• Categories:Industry News
• Time of issue:2024-01-19

(Summary description)According to Dell'oro Group's latest "AI Networks for AI Workloads" report, switch expenditures deployed in the artificial intelligence back -end network are expected to expand the data center switch market by 50% 200G QSFP56 AOC. The current data center switch market expenditure is mainly used to connect the front -end network of universal servers. 200G QSFP56 AOCArtificial intelligence workloads will need to establish a new back -end infrastructure. With the market leadership of major companies competing for artificial intelligence back -end networks, competition between Infiniband and Ethernet is increasing. Although Infiniband is expected to maintain a leading position, it is expected that Ethernet will make significant progress, such as its share (referring to income) by 2027 will rise by 20 percentage points 200G QSFP56 AOC. 200G QSFP56 AOC Sameh Boujelbene, Vice President of Dell'oro Group, said: "Genesis artificial intelligence applications have created a new era in the era of artificial intelligence, because they must handle the number of parameters. At present, some large artificial intelligence applications process trillions of parameters. This number is this number. Ten times each year. This rapid growth requires the deployment of thousands of or even hundreds of thousands of acceleration nodes. Connecting these accelerated nodes in large clusters requires a structure of data center scale, called an artificial intelligence back -end network. It is different from the main use200G QSFP56 AOC. Traditional front -end network connecting the universal server 200G QSFP56 AOC. " "This dilemma has raised a key question: what is the most suitable structure that can expand to hundreds of thousands or even millions of acceleration nodes, and at the same time ensure the minimum job completion time (JCT)? Said that Ethernet is faster than Infiniband. However, 200G QSFP56 AOC the speed of the network is not the only factor. Celection control and adaptive routing mechanism is also important. We analyze the main cloud service providers (such as Google, Amazon, Microsoft, Meta, Ali The artificial intelligence back -end network built by Baba, Tencent, Bytes, Baidu, etc.), as well as various considerations to push them to choose the back -end structure, can draw our predictions, "Boujelbene continued 200G QSFP56 AOC. Other highlights of "Artificial Intelligence Networks of Artificial Intelligence Working 200G QSFP56 AOC": 1. The artificial intelligence network will accelerate to a higher speed. For example, in 2025, in the two years when the latest 800Gbps products are launched, 800Gbps will account for most ports of the artificial intelligence back -end network 200G QSFP56 AOC. 2. Although most market demand will come from Tier 1 (Tier 1) cloud service providers, it is expected that the needs of level 2/3 (Tier 2/3) and large enterprises will be very huge, and it will be close to $ 10 billion in the next five years. The latter tends to Ethernet 200G QSFP56 AOC. The company has a dust -free production workshop of 10,000 square meters. The production capacity reaches 24 million source devices/year, and 30 million yuan/year light -free device is one of the largest companies in the field of optical communication and packaging production in China. The company's products are widely served in various types of application scenarios such as access networks, mobile wireless communication, transmission networks, data centers, cloud computing, HPCs, cable TV, security monitoring, smart grids, industrial and consumer sensing, and Lidar (LIDAR) 200G QSFP56 AOC.