"Heavy" mobile network performance Disclosure (on)--mobile network components in detail

Brief introduction

Over the past few years we have made significant progress in mobile cellular network performance. However, due to the expansion of network latency, many applications have not been improved.

Latency has long been a constraint on mobile networks. Although some progress has been made in recent years, the reduction in network delays has not been keeping pace with the delayed growth rate. It is this misalignment that causes latency, not throughput, to be the biggest factor affecting network performance.

This article logically consists of two chapters. The first section discusses the details of the delay that causes the mobile network, and the second part introduces the software technology to improve the latency of network performance reduction.

What are you waiting for?

Latency represents the time required for a packet to pass through one or a series of networks. Due to a number of factors, mobile networks can have higher latencies that already exist in most network traffic, including network types (such as hspa+ and LTE), carriers (at and/or Verizon), or environmental factors such as standing, driving, geography, and different points of time during the day. Therefore, it is difficult to accurately estimate the latency of the mobile network, but we can see that its latency ranges from dozens of to hundreds of milliseconds.

Round-Trip Time (RTT) is a way to test latency by testing the return time of a packet from the source server to the destination. And the size of the RTT has a significant impact on most network performance. This reason can be explained by the table tennis movement.

In the usual table tennis, the time it takes to move between the players in the table tennis game is hardly noticeable. However, the farther the athletes stand, the longer they wait for the ball, and they cannot do anything else while waiting. Athletes at the normal distance, 5 minutes of table tennis match, if the distance to 1000 feet, then it will take a few hours to complete, although this sounds more ridiculous. However, if you think of the source server and the destination server as the two athletes in table tennis, the round trip time represents the time between two athletes, then you will begin to understand this problem.

Part of the general operation of most network protocols is like playing table tennis. These "out-of-the-box", if you prefer, are the two-way exchange of information (such as TCP) that is required to establish and maintain a network link session (such as HTTP) or to perform a service request (such as HTTPS). These messages have little or no data transfer during the exchange process and the network bandwidth is largely unused. As a result, latency is largely not fully utilized, which also results in a delay of at least one RTT for each information exchange, which constantly accumulates the attention that leads to a significant performance impact.

Imagine that an HTTP request to download 10K will have 4 exchanges of information. If each RTT is 100 milliseconds (which is very reasonable for the mobile network) and all 4 interchanges are taken into account, then the throughput is 10k/400ms or 25k/s.

Note that the above example is completely not related to bandwidth-no matter how fast the network results are the same, it is 25k/s. However, the performance of the above operations, or any similar operation, can be improved by a simple, clear strategy: Avoid the network client and the server to exchange information.

Mobile cellular networks

The next step is to introduce a simple introduction to the components and conventions that affect latency in mobile cellular networks.

Mobile cellular networks are representatives of a number of interconnected components with highly specialized functions. Each of these components has an impact on latency, but the individual components have varying degrees of impact. And in the mobile network alone, such as the management of radio resources, has become one of the factors of mobile cellular network delay.

Figure 1: Mobile cellular network components

Baseband processor

The majority of mobile devices are actually two very complex computers. The application processor is responsible for managing operating systems and applications, similar to computers and laptops, and the baseband processor is responsible for all wireless network functions, equivalent to a computer modem, except that the modem uses radio waves rather than telephone lines.

The baseband processor is a consistent, but often delayed, resource. The high-speed wireless network is an incredibly complex thing, and the complex signal processing it requires leads to a fixed, inevitable delay for most network traffic in microseconds to milliseconds.

Cellular Base Station

The cellular base station, which is synonymous with the transceiver Base station or the transmitter tower, acts as an access point for the mobile network. The responsibility of cellular base stations is to provide an area of network coverage, also known as Honeycomb.

Cellular base stations, like mobile devices, deal with complex high-speed wireless networks, with the same mostly insignificant latency, but a cellular base station must serve hundreds of thousands of mobile devices at the same time, and the difference in system load will result in different throughput and latency. At the same time, slow, unreliable network performance tends to limit the processing delivery rate of cellular base stations in crowded public activities.

The latest generation of mobile networks has expanded the role of cellular base stations, including the management of its mobile devices, many of the previously infinite Network controller functions have been given to the cellular base station processing, such as: network registration and transmission scheduling. The reason for doing so is explained later in this chapter that this role transformation has greatly improved the latency of the next generation of mobile cellular networks.

Backhaul Network

The Backhaul network is a dedicated wide area network (WAN) connection between the cellular base station, the controller of the base station, and the core network. Backhaul networks have been and continue to be notorious for delays.

A backhaul network is a transition from a classic circuit or from a frame-based old mobile network (such as a gsm,ev-do) transmission protocol. This protocol has a delay, mainly because of its synchronization characteristics, whereas a logical connection represents a channel that simply accepts or sends data in a short, pre-allocated period of time. In contrast, the latest generation of mobile networks uses IP packet-switched backhaul networks and supports asynchronous data transfer. This switchover greatly reduces the return delay.

The bandwidth constraints of physical facilities have always been a bottleneck. Many backhaul are not designed to handle peak traffic loads, so modern high-speed mobile networks exhibit significant latency and throughput differences in the event of network congestion. While operators are trying to upgrade these networks as quickly as possible, the component remains a weakness in many network infrastructures.

Wireless network Controller

In general, wireless network controllers manage neighboring cellular base stations and the mobile devices they serve.

The wireless network controller coordinates the mobile device directly through a message-based management scheme called signaling. Because of the topology of the mobile network, all messages for mobile devices and controllers must be sent over a high-latency backhaul network. This itself is not ideal, but due to many network operations, such as network registration and transport scheduling, this need to back and forth multiple exchanges of information operations to make the delay worse; Typically, an infinite network controller is also known for this reason as an important factor in latency.

As mentioned earlier, the controller in the latest generation of mobile networks out of the responsibility of equipment management; So many tasks are now handled by the cellular base station itself. This design also determines that the backhaul network latency in most networks is eliminated.

Core network

The core is the gateway between the operator's internal network and the public network, and it is here that operators use inline network devices to implement network service quality policies or bandwidth metering. As a rule, any interception of network traffic will incur a delay. And this delay is prevalent in the real situation, but it is also important to note.

Power protection

One of the most important sources of mobile network latency is directly related to the limit on mobile phone battery usage.

The power consumption of the high-speed mobile device's network station is 3 watts; This power consumption is so large that the iphone5 battery can only support one hours, and that's why mobile devices seize every opportunity to remove or reduce the radio path. This approach not only prolongs battery life but also delays the activation of the radio route, providing power to the wireless circuitry only when the data is received or transmitted.

All mobile phone networks use standard formalized wireless resource management (RRM) programs to conserve energy. Most RRM conventions define 3 states-activation, idle, unlinked-each state represents a tradeoff between some startup latency and power consumption.

Figure 2: Wireless resource management state transitions

Active (Active)

The activation status represents the data perhaps with the smallest delay in telling the transmission or receiving.

This status can be very large even in idle situations. In this state, the network is idle for a short time, usually less than 1 seconds, and the departure state is converted to an idle state. This impact on performance is noteworthy: prolonged pauses in network processing can cause the device to switch back and forth between activation and idle state, which can also lead to increased latency.

Free (idle)

Idle state is a tradeoff between low-power usage and proper lazy loading.

In this state, the mobile device's network is still connected, unable to send or receive data, but can receive network requests to satisfy the need to transition to the activation state (such as input data). After a reasonable period of time, if the network is still inactive for one minute or less, the mobile device switches to the disconnected state.

Idle state can cause delays in two ways. First, it takes some time to re-provide the energy and synchronous analog circuitry for the infinite communication device, and secondly, to save more power, the wireless communication device is merely a intermittent monitoring of the network notification, and the response to the notification is slightly delayed.

Not Connected (disconnected)

The non-connected state consumes the least power but has the highest startup latency.

In this state, the device's network is not connected and the wireless device is inactive. The wireless communication device is activated and listens to the arrival of the network request through a special broadcast channel , but this activation is not very frequent .

The latency of the non-connected delay is the same as that of the idle state, but the difference is that the non-connected state also counts the delay of reconnecting the network. Connecting to a mobile network is a complex process that includes multiple back-and-forth information exchanges (that is, signaling). Restoring a network connection takes at least hundreds of milliseconds, and the link time is often a matter of seconds.

(To be continued, the next chapter is more exciting ...) ）

1. This article is translated by the programmer architecture

2. This document is translated from the performance of Open Source software | Secrets of Mobile Network performance

3. Reprint Please be sure to indicate this article from : Programmer Architecture (No.:archleaner )

4. More articles please scan the code:

"Heavy" mobile network performance Disclosure (on)--mobile network components in detail