Technical Requirements

Copyright © 2021 by Symetry, Inc. 14 Pine Street, Ste 6 Morristown, NJ 07960 All Rights Reserved April 15th, 2021

Technical Requirements

Technical Requirement for SymetryML

Requirement	Description
Operating System	Centos 7.x or Amazon Linux Based on Red Hat 7.x
Application Server	Certified with Jetty 9.4.49
Additional Drivers	CUDA version 10 or up
GPU	NVIDIA Tesla or GPU with minimum compute capability of 3.0
Spark	Spark 2.4.x
Java	Java 11
Redis	Redis 2.8.x and up

System Recommendations

Recommendation	Description
Spark Cluster worker memory	Minimum: 8 GB Recommended: 16 GB and more

GPU Support

GPU Support	Description
CUDA library	CUDA Version 10.x
Intel MKL	Version 11.0+

SymetryML Memory Requirements

SymetryML memory requirements depend on the number of projects and the number of attributes in each project. The following table provides a rough guideline as to the compute and memory requirements for your projects. These numbers are for dense datasets. For sparse datasets, the number will vary.

Number of Attributes	Type of Project	RAM Memory Requirement
10	CPU	512m
64	CPU	512m
128	CPU or GPU	1g
512	GPU	1g
1024	GPU	1g
4096	GPU or multi-GPU	4g
8192	GPU or multi-GPU	16g
12,500	GPU or multi-GPU	24g
20,000-25,000	multi-GPU	64g
25,000-100,000	multi-GPU on Nvidia V100	64G to 256G

Data Source Requirements

Data Source	Description
JDBC / ODBC
S3
SFTP
HTTP / HTTPS
Local File	Local file where the application server resides
Kafka Streams
RedShift
Plugins	It's possible to add new 'plugin' data source to SymetryML
SymetryML Rest API	One can use the SymetryML Rest API to push data into a SymetryML project. Please consult the following section for details (section called Real Time Streaming of Data into SymetryML Projects).