parameters
GP model
Kafka
Vn
V1 V2
System Under Test
historical
data
Workload
Generator
Technology Interface
Storm
Cassandra
Spark

11. Performance Repository
11©DICE
spouts max_spout sorters emit_freq chunk_size message_size throughput latency
1 10 1 1 1.00E+05 1000 22657 3.9833
1 10 1 1 1.00E+05 10000 3596.3 18.415
1 10 1 1 1.00E+05 1.00E+05 112.56 217.63
1 10 1 1 1.00E+06 1000 12273 5.1952
1 10 1 1 1.00E+06 10000 1174.9 24.247
1 10 1 1 1.00E+06 1.00E+05 111.88 205.49
1 10 1 1 2.00E+06 1000 12024 5.2935
1 10 1 1 2.00E+06 10000 1151.3 25.039
1 10 1 1 2.00E+06 1.00E+05 94.294 220.62
1 10 1 1 1.00E+07 1000 11552 6.2867
1 10 1 1 1.00E+07 10000 1228.1 24.975
1 10 1 1 1.00E+07 1.00E+05 102.29 236.19
1 10 1 10 1.00E+05 1000 25978 3.4782
1 10 1 10 1.00E+05 10000 10112 9.2847
1 10 1 10 1.00E+05 1.00E+05 1023.8 83.236
1 10 1 10 1.00E+06 1000 24147 3.6594
1 10 1 10 1.00E+06 10000 8400.2 11.804
1 10 1 10 1.00E+06 1.00E+05 1197.4 73.786
1 10 1 10 2.00E+06 1000 22858 3.7151
1 10 1 10 2.00E+06 10000 7141.3 10.755
1 10 1 10 2.00E+06 1.00E+05 1095.1 78.624
1 10 1 10 1.00E+07 1000 22693 4.3637
1 10 1 10 1.00E+07 10000 6281.5 14.308
1 10 1 10 1.00E+07 1.00E+05 951.27 71.492
1 10 1 60 1.00E+05 1000 25862 3.8521
1 10 1 60 1.00E+05 10000 10859 8.6452
1 10 1 60 1.00E+05 1.00E+05 1128.8 79.862
1 10 1 60 1.00E+06 1000 23553 3.9048
1 10 1 60 1.00E+06 10000 9734.3 9.345
1 10 1 60 1.00E+06 1.00E+05 982 66.852
1 10 1 60 2.00E+06 1000 25408 3.5738
1 10 1 60 2.00E+06 10000 7993.9 9.2784
Configuration
Metrics
Measured
The performance
repository
mediates between
the optimization
component and
experimental suite

12. Configuration Parameters (Output)
12©DICE

13. Technology Support
13©DICE
Configuration
Optimisation Tool
performance
repository
Monitoring
Deployment Service
Data Preparation
configuration
parameters
values
configuration
parameters
values
Experimental Suite
Testbed
Doc
Data Broker
Tester
experiment time
polling interval
previous versions
configuration
parameters
GP model
Kafka
Vn
V1 V2
System Under Test
historical
data
Workload
Generator
Technology Interface
Storm
Cassandra
Spark
Technologies: Storm, Spark, Cassandra