JET INFOSYSTEMS The main approach to Big Data

JET INFOSYSTEMS The main approach to Big Data parallel processing: Oracle way Aleksey Struchenko Database Department Leader 3/7/2021

Parallel Processing of Data • Parallel processing is the main principle of big data computing • Clusters with hundreds of nodes compute the Physics data • In databases the parallel querying methods for SQL and Non-SQL Databases can be different, this presentation focuses on SQL DB (first of all – Oracle as a leader in databases) • In databases area there are two main approaches for clusters in architecture: Shared Nothing and Shared Everything © 2015 Jet Infosystems 2

Shared Nothing Cluster • Every computing node has its own storage and data, the coordinator distributes the query between nodes and aggregates the results • Shared Noting Clusters are used for parallel processing in all Non-SQL Databases and most SQL Databases • For SQL Databases this approach is often called MPP (Massive Parallel Processing): Teradata, IBM Netezza, EMC Green Plum • The main benefit is the absolute scalability, but cluster reconfigurations and multi-table join queries are the real problems for MPP © 2015 Jet Infosystems 3

Shared Everything Cluster • All cluster nodes query the data from shared storage, in any case the memory is shared between nodes with special technology • The best-known implementation is RAC (Real Application Clusters) – Oracle Database Enterprise Edition Option • No limitation for multi-table join queries and easy cluster reconfiguration, but the scalability of RAC needs special testing © 2015 Jet Infosystems 4

Oracle Exadata Database Server RAC Infini. Band Exadata Storage Server Exadata Software: • Smart Scan (offload) • Compression Disks + Flash (or Flash only) © 2015 Jet Infosystems Exadata Storage Server Exadata Software: • Smart Scan (offload) • Compression = Disks + Flash (or Flash only) 5

Oracle Engineered Systems © 2015 Jet Infosystems 6

About Jet Infosystems • A full-profile systems integrator, and a leader in IT market of Russia and CIS, more than 1000 employees • An Oracle Platinum Partner (15+ years) with OCM specializations • The first Oracle Exadata demo-center in Russia (since 2010) • The largest service centre for corporate class solutions (including Oracle Database and Oracle Exadata outsourcing) • Collaborated with JINR since 1993 (Sun, Cisco, Brocade, Huawei) • Questions: struchenko@jet. msk. su (Aleksey Struchenko) © 2015 Jet Infosystems 7