Primary tab navigation

OpenPOWER Cloud

Accelerating Cloud Computing



OpenPower

19 November 2014, IBM announced a new OpenPOWER/POWER technology cloud,
the SuperVessel Cloud

 

 

OpenPower

Group photo of SuperVessel Cloud 2nd salon,
which was held in Xi'an

SuperVessel, an OpenPOWER cloud platform created by IBM Research – China, is open to students, faculty and other developers who want to participate in the OpenPOWER movement. It acts as a virtual R&D engine for the creation and testing of applications serving burgeoning areas such as mobile and data analysis.

IBM partners and developers industry-wide can leverage SuperVessel to speed up their application development on OpenPOWER. With the help of accelerators that are available on SuperVessel, enterprises and clients can apply them to virtual applications, design and verify customization in just a few minutes.

Technological Innovations

SuperVessel demonstrates the latest achievements of IBM’s open-computing technology. The cloud is based on POWER processors with FPGAs and GPUs to provide heterogeneous acceleration service, and uses OpenStack to manage the whole cloud. SuperVessel is divided into online "labs" where users can access open-source software, build and test applications and share experiences and best practices. The labs include Big Data, Internet of Things, acceleration and virtualization on POWER.

Heterogeneous computing (CPU + FPGA) will become a strong trend of the future, as it offers low-cost, low-power consumption, and high-performance computing capabilities. In comparison to traditional CPUs, the power consumption for FPGA is 90% lower. By using accelerator and FPGA together, different applications can become 300+ times faster in pattern matching, 200+ times faster for compression, 100+ times faster in machine learning and more.

Based on the FPGA accelerator virtualization technology, SuperVessel launched the first Application Acceleration Store in 2015, creating a new cloud-based market model for developing and sharing applications with accelerators. It is an important innovation that makes application easily accelerated in the cloud, allowing the accelerators broadly tried by different applications on the cloud. Application developers are able to apply a virtual machine (VM) to the SuperVessel Cloud and select an accelerator from the Application Acceleration Store. In less than a minute, developers will access the VM with FPGA accelerator and start application development. The design and trial of applications are simplified and a single FPGA chip could be virtualized and used by multiple virtual machines. It is also capable of running different accelerators at the same time.

Project Achievements

  1. FPGA virtualization in the cloud is creating new ways to virtualize and share FPGA accelerators among virtual machines. SuperVessel Cloud is a one-of-a-kind technology, providing pay-as-you-go services in the domain of virtualized FPGA resources.
  2. The Application Accelerator Store launched by IBM Research – China created a bridge between accelerator developers and clients. On the one hand, accelerator developers can upload and list their accelerator IPs in the store after a full automation process of compilation and verification. On the other hand, as clients they can apply their virtual machines online, specify and try out different accelerators. Once the accelerator and application are combined and tested, developers and businesses can move into further collaboration and new IP licensing can be initiated offline.
  3. IBM GCG and partners jointly announced this new business model around accelerator store spearheaded by SuperVessel Cloud for the OpenPOWER ecosystem.
  4. IBM signed the agreement with xuetangX.org, the biggest MOOC in China. In this agreement, SuperVessel is providing cloud services to support the new online education model created by xuetangX.org. With its FPGA virtualization technology and innovation capabilities, SuperVessel is leading the way in creating cloud environment for heterogeneous computation (CPU + FPGA/GPU) in the education and training domain.

Media Coverage