Browse Prior Art Database

Scheduling Latency Driven Throttling Request to LTE L2 Stack

IP.com Disclosure Number: IPCOM000240774D
Publication Date: 2015-Feb-27
Document File: 6 page(s) / 93K

Publishing Venue

The IP.com Prior Art Database

Abstract

Due to exponential growth in mobile devices, deployment of small cell based eNodeB endpoints is increasing in the cellular network. Mobile Network Operators (MNOs) are under pressure to reduce the cost of the network which in turn demands cost reduction in each node of the network. Due to this pressure, low cost eNodeB SoCs (System On Chip) are being designed that demand hosting of LTE stack software on the General Purpose Processors (GPP). LTE stack execution is extremely deadline-sensitive and failure in meeting deadlines result in system failure. Since these GPPs also host other non-LTE software, it becomes a challenge for LTE stack to meet LTE standards defined deadlines if the LTE stack is not scheduled timely. This paper describes how a minor enhancement in the SoC hardware provides a delay hint to the LTE stack. And using this hint, LTE stack takes some preventive actions to complete its job and avoids any potential deadline miss.

This text was extracted from a Microsoft Word document.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 35% of the total text.

Scheduling Latency Driven Throttling Request to LTE L2 Stack

Abstract

Due to exponential growth in mobile devices, deployment of small cell based eNodeB endpoints is increasing in the cellular network. Mobile Network Operators (MNOs) are under pressure to reduce the cost of the network which in turn demands cost reduction in each node of the network. Due to this pressure, low cost eNodeB SoCs (System On Chip) are being designed that demand hosting of LTE stack software on the General Purpose Processors (GPP). LTE stack execution is extremely deadline-sensitive and failure in meeting deadlines result in system failure. Since these GPPs also host other non-LTE software, it becomes a challenge for LTE stack to meet LTE standards defined deadlines if the LTE stack is not scheduled timely. This paper describes how a minor enhancement in the SoC hardware provides a delay hint to the LTE stack. And using this hint, LTE stack takes some preventive actions to complete its job and avoids any potential deadline miss.

Keywords

GPP: General Purpose Processor, L2: LTE Layer2 stack, LTE: Long Term Evolution; MNO: Mobile Network Operator, RTOS/OS: Operating System Software, RT: Real Time, SoC: System on Chip

General Introduction

Constantly decreasing margins are forcing MNOs to look for cost reductions. Due to the proliferation of small-cells (Pico, Femto), eNodeB is one such component which attracts a lot of pressure for cost-reduction as even a minor cost-saving in eNodeB results in a big overall saving due to deployment of eNodeB at mass-scale. Eventually that burden comes to the solution providers and finally to the hardware & software designers of the network components. To reduce the cost of the SoC, the system designers are coming up with single GPP based solutions. In such solution, the same GPP executes time-sensitive LTE stack and other non time-sensitive general functions like operating systems, device drivers of all the devices, management-software.

Background of the Problem

Any operating system suffers from latencies. Latency here refers to delay incurred by the OS (operating system software) in scheduling a user task. In LTE eNodeB, every 1 milli-second (1ms) the LTE Layer2 task needs to start and complete certain workload. The 1ms duration is also called as subframe-duration. The 1ms timer indication is generally implemented through a hardware-interrupt. Once the hardware asserts the 1ms interrupt, OS is required to schedule the LTE Layer2 stack at earliest otherwise the LTE standard defined deadlines won't be met by the Layer2. For example, please refer to Fig 1 below. Suppose for a given implementation and throughput use case, the LTE Layer2 needs 900 µsec of CPU time to complete its workload in a subframe-window. Once the 1ms hardware interrupt asserts, the OS incurs a latency[1] of 70µs and then LTE L2 task is scheduled. L2 task consumes 900µs and then it relinquishes the CPU. Afterwards, non-real-time (non RT) tasks are scheduled for 3...