pdgf: a data generator for cloud-scale benchmarking
DESCRIPTION
This is a presentation that was held at the Second TPC Technology Conference on Performance Evaluation & Benchmarking, 17.09.2010, Singapore. Abstract: In many fields of research and business data sizes are breaking the petabyte barrier. This imposes new problems and research possibilities for the database community. Usually, data of this size is stored in large clusters or clouds. Although clouds have become very popular in recent years, there is only little work on benchmarking cloud applications. We present a data generator for cloud sized applications. Its architecture makes the data generator easy to extend and to configure. A key feature is the high degree of parallelism that allows linear scaling for arbitrary numbers of nodes.We show how distributions, relationships and dependencies in data can be computed in parallel with linear speed up.TRANSCRIPT
T. Rabl, M. Frank, H. Mousselly Sergieh, and H. Kosch17.09.2010Second TPC Technology Conference on Performance Evaluation & Benchmarking
Data sizes grow beyond petabyte barrier Cloud computing to the rescue
Need for cloud-scale benchmarking Need for realistic, cloud-scale data sets
Problems: Generating petabytes Storing petabytes Transporting petabytes
Solution: Parallel, on-site generation
2A Data Generator for Cloud-Scale Benchmarking - Tilmann Rabl
3 tables with primary and foreign keys Non-uniform distributions (lognormal) Replicated data
How to generate consistent data in parallel?3A Data Generator for Cloud-Scale Benchmarking - Tilmann Rabl
No references Only uncorrelated data Simple statistical references
Scanned references Read data to generate reference Generate tuple pairs
Computed references Data is generated deterministically Compute data to generate reference
A Data Generator for Cloud-Scale Benchmarking - Tilmann Rabl 4
Parallel Data Generation Framework
Parallel pseudo random number generator Deterministic Fast skip ahead
Generation of values as a function Deterministic
Seeds + row id + generator allows recalculation Every value can be computed independently
A Data Generator for Cloud-Scale Benchmarking - Tilmann Rabl 5
Seeding strategy All seeds can be cached Generation of value in row n with n-th
random number
A Data Generator for Cloud-Scale Benchmarking - Tilmann Rabl 6
useruser_idrow_id degree_program …
123
Table RNGseed
Column RNGseed
Row RNGseed
Generatorrn
c_id
t_id
r_id
Reference generation Lognormal indexes row_id of user Equal seeds and lognormal generator for user_id
A Data Generator for Cloud-Scale Benchmarking - Tilmann Rabl 7
seminar_useruser_idrow_id degree_program …
123
Table RNGseed
Column RNGseed
Row RNGseed
Lognormalrn
id
id
r_id
Row RNGseed
Generatorrn
r_id
Java based Plug-in concept Generators Distributions RNGs Output Scheduler
TPC-H plug-in
8A Data Generator for Cloud-Scale Benchmarking - Tilmann Rabl
XML file Reflects SQL schema Tables Fields
Seed Size Scale factor Output
<project name="simpleUserSeminar">[..]<table name=“seminar_user">
<size>201754</size><fields><field name="user_id"><type>java.sql.Types.INTEGER</type><reference><referencedField>user_id</referencedField><referencedTable>user</referencedTable></reference><generator name="DefaultReferenceGenerator"><distribution name="LogNormal"><mu>7.60021</mu><sigma>1.40058</sigma></distribution></generator></field><field name="degree_program">
[..]
9A Data Generator for Cloud-Scale Benchmarking - Tilmann Rabl
Every value can be computed independently No communication Workload distribution Configuration for every node Automatic distribution
<?xml version="1.0" encoding="UTF-8"?><nodeConfig><nodeNumber>1</nodeNumber><nodeCount>2</nodeCount><workers>2</workers>
</nodeConfig>
Processors
Nodes
Cluster/Cloud seminar_user
100000 rows
Row 1 – 50000
Row 1 –25000
Row 25001 –50000
Row 50001 – 100000
Row 50001 –
75000Row 75001 –
10000010A Data Generator for Cloud-Scale Benchmarking - Tilmann Rabl
16 node HPC cluster 2 Intel Xeon QuadCore processors 16GB RAM 2 x 74GB HDD, RAID 0
SetQuery data set 1 table “Bench”, 21 columns 12 Random numbers, 8 strings
TPC-H data set 8 tables, 61 columns Data types: integer, char, varchar, decimal, date
11A Data Generator for Cloud-Scale Benchmarking - Tilmann Rabl
SetQuery data set 100 GB data set (SF 460) 1 to 16 nodes
12A Data Generator for Cloud-Scale Benchmarking - Tilmann Rabl
TPC-H data set Data sizes: 1 GB, 10 GB, 100 GB
Single node (8 cores)13A Data Generator for Cloud-Scale Benchmarking - Tilmann Rabl
Parallel data generation framework Linear speed up Fast generation Large data sets Realistic data
Independent generation of tables / colums / values
Easy configuration and extension
14A Data Generator for Cloud-Scale Benchmarking - Tilmann Rabl
More implementation generators, distributions Benchmarks Graphical user interface Scheduler
Query generator Consistent inserts, updates, deletes Precomputed query results Time series
15A Data Generator for Cloud-Scale Benchmarking - Tilmann Rabl
SetQuery data set Data sizes: 220 MB, 2.2 GB, 10 GB, 22 GB, 100 GB
Single node17A Data Generator for Cloud-Scale Benchmarking - Tilmann Rabl
TPC-H data set Data sizes: 1 GB, 10 GB, 100 GB, 1TB
1, 10, 16 nodes18A Data Generator for Cloud-Scale Benchmarking - Tilmann Rabl
19A Data Generator for Cloud-Scale Benchmarking - Tilmann Rabl
20A Data Generator for Cloud-Scale Benchmarking - Tilmann Rabl