Gcore, a European CDN provider that offers high-performance, low-latency, multinational cloud and edge solutions, has introduced Cloud IPU Virtual vPODs, which expand the capabilities of Gcore’s artificial intelligence (AI) infrastructure. This innovative solution provides users with direct access to BOW host machines and significantly cuts down the amount of time required for deployment. As a result, Gcore would offer users a cutting-edge and powerful means of accessing their AI infrastructure.
Gcore’s IPU-based AI Cloud is intended to be of assistance to businesses operating in a wide variety of industries, including but not limited to healthcare, manufacturing, scientific research, and the financial sector. It is designed to support every stage of their journey toward adopting AI, beginning with the construction of proofs of concept and continuing through training and deployment.
Gcore has collaborated with Graphcore in order to satisfy the quickly expanding demand for powerful, efficient, and private artificial intelligence processing in the cloud. To this end, Gcore has implemented Graphcore’s “cutting-edge” IPU technology.
The capabilities of Gcore’s AI Cloud are currently undergoing expansion. The Cloud IPU Virtual vPOD is a variant of AI cluster in which a server is deployed on a virtual computer rather than a dedicated vPOD, which deploys a server on a dedicated bare metal server. This is in contrast to the traditional deployment method of deploying a server on a dedicated bare metal server.
Users are given simplified access to IPU hardware through the use of Cloud IPU Virtual vPODs. The users of each IPU virtual instance are granted full access, which enables them to install and run any code of their choosing. This would allow for a connection with IPU accelerators that is extremely fast, ephemeral storage, execution of user-defined code in input pipelines, and improved integration of Cloud IPUs into research and production workflows.
“Cloud IPU Virtual vPODs are an excellent way for businesses to improve their machine-learning performance,” said Seva Vayner, Director of Edge Cloud stream at Gcore. “Because of this technology, the construction of dependable infrastructures that are capable of being reallocated rapidly is made possible. As a result of its ability to cut down on wasteful expenditures in areas such as provisioning, electricity, and maintenance, it is an extremely cost-effective solution. Cloud IPU Virtual vPODs provide companies with increased performance and scalability while simultaneously saving them money.”
IPU Technology
IPU stands for ‘Intelligence Processing Unit’ and refers to a specialized processor designed to accelerate the processing of artificial intelligence (AI) workloads. The technology is primarily associated with the company Graphcore, which has developed a line of IPU chips that can be used to power machine learning and other AI applications.
The IPU is designed to work in tandem with a host processor, such as a CPU or GPU, to accelerate AI computations. While traditional processors are optimized for general-purpose computing, IPUs are optimized for the specific requirements of AI workloads, such as neural network training and inference. They typically feature a large number of parallel processing units, specialized memory structures, and unique software and hardware architectures to deliver high performance on AI workloads.
In addition to its IPU chips, Graphcore also offers a software development kit and other tools to help developers build AI applications that can take advantage of the technology. Other companies, such as Google and Intel, have also developed their own AI-specific processors, although these may not be referred to specifically as IPUs.
