Huawei Shares Research on Adaptive Randomized Parallelism Theory for Application-Driven Network Control Systems at ICCCN 2016
[Hawaii, US, August 11, 2016] Gong Zhang, Director of the Future Network Theory Laboratory under Huawei's 2012 Laboratories, gave a plenary speech with the theme of ‘Application-driven Network: From Network Engineering to Network Science’, at the International Conference on Computer Communication and Networks 2016 (ICCCN). During the conference, he also shared Huawei's latest research achievements into Adaptive Randomized Parallelism (ARP) theory.
ARP is a crucial element of the Application-driven Network (ADN) and research results reveal that ARP can measure network statuses in real time at an extremely low network measurement cost. This theory marks a new milestone for intelligent networks, and is applicable to both networks and large-scale distributed systems, such as cloud computing and cloud storage systems, which can become more automated and intelligent with the help of ARP.
Within the ADN architecture, an intelligent control system – which functions like the human neural systems that manage people's activities – is crucial to ADN control. The smart control system is built upon a network measurement system that is able to measure the communications network environment and then reschedule network resources in real time. The network measurement system can also learn from network changes and apply what it has learned in future decisions on networks.
The ARP theory is a research breakthrough made by Huawei's Future Network Theory Laboratory using a distributed application measurement method to define the relationship between the precision and efficiency of network data sketch. This theory can be used to abstract data planes and cost-effectively measure different network statuses. More importantly, ARP can ensure highly precise and efficient measurements of network statuses when network traffic bursts unexpectedly. In addition, ARP supports various intelligent network services, such as network anomaly detection, traffic engineering, and network tomography.
Huawei's Future Network Theory Laboratory has also designed an ADN-based lightweight network status measurement algorithm and system. According to existing industry research results, this network status measurement algorithm reduces the memory overhead of data planes by 50 to 100 times, increases the computing speed by three-fold, and improves the precision rate of network status measurements by 40%. This algorithm has eliminated the theoretical limitations on the cost and efficiency of traditional network measurement models.
As Huawei's ADN research leader, Gong Zhang said, "Network measurement systems lay the groundwork for intelligent control in future networks. Supported by the new ARP theory, network measurement systems on ADNs can more efficiently measure network statuses in real time, making it possible to realize automated and intelligent network control. An improvement in network measurement by us will be a great leap for intelligent networks."