Line Rate Compression
"Eideticom’s NoLoad provides hardware-based compression that enables increased capacity (lower $/GB) without sacrificing performance"
"Eideticom’s NoLoad CSP achieves higher Compression Ratio and Throughput than software while using 70% less CPU"
Today's world of faster and more virtualized-servers, storage, and network connections, means that workloads are growing in scale and complexity. A critical application to guarantee improved Quality of Service (QoS) is the database, and a database's interactions with storage are critical to improve overall speed and user experience. Eideticom's Noload® Computational Storage Processor (CSP) delivers the flexibility to accelerate a range of database applications by leveraging hardware-based computational offloads with unmatched scalability, performance and efficiency.
Eideticom's NoLoad® Computational Storage Processor (CSP) is the ideal solution to accelerate MapReduce workloads. Hadoop MapReduce is the original framework for writing applications that process large amounts of structured and unstructured data stored in the Hadoop Distributed File System (HDFS). Eideticom’s NoLoad solution accelerates both Compression and Erasure Coding, making it the ideal solution to accelerate MapReduce workloads. The end result is significant improvements in performance and cost efficiency when compared to a software-only solution.
Hadoop Video (coming soon)
Eideticom accelerates production grade key-value store RocksDB databases with our NoLoad® Computational Storage Processor (CSP). The result is dramatically improved data-center Total Cost of Ownership (TCO) and significantly increased database Quality of Service (QoS).
Cassandra is an incredibly popular open-source and distributed NoSQL database that is deployed in the most demanding cloud environments. Since Cassandra deployments often scale to hundreds or thousands of nodes, efficiency is of the utmost importance.
Eidieticom's NoLoad CSP can be used of offload key database tasks and to reduce data movement which leads to more efficient systems. This allows customers to scale out over less servers but still satisfy their customers’ requirements.
MySQL is an open-source relational database used to store some of the most important business critical information. Safely deploying MySQL across many servers is a key requirement of most companies and Eideticom's NoLoad CSP can facilitate this by offloading key MySQL tasks. This leads to more efficient systems, reduced total cost of ownership and better performing systems.
Peer to Peer (P2P)
PCIe End-Points (EPs) are getting faster and faster e.g. NVMe SSDs, RDMA NICs & GPGPUs. Bounce buffering all I/O data through system memory is a waste of system resources and reduces QoS for CPU memory. Eideticom's NoLoad P2P allows PCIe EPs to DMA to each other whilst under host CPU control. 99.99% of DMA traffic now goes directly between EPs, so P2P avoids CPU's memory subsystem.
Eideticom's NoLoad Accelerators identify as NVMe Namespaces, which can be accessed/shared using NVMe-oF. NoLoad Accelerators located in a remote server can be accessed by any client with a RDMA or TCP/IP connection. Eideticom enables the disaggregation of FPGA Accelerators using NoLoad CSP and NVMe-oF. We help our customers get their accelerators out-of-the-box and shared across the datacenter.
AI and ML are becoming dominant workloads in both data-centers and at the edge. However, AI and ML algorithms run very inefficiently on standard CPUs. Eideticom's NoLoad CSP allows customers to run AI and ML workloads at scale in a cost-effective and efficient manner. By offloading key AI and ML tasks to our NoLoad CSP customers get results faster, for lower cost and more efficiently than via CPUs.
With the explosive growth in data, the ability for conventional computer architectures to perform analytics has been compromised. Eideticom's NoLoad CSP offers a new way to perform data analytics that leads to scalable, efficient and lower cost solutions. By offloading key computation tasks to our NoLoad Computational Storage Processor, and by reducing data movement, we can scale data analytics in a cost and power efficient way.