IBM Advances its High Performance Data Analytics Arsenal with its Spectrum Computing Platform
As the need for gathering data continues, organizations keep dealing with increasing amounts of information that need to be stored, processed, and analyzed faster and better, stimulating the growth and evolution of the high performance computing (HPC) market.
One key segment in this market the continues to grow, especially in recent years, is the high performance data analytics (HPDA) as organizations continue to adopt and evolve their big data and data lake initiatives, to the point that IDC forecasts that in 2018, the HPDA server market will reach $2.6 billion (23.5% CAGR) and the HPDA external storage market will add $1.6 billion (26.5% CAGR).
Is not strange then, that software companies like IBM are keen to develop solutions capable to address the specific HPDA market segment, and IBM haws been working specifically on it with its Spectrum software line.
IBM Evolves Spectrum to Keep the Pace with the HPDA Market Segment
Late in November of last year, IBM announced a brand new software offering called IBM Spectrum Computing. The announcement signals how IBM is setting itself on route to address a not huge but important market segment of the IT industry: The provision of analytics solutions for high performance workloads or High Performance Data Analytics (HPDA).
With this, IBM aims to continue its strategy to serve and provide advanced data management and analytics solutions for an increasing set of computing ecosystems and environments. The new offering will enable organizations to work on high data volumes being analyze within advanced analytics software solutions including Spark, TensorFlow or Caffé to deploy applications that use machine learning, artificial intelligence and deep learning.
As IBM mentions:
IBM Spectrum Computing uses intelligent workload and policy-driven resource management to optimize resources across the data center, on premises and in the cloud.
According to IBM, the new software platforms is designed to set up distributed, mission-critical, HPC infrastructures, scalable to over 160,000 cores and enable execution of big data analytics applications at speeds up to 150x faster. The new Spectrum Computing offering from IBM is now set to deliver open offerings developed especially to increase speed of adoption, and production of parallel processing and clustered computing tasks to allow organizations to:
- Simplify deployment via cluster virtualization and enabling disperse systems to work together as one by including shared computing, data services and cluster management specifically configured deep learning/machine learning tasks.
- Deliver artificial intelligence (AI) for the enterprise, enable centralized management and reporting and allow multi-tenant access with end-to-end security.
- Openness to the latest technology so it can provide support for the latest open source tools and protocols, including Spark and other deep learning software options.
- Ease the deployment and adoption of cognitive workloads so end users will have access to the consumption of cluster resources across applications without having the need for specialized cluster knowledge.
- Elasticity for hybrid clouds by simplifying cloud usage in distributed clustered environments, incorporating cloud automated workload-driven provisioning and de-provisioning routines to encourage intelligent workload and data transfer to and from the cloud
According to the “big blue”, the Spectrum Computing family has been designed to provide support for hybrid architectures so it can accommodate to specific workloads as well as to specific platforms to enable optimal performance. Interestingly, IBM Spectrum Computing offers support for mixed x86-IBM POWER platforms environments. Scalable to over 160,000 cores, IBM aims to enable users to fulfill their high performance analytics deployment needs with state-of-the-art software-defined computing technology for complex distributed, mission-critical, high-performance computing (HPC) infrastructures. Core components of the Spectrum Computing Platform include:
- IBM Spectrum LSF. Spectrum platforms’ workload management solution which provides user management and workload administration tasks.
- IBM Spectrum MPI. Spectrum’s message passing interface (MPI) with high-performance capabilities and configured specifically to support distributed computing environments.
- IBM Spectrum Cluster Foundation. IBM Spectrum’s infrastructure life cycle management solution for scale-out environments and coming at no charge.
- IBM High Performance Services. Hosted on the Softlayer (IBM Cloud), this is IBM’s HPC and storage clusters as a service infrastructure.
Additional to the launch is the coming of a new set of tools developed to help users to design artificial intelligence (AI) models using much of the leading deep learning frameworks including Tensor Flow and Caffé. One key aspect of this set of releases includes the latest version of IBM Spectrum Scale software ―Spectrum’s advanced storage management― which now provides full support for moving workloads including unified files, objects and HDFS right from where they´re storage to where they´re analyzed. This latest version of IBM Spectrum LSF Suites is now generally available, along with the new versions of IBM Spectrum Conductor with Spark, Deep Learning Impact and IBM Spectrum Scale.
So What?
Well, as solutions for HPDA evolve, so they market reach, we are seeing continues adoption within many industries and business segments: finance, security, retail, and the adoption of new technologies that will require solutions capable of analyzing petabytes of data through parallel processing compute resources.
The latest version of IBM Spectrum Computing aims to put IBM in the forefront of the HPDA segment and better compete with other giants in this market like HPE and Hitachi. A competition field that will be interesting to keep an eye on in the coming years.
Finally, feel free to check this Spectrum intro video from IBM and of course, to let me know your thoughts right below.
Originally published at www.dofthings.com.
