KR20200080416A - Data processing apparatus for mobile edge computing - Google Patents

Abstract

The present invention relates to a data processing device for mobile edge computing. According to the present invention, an IT-based server is operated by the technological concept of cloud computing being applied to a network edge in the process of super multi-view content processing. As a result, massive ultra-realistic content can be effectively processed so that personalized interaction services can be provided on a hologram/super multi-view basis anytime and anywhere. The data processing device includes a data plane library and an NIC driver executing packet processing with the data plane library. The data plane library includes: a memory manager allocating a pool of objects in a memory; a buffer manager pre-allocating and managing fixed-length buffers stored in a memory pool in order to reduce the buffer allocation/release time of an OS; a queue manager realizing a secure lockless queue in order to avoid unnecessary latency when different software elements process network packets in parallel; a flow classification generating hash information on a network packet header using Intel streaming SIMD extension technology so that the network packets are allocated to the same flow and processed at high speed; and a poll mode driver increasing the packet pipeline speed using synchronous poll mode interrupt processing routine.

Description

Data processing apparatus for mobile edge computing

The present invention relates to a data processing device for mobile edge computing, and in particular, in the ultra-multi-point content processing process, the cloud computing technology concept is applied to a network edge to operate an IT-based server, thereby hologram/second anytime, anywhere It relates to a data processing device for mobile edge computing that effectively processes large-capacity ultra-realistic content to provide individual interaction services based on multi-view.

Mobile edge computing (MEC) refers to a technology that processes a user's data by placing a server close to a user who intends to use a communication service. This means that it is spread like a fog so that users can access the data transmitted by the mobile network at any time. It is also called fog computing.

When MEC is applied, data transmission time is drastically shortened and customized service is possible. For example, when the MEC is applied to the base station around the highway, the procedure for sending to the central server can be omitted, so that various traffic information can be directly transmitted to the user vehicle. In other words, traffic accident information that occurred in Busan can be quickly distributed by storing it on a server installed in a nearby area without processing it in Seoul.

Because of these characteristics, MEC is attracting attention as a core technology for 5G networks that will be commercialized in 2020. When MEC is applied to 5G, ultra-low latency can be achieved because devices connected to the network do not go through a central server, and servers adjacent to the base station process data in real time.

Conventionally proposed technologies for mobile edge computing are disclosed in <Patent Document 1> to <Patent Document 12>.

In the prior art disclosed in <Patent Document 1>, a cache server located at a network edge such as a point of presence (POP) adjacent to a user device receives media content provided from a content provider, and POP is a transcoding service. And transcode the received media content. At this time, based on the request received from the user device, for example, the transcoded media content is provided to the user device as streaming based on information such as an application, an OS, and a browser type.

In addition, in the prior art disclosed in <Patent Document 2>, the multimedia edge cloud (MEC) of the cloud edge stage directly communicates with the user device, and evaluates service level quality, individual capabilities of the user device, and the like, and provides multimedia data suitable for the multimedia type. It provides a load balancing server that provides a user device, and also provides a handover function to a second MEC according to the movement of the user device.

Further, in the prior art disclosed in <Patent Document 3>, each computing device communicates indirectly with each other through cloud-based resources, is configured without direct communication, and queries based on resource usage for cloud edge topologies. Provides a technique for deciding on which computing device to run operators of data.

In addition, in the prior art disclosed in <Patent Document 4>, the "content processor" located at the edge of a computer network stores content data, processes the stored content data according to attributes of a user request, and a plurality of "content processors", That is, by providing computing resources among edge computers, a technology provided with a processor that personalizes a service transmitted to a computer network edge is provided.

In addition, the prior art disclosed in <Patent Document 5> is composed of a plurality of servers for distributing digital multimedia streaming content, and the data center further includes a database having a plurality of content records, and each database stored in the data center. Content records provide a link between content and one or more intermediate points of presence (POP).

In addition, the prior art disclosed in <Patent Document 6> is an apparatus for activating a mobile edge service (MES), a GGSN interface configured to receive at least one of wireless applications, a switch for determining transmission of a packet, and at least one packet. It includes a core service layer that utilizes requests associated with packets to apply to the mobile edge service of and a protocol stack for controlling communication between services. As a technology for a system and a method instructing a mobile communication network to allow a mobile edge service (MES), a content portal and a value-added service can be provided to maximize a user's value.

In addition, the prior art disclosed in <Patent Document 7> provides a technique for an edge node that receives content from a Network Operations Center (NOC) through a satellite broadcast content distribution network and distributes it to the last mile provider. Media servers are connected to public and private VLANs, and media servers, private VLANs and public VLANs exist in one computer.

In addition, the prior art disclosed in <Patent Document 8> prevents the connection request from being processed in the enterprise CDN area in order to prevent the connection request transmitted from outside the enterprise and associated with a part of intranet content from accessing the intranet content from the enterprise CDN area. . By building the same software and ICDN in the company's internal system, it is possible to efficiently provide Internet content to the company.

In addition, the prior art disclosed in <Patent Document 9> assumes that the name included in the host header is not recognized before determining whether to service a request including an unknown host header using DNS. Then, a list of DNS CNAMEs is obtained from the edge server, and the CNAME chain returned to the edge server is analyzed to determine the response method.

In addition, in the prior art disclosed in <Patent Document 10>, allocating content from a network edge cache to a user device group, identifying a difference between content to be delivered to the user device and content to be delivered to the user device group, identification And selecting and transmitting a part of the content to be transmitted to the user device based on the one or more differences, and receiving the content from the user group device and pre-delivering the unreceived portion through the difference in the content delivered to the user device. It provides the technology to do.

In addition, the prior art disclosed in <Patent Document 11> relates to a method for streaming a media object, determining a prefix size, determining a transmission rate based on the size of the prefix for streaming at a bit rate, necessary And providing bandwidth between servers to achieve transfer rates.

In addition, the prior art disclosed in <Patent Document 12> constructs an Overlay Network using a plurality of edge servers (ESs), allocates service resources to each ES, and handles service resources requested by the user device. And selecting and supplying a second ES to be performed.

US registered patent US 9088634 (Registered on July 21, 2015) (Dynamic media transcoding at network edge) US registered patent US 9088577 (registered on July 21, 2015) (Multimedia aware cloud for mobile device computing) US registered patent US 9088344 (2015.08.04. registered) (Cloud-edge topologies) US registered patent US 8243596 (registered on August 14, 2012) (Distributing intelligence across networks) European Registered Patent 1269714 (Registered on August 30, 2006) (METHOD AND DEVICE FOR DISTRIBUTED CACHING) US registered patent US 6891842 (registered on May 10, 2005) (System and method for enabling mobile edge services) US registered patent US 7237071 (registered on June 26, 2007) (Micronode in a satellite based content delivery system) U.S. registered patent US 7600025 (Registered on October 6, 2009) (Extending an internet content delivery network into an enterprise) US Patent Publication US 2013-0191499 (published July 25, 2013) (Multi-domain configuration handling in an edge network server) European Patent No. 2973352 (published on January 20, 2016) (PRE-DELIVERY OF CONTENT TO DEVICES) US Patent Publication No. 2005-0071496 (published on March 31, 2005) (PRE-DELIVERY OF CONTENT TO DEVICES) European Patent Publication No. 2063598 (published on May 27, 2009) (PRE-DELIVERY OF CONTENT TO DEVICES)

However, such prior arts provide various data processing methods for mobile edge computing, but there are limitations in high performance packet processing.

Therefore, the present invention has been proposed to solve all the problems encountered in the high-performance packet processing and the prior art required by the general mobile edge computing as described above, and the concept of cloud computing technology in the process of ultra-multi-point content network edge (edge) It provides a data processing device for mobile edge computing that effectively handles large-capacity ultra-realistic content to provide personal interaction service based on hologram/ultra-view point by operating IT-based server by applying to There is a purpose.

In order to achieve the above object, a data processing apparatus for mobile edge computing according to the present invention includes: a data plan library; It includes a NIC driver for performing packet processing using the data plan library,

The data plan library is characterized by including a memory manager, a buffer manager, a queue manager, flow classification, and a poll mode driver.

According to the present invention, the cloud computing technology concept is applied to the network edge in the process of ultra-multi-point content processing to operate the IT-based server to provide individual interaction services based on the hologram/ultra-multipoint point anytime, anywhere. It has the advantage of being able to effectively process ultra-realistic content.

1 is an overall configuration diagram of a data processing apparatus for mobile edge computing according to the present invention,
FIG. 2 is a block diagram of an embodiment of the Poll Mode Driver of FIG. 1.

Hereinafter, a data processing apparatus for mobile edge computing according to a preferred embodiment of the present invention will be described in detail with reference to the accompanying drawings.

1 is an overall configuration diagram of a data processing apparatus for mobile edge computing according to a preferred embodiment of the present invention, including a data plan library (Data Plane Library), a NIC driver performing packet processing using the data plan library The data plan library includes a memory manager, a buffer manager, a queue manager, flow classification, and a poll mode driver.

DPDK stands for Data Plane Development Kit and is a system for optimizing high-performance packet processing based on Intel architecture. DPDK consists of a set of data plane libraries and a set of NIC drivers for packet processing optimization.

As shown in FIG. 1, the biggest feature of the DPDK is that, unlike the existing Linux-based packet processing process, the packet processing process is not performed in the kernel, but the application (Customer Application) uses the Intel DPDK library API of the User Space. It uses EAL (Environment Abstraction Layer) to provide a path to directly access the NIC through the Linux kernel.

In addition, DPDK provides a NIC driver optimized for packet processing as well as the Data Plane API of the Linux User Space.

The library, which is a component of the DPDK, can be largely divided into Memory Manager, Buffer Manager, Queue Manager, Flow classification, and Poll mode Driver. The roles of each library are as follows.

The memory manager performs a function of allocating a pool of objects in memory. Each object pool is created in a huge page memory space and uses a ring to store usable objects. This memory manager provides an alignment help function to ensure that objects are evenly stored in the DRAM Channel and that the objects are padded.

The buffer manager manages pre-allocated fixed-length buffers stored in the memory pool in order to reduce the time that the OS spends allocating and releasing buffers.

Instead of using the existing spinlock, the Queue Manager implements a secure lockless queue to prevent unnecessary latency when different software elements process network packets in parallel.

Flow Classification effectively generates hash information for the network packet header using Intel streaming SIMD extension technology, so that network packets are allocated to the same flow and processed at high speed.

The Poll Mode Driver speeds up the packet pipeline by using an optimized synchronous poll mode interrupt handling routine instead of processing network packets based on the existing asynchronous interrupt.

2 shows a specific embodiment architecture of the Poll Mode Driver.

In FIG. 2, rte_timer performs a Timer function. Based on a timer interface provided by EAL, a timer service is provided in a DPDK execution unit. rte_malloc performs memory allocation for hugepages, rte_mempool allocates a pool of objects in memory, and rte_eal + libc supports application loading, memory allocation, time interface, PCI access and loading. Perform. In addition, rte_mbuf performs a function of supporting a buffer of IP traffic and message data packets, and rte_ring stores a fixed object or fixes a FIFO that allows communication between cores.

The invention made by the present inventors has been described in detail according to the above embodiment, but the present invention is not limited to the above embodiment and can be changed in various ways without departing from the gist thereof. It is obvious to those who have it.

Claims (2)

Data Plane Library; And
And a NIC driver that performs packet processing using the data plan library.
The method according to claim 1, The data plan library is a memory manager for performing a function of allocating a pool of objects in memory (Memory Manager); A buffer manager for pre-allocating and managing fixed-length buffers stored in a memory pool in order to reduce the time that the OS spends in allocating and freeing buffers; A queue manager that implements a secure lockless queue to prevent unnecessary latency when different software elements process network packets in parallel; Flow classification to generate hash information for a network packet header using Intel streaming SIMD extension technology so that network packets are allocated to the same flow and processed at high speed; And a Poll Mode Driver for increasing the packet pipeline speed using a synchronous poll mode interrupt processing routine.

Images