How and where the technologies will disrupt IT spend
The development and evolution of devices that accelerate data handling and offload the computation processor continues. The Evaluator Group paper on Data Flow Acceleration and how it addresses the next bottleneck in storing data first explored this topic. As developments have progressed, the need for an update and a perspective on potential impacts on the industry have become evident. Along with the update is the need for some clarity. As with anything in the information technology area, there are few clear definitions and market categories. These become even more diffused over time with variations and different messaging. Data acceleration and offload is no different and further explanations are necessary.
A number of companies have recognized the need to accelerate access to data and functionalities related to handling data rather than burdening the computation element in a system. The need leads to opportunity to provide high-value solutions. Many of these companies are startups or established vendors that offer specialized solutions used in systems. When delivered by the larger system vendors, the solutions will have the greatest impact on the industry because of their market presence.
With the lack of concise definitions, understanding the goals and value delivered will help to gauge the long-term impact on the industry.
Context: Understanding Data Acceleration and Computation Offload
The names used by different vendors and some writing about the topic are inconsistent and are not really self-descriptive – the capabilities go beyond what the names imply. The functional differences and what can be done with the technology for data acceleration and computation offload are what is important rather than some of the nuanced implementation details. Some details are meaningful, but the industry is at an early stage where differential aspects are less important than what can be achieved in both function and, ultimately, economic value.
For a general understanding data acceleration and offload technology, an explanation is in order for the terms used. Solutions fall into four categories:
- Data Processing Units
- Computational Storage
- Storage Processing Units or Service Processing Units
This explanation will not necessarily coincide with vendor marketing messages. The goal here is for greater clarity because there is such a mixture of messaging and information. Examples of each are included.
Vendors mentioned: Fungible, Nvidia, Pensando, Broadcom, Marvell, Netronome, NGD Systems, Samsung, ScaleFlux, Eideticom, Kalray, Nebulon, Pliops
Download now to read the complete free Technical Insight report.