Hyperkernel, a key technology of TidalScale, is an inverse hypervisor technology that allows a software-defined server to consolidate all the resources of multiple nodes such as CPU, memory, storage, and network into a single system. Located between hardware and guest OS, Hyperkernel optimizes the virtual CPU and memory moving through patented machine learning in microseconds and moves them to a completely redesigned interconnection network, resulting in increasingly faster and evolved.
TidalScale solution supports to consolidate multiple commodity servers into one or more virtual servers, regardless of operating system, thus implementing server environment much easier than the method that the user is currently using, as well as extending the useful life and value of the data center. This enables enterprises to build supercomputer-class systems for large-scale analysis, memory-intensive computing, computing-intensive workloads, and modeling. On top of that, TidalScale stated that it will strengthen its position in the Korean and Asian markets and expand the market by using the momentum of establishing a Korean branch.
Dave Ferretti, Vice President of Worldwide Sales, TidalScale, stated that the current data is soaring and new economics are required for handling data. In addition, he emphasized that data center infrastructure also requires new definitions, and the use of memory, storage, compute and applications is dramatically changing. Memory on the server will eventually affect performance, and data must be handled through in-memory technology. He added that the data must be handled in a larger memory configuration that fits the cloud environment.
IT infrastructure needs to be able to quickly support the right size for the business, and technology is needed to respond quickly to changes such as rapid growth. TidalScale supports these requirements by using multiple commodity servers to configure the size of the computing resources at that time, and to substitute a two socket server for a large server that was replaced for every few years.
Larger computational resources, which can be obtained through the new architecture, can perform more complicated simulations. They are also expected to quicken decision-making in real time by handling multiple pieces of information within memory in a situation where the transaction size of financial analysis is getting bigger. Meanwhile, TidalScale has many partners and customers around the world, and Oracle has recently launched a service that incorporates TidalScale’s technology in its cloud. Also, SAP is overcoming the aspect of large memory requirements in HANA operations with the technology of TidalScale.
The Korean branch of TidalScale is the first overseas branch, and the reason for the selection is that Korea is the center of the world's memory and flash technology. It is also emphasized that although the domestic business is important, it has a goal to prepare by cooperating closely with memory and flash-related companies in product selling and sales to the world, and to introduce new data center architecture. Moreover, TidalScale configures and operates resources at the right size within the data center through rack-level virtualization, and extends the use of flash tied to a single server.
As things necessary for TidalScale’s ‘software-defined server’, Jamon Bowen, who is sales engineering director at TidalScale, pointed out that it must be able to operate at an appropriate size, be able to produce in-memory performance, be self-optimizing, and be available in commodity hardware. And both the application and the operating system must be transparently usable as if they were used on a hardware basis, without modification.
In a traditional software-defined data center, innovation based on “software-defined” has been reflected to storage, network, and management part, and paradoxically, “server” was mentioned as the last piece for making resources be operated at an appropriate size. The existing virtualization technology has a virtualized processor and memory, and is connected to the hardware by a hypervisor, which allows creating only small virtual machines on one server. When adding more capacity, there is a difficulty that the new instance and the changed application have to be applied.
Contrary to the existing virtualization where multiple operating systems run on one server, TidalScale's core technology, 'Hyperkernel', is introduced as a “inverse hypervisor” technology, which can integrate all resources from multiple nodes, including processors, memory, storage, and networks, into a single system. This allows a single VM to expand directly to this added resource when adding a resource pool, making it possible without application modification. This allows flexibility in the data center, enabling servers to operate at the right size as needed.
In this virtualization environment, operating systems and applications use virtualized resources, and the advantage of virtualization technology is that the resources can be moved. It is difficult to avoid performance degradation when going through the network, but TidalScale can increase infrastructure utilization and optimize performance by making not only the processor but also the memory move.
Located between hardware and guest OS, Hyperkernel optimizes the virtual processor and memory moving through machine learning in microseconds and moves them to a completely redesigned interconnection network, resulting in increasingly faster and evolved. Based on this technology, it is possible to configure as a single virtual mainboard in one large VM with multiple commodity two-socket systems. In addition, a commodity server can be used as a building block to produce necessary server on a software basis, to implement an operating system and an application without modification, and to implement various configurations with flexibility.
Meanwhile, TidalScale's solution is based on Intel processors, and while pool configuration and integrative utilization with different generational platforms are not recommended, it can be configured and used in any direction. What’s more, moving the processor in the movement of resources does not consume much network resources, but memory movement requires bandwidth. In this respect, the performance improvement of network interface on recent platforms will be very helpful for the spread of such solutions. It is also added that a minimum 10Gbps infrastructure is recommended for this solution, and there is a need to look at latency aspects as well.
Park Un-young, Country Manager of TidalScale Korea, said that TidalScale's Korea branch has plans to advance into the third country as well as the domestic market. In addition, he pointed out all the markets that need IT as the strategic market, and introduced that it will practically focus on big data and cloud domains. In the big data area, it is possible to enhance the competitiveness of the company by overcoming the difficulties of distributed processing implementation in R analysis. In the cloud area, it is possible to provide flexibility of change in infrastructure scale in software base. It also provides a new profit model for the business and convenience benefits for the customer.
Copyright ⓒ Acrofan All Right Reserved