2 Min Read
KT Corporation has unveiled an "AI-only" line service, offering bandwidth options of either 400 Gbit/s or 800 Gbit/s to support the large-volume data transfers required by AI data centers.
On Wednesday, South Korean telecom operators told local media that networks capable of reliably connecting AI data centers and handling large volumes of data are emerging as critical infrastructure as these centers become increasingly distributed across the country.
The new bandwidth capacity is up to eight times greater than the previous maximum of 100 Gbit/s. While 100 Gbit/s is sufficient for most enterprise needs, including data replication, video backhaul and standard cloud connectivity, it falls short of what is required for synchronizing distributed AI training runs across multiple AI data centers.
In essence, although the speed of data transfer remains the same, the increased bandwidth capacity is comparable to a one-lane street being widened into an eight-lane highway, enabling far greater volumes of traffic to flow simultaneously and preventing congestion.
Related:Orange Business to provide European Union's backbone network for trusted data exchange
This distinction is critical for infrastructure such as AI data centers. Training large language models requires moving massive, terabyte-sized datasets between clusters of GPUs. While upgrading to 800 Gbit/s does not make the first byte arrive any quicker, it drastically reduces the time taken to transfer large amounts of data by preventing network bottlenecks.
KT Corporation did not disclose pricing for its two-tiered dedicated AI line.
An all-photonic network architecture
The South Korean operator said it has implemented all-photonic network (APN) technology to improve the ultra-low latency of the new AI-only lines. APN technology minimizes conversions between optical and electrical signals along the data transmission path.
"Our infrastructure strengths, including nationwide optical cable coverage of 840,000 kilometers and diversified routes, allow us to design optimal shortest and dual paths between multiple data centers or sites, minimizing latency and ensuring stability," the company said as quoted by the Seoul Economic Daily.
Looking ahead, KT Corporation also said it intends to extend the use of its AI-only line to data center operators and enterprises developing large-scale AI infrastructure.
About the Author
Senior Editor, APAC, Light Reading
Gigi Onag is Senior Editor, APAC, Light Reading. She has been a technology journalist for more than 15 years, covering various aspects of enterprise IT across Asia-Pacific.
She started with regional IT publications under CMP Asia (now Informa), including Asia Computer Weekly, Intelligent Enterprise Asia and Network Computing Asia and Teledotcom Asia. This was followed by stints with Computerworld Hong Kong and sister publications FutureIoT and FutureCIO. She had contributed articles to South China Morning Post, TechTarget and PC Market among others.
She interspersed her career as a technology editor with a brief sojourn into public relations before returning to journalism, joining the editorial team of Mix Magazine, a MICE publication and its sister publication Business Traveller Asia Pacific.
Gigi is based in Hong Kong and is keen to delve deeper into the region’s wide wild world of telecoms.