Japan’s NTT demonstrates ultra-low latency between data centres

Japanese companies NTT Corporation and NTT Data have carried out successful demonstrations in two different countries of ultra-low latency connections between datacenters using the All-Photonics Network (APN) solution. The aim of the initiative is to evaluate the transformation of geographically distributed IT infrastructures into functional equivalents of a single datacenter.

In the UK, datacenters were connected to the north and east of London via NTT’s Innovative Optical Wireless Network (IOWN), and communication between the facilities was achieved with a round-trip delay of less than 1 millisecond. In the United States, datacenters in Northern Virginia achieved similar results.

The two companies point out that the datacenter sector is subject to severe restrictions in several locations. Emissions restrictions and land scarcity are making it difficult to build datacenters in urban areas, causing operators to turn to suburban areas. However, geographically distant datacenters can generate higher latencies and make it more difficult to meet customers’ communication demands. Hence the validity of NTT Data and NTT’s demonstration – to virtually transform distributed datacenters into a single facility.

The datacenters tested in the United Kingdom and the United States are 89 km and 4 km apart respectively. Measurements carried out using 100 Gbps and 400 Gbps links showed that the two datacenters connected by the APN solution in the UK operated with less than 1 millisecond (approximately 0.9 ms) and a delay variation (jitter) of less than 0.1 ms. According to cloud connectivity provider Megaport, the typical delay between datacenters at an equivalent distance is more than 2 milliseconds.

In the case of the United States, the delay was approximately 0.06 milliseconds; the delay variation was less than 0.05 microseconds. In comparison, conventional networks with layer 2 switches have a delay variation of a few microseconds to tens of microseconds.

In summary, the APN solution has halved latency and jitter by orders of magnitude, proving capable of meeting the requirements of current and emerging use cases, such as distributed and real-time Artificial Intelligence analysis for industrial IoT and predictive maintenance, intelligent surveillance systems, smart grids, energy management and natural disaster response. NTT Data is carrying out further demonstrations in the financial sector, which requires low latency for remittances, settlements and transactions.

According to the companies, another advantage of the IOWN solution is that it can be used simply by adding wavelengths, without the need to install new dark fibres. In this way, datacenter operators will be able to respond very quickly to customer demands.

Other factors and strategies

It’s worth remembering that in addition to distance, other factors can affect latency, such as bandwidth and throughput.

Among the strategies employed to reduce communication overload in distributed datacenters are parallel processing, asynchronous communication, load balancing, caching, resource allocation algorithms and, of course, proximity to end users.

NTT’s track record with ultra-low latency

NTT is actively involved in the development and implementation of ultra-low latency network solutions. In 2022, it presented a technology capable of transmitting and receiving video in uncompressed 8K120p format over optical networks. According to the company, this technology aims to offer very high quality video communication regardless of distance, with almost instantaneous response rates also for applications other than video, such as high frequency rate negotiations.

More recently, in partnership with Olympus, NTT started the world’s first joint trial of a cloud endoscopy system. The solution also uses the ultra-low latency All-Photonics Network IOWN technology.

The system will allow endoscope images to be processed in real time and in the cloud, establishing a reference model to overcome current maintenance performance limitations. According to the companies, this would be difficult to achieve with traditional network technologies.