A Global Data Platform Architected to Break AI and Performance Barriers

• Read original article here

New IBM Storage Solution: A Global Data Platform Architected to Break AI and Performance Barriers
By David Wohlford posted 9 hours ago
  
Register for the webinar Breaking performance barriers for AI and hybrid cloud on Tuesday, May 17 at 11 AM ET. 
Artificial intelligence (AI) is rapidly being adopted and embedded into business and society.1 AI allows organizations to make better decisions and improve core business processes2.   The beginning of an AI pipeline is data with a continual need of more data to achieve more results3.   With more data comes more challenges and we start to see there is no AI without an IA (information architecture).  Innovation is needed if we are going to speed up and solve the difficult data challenges and barriers that exist.  Barriers such as data silos, infrastructure complexity, data mobility, application performance and the continual drive for forklift upgrades.
The goal should be to break these barriers and create a data platform that provides:
performance with multiple data interfaces
data connections that break silos
data automation for the management of all this data
data security that creates secure connections and secure data resources
That is why IBM® Storage has engineered a global data platform. This software platform consists of a set of core data services, with each core data service containing one or more specific data services to help break data barriers and solve customer application requirements.
The global data platform is designed with an integrated set of data services that are easy to upgrade without data disruption, easy to use with a GUI and API access, and is included with each deployment of our IBM Spectrum® Scale software.   To learn more about the value that customers have obtained with the IBM global data platform and IBM Spectrum Scale software, read  Forrester’s The Total Economic Impact™ Of IBM Spectrum Scale.
The functionality of the global data platform has its roots in over 23 years of customers that started in 1998 as IBM Global Parallel File System (GPFS™).   Since that time, it has grown to over 4000 customers and has become both a software defined platform with IBM Spectrum Scale and simple to deploy building blocks with IBM Elastic Storage® System (ESS).
Today we are pleased to announce an important milestone in the history for IBM Spectrum Scale and IBM ESS, the IBM ESS 3500.  This exciting new addition to IBM ESS product family will be generally available on May 20 and is available now for ordering.
The IBM ESS 3500 has been enhanced for easy installation and scalability which is powered by IBM Spectrum Scale software.  It is targeted for AI, machine learning (ML), high-performance computing (HPC), and analytics workloads.  With the ability to run applications in parallel the system is also well suited for data collaboration, critical business applications, hybrid cloud applications and consolidation of multiple workloads on a single system. (See table).
Workloads and Applications
AI/ML/HPC
NVIDIA, TensorFlow, Caffe, Pytorch, IBM Cloud® Pak for Data, ML Kit, AWS Lambda, AWS SageMaker …
Faster GPU and AI analysis with access to more data and connected data with high-performance data access services
Analytics
Cloudera, Hadoop, Apache Spark™, SAS®, Tableau, Python, Power BI, …
Eliminate silos and speed time to value with data caching services and combined HDFS and S3
Collaboration
Media Development, AI Model Development, Computer Aided Engineering
Increase workgroup productivity with easier access to more data from edge to core to cloud
Critical Apps
Business critical applications that have expanding data requirements
Maintain business continuity with data that is always online and secure with security services
Hybrid Cloud
Red Hat® OpenShift® workloads, backup and archive, applications that move data or access on-prem and cloud
Faster time to production as data is consumed anywhere and connected everywhere
Consolidation
Data lake, data ocean, and big data
Lower costs as data is resilient and accessible without duplication
2u per node