对等商务

对等–混合系统 (Peer-to-Peer – Hybrid Systems)

Centralized systems are simple to deploy and manage, but they are a potential bottleneck, since the central server has limited capacity and can not withstand the increased demand. On the other hand, decentralized systems are scalable and robust, but it takes some implementation complexity, especially on issues of fault tolerance and resource discovery. Many distributed systems combine features of both architectures, part of the system in the traditional client-server model and part peer-to-peer.

集中式系统易于部署和管理，但由于中央服务器容量有限且无法承受需求的增长，因此它们是潜在的瓶颈。 另一方面，分散式系统具有可伸缩性和健壮性，但它会带来一些实现复杂性，尤其是在容错和资源发现方面。 许多分布式系统结合了两种体系结构的功能，传统客户端-服务器模型中的部分系统以及对等网络的一部分。

Hybrid structures are deployed in distributed systems remarkably collaborative. The main concern in many of these systems is how to join the system, for which often a traditional client-server model is adopted. Once the node joins the system, it can use a fully decentralized scheme for collaboration. An example of a server that uses this approach is BitTorrent.

混合结构部署在非常协作的分布式系统中。 许多这些系统中的主要关注点是如何加入系统，为此通常采用传统的客户端-服务器模型。 一旦节点加入系统，它就可以使用完全分散的方案进行协作。 使用这种方法的服务器的一个示例是BitTorrent。

For a user of an instant messaging service, the application appears to peer-to-peer to send data directly to the friend being contacted. But all the instant communication services have some sort of server behind that facilitates communication between the nodes. For example, the server maintains an association between the user name and your IP address, record messages when the user is offline, and routes messages to users who are behind firewalls. A fully decentralized system of instantaneous communication would not work in today’s Internet, but there are great advantages in scalability, allow the client-client communication when possible. Thus, most systems use a centralized directory scheme while the light is distributed.

对于即时消息传递服务的用户，该应用程序似乎是点对点的，直接将数据发送给正在联系的朋友。 但是所有即时通信服务背后都有某种服务器，可促进节点之间的通信。 例如，服务器维护用户名和IP地址之间的关联，在用户脱机时记录消息，并将消息路由到位于防火墙后面的用户。 完全分散的即时通信系统无法在当今的Internet中运行，但是可伸缩性具有很大的优势，并在可能时允许客户端与客户端之间的通信。 因此，大多数系统在灯光分布时使用集中式目录方案。

对等应用对等的要求 (Requirements for an Application Peer to peer)

To work effectively, peer-to-peer should be concerned with the following non-functional requirements:

为了有效地工作，对等应关注以下非功能性要求：

全局可扩展性 (Global Scalability)

One of the goals of peer-to-peer is to exploit the hardware resources of a large number of hosts connected to the Internet. Thus, these applications should be designed so as to support access to millions of objects in tens or hundreds of thousands of hosts.

对等网络的目标之一是利用连接到Internet的大量主机的硬件资源。 因此，应将这些应用程序设计为支持访问数以万计的主机中的数百万个对象。

优化相邻节点之间的本地交互 (Optimization of local interactions between neighboring nodes)

A “network distance” between nodes that interact have a substantial impact on the latency of individual interactions, for example, clients requesting access to resources. The load of network traffic is also impacted by this. Applications should put the resources close to us who use them more.

进行交互的节点之间的“网络距离”对各个交互(例如，客户端请求访问资源)的等待时间具有重大影响。 网络流量的负载也受此影响。 应用程序应该使资源靠近我们，而资源的使用更多。

可用性 (Availability)

Most peer-to-peer networks are composed of host computers that are free to join or leave the system at any time. In addition, the network segments used in peer-to-peer networks are not managed by any authority, nor have guarantees of service quality. A major challenge for peer-to-peer is to provide a reliable system despite these facts. When new hosts are added, they should be integrated into the system and the load should be redistributed to explore these new features. When they leave the system voluntarily or involuntarily, the system should detect their departure, and redistribute its loads and resources.

大多数对等网络由主机组成，这些主机可以随时自由加入或离开系统。 另外，对等网络中使用的网段不由任何机构管理，也不保证服务质量。 尽管存在这些事实，对等网络的主要挑战是提供可靠的系统。 添加新主机后，应将它们集成到系统中，并应重新分配负载以探索这些新功能。 当他们自愿或非自愿离开系统时，系统应检测到他们的离开，并重新分配其负载和资源。

数据安全 (Data security)

In systems of global scale with participants from different backgrounds, trust must be built with the use of authentication and encryption mechanisms to ensure privacy of data and information.

在具有来自不同背景的参与者的全球规模的系统中，必须使用身份验证和加密机制来建立信任，以确保数据和信息的私密性。

匿名，拒绝和拒绝审查 (Anonymity, deniability and resistance to censorship)

Anonymity is a legitimate concern in many situations that require resistance to censorship. A related requirement is that hosts that store data must be able to plausibly deny responsibility for the possession and supply them. The use of a large number of hosts in peer-to-peer can be useful in achieving these properties.

在许多需要抵制审查的情况下，匿名是一个合理的问题。 一个相关的要求是，存储数据的主机必须能够合理地否认拥有并提供数据的责任。 在点对点中使用大量主机可以实现这些属性。

P2P网络结构化和非结构化 (P2P networks structured and unstructured)

Overlay Network

叠加网络

An important concept of the paradigm is the P2P overlay network, or overlay network. In the overlay network, nodes are formed by the processes and the links are represented by the possible communication channels (which are typically TCP connections). In general, a process can not communicate directly with other arbitrary process, but can only send messages through the communication channels available. Accordingly, each node has a neighbor set, which in turn possess other sets of neighbors. The figure below illustrates the overlay network on the network layer (e.g., internet). Note that two nodes that are neighbors in the overlay network are not necessarily neighbors in the physical network.

范式的重要概念是P2P覆盖网络或覆盖网络。 在覆盖网络中，节点由进程形成，链路由可能的通信通道(通常为TCP连接)表示。 通常，一个进程不能与其他任意进程直接通信，而只能通过可用的通信通道发送消息。 因此，每个节点具有一个邻居集合，该邻居集合又拥有其他邻居集合。 下图说明了网络层(例如Internet)上的覆盖网络。 注意，在覆盖网络中作为邻居的两个节点不一定是物理网络中的邻居。

The location of nodes and objects in the overlay network is accomplished through a distributed routing algorithm. This algorithm is implemented on the application layer, having no connection with the routing implemented by routers in the network layer. Through this algorithm that the client requests are routed to a host containing the object for which the request is addressed. The objects of interest are placed and relocated on any node in the network without the involvement of the client.

覆盖网络中节点和对象的位置是通过分布式路由算法完成的。 该算法在应用层上实现，与网络层中的路由器实现的路由没有任何关系。 通过这种算法，客户端请求被路由到包含请求所针对的对象的主机。 在客户端不参与的情况下，将感兴趣的对象放置并重定位到网络中的任何节点上。

The routing ensures that any node can access any object in the overlay network by exploiting the knowledge of each node in the network to locate the target object. Peer-to-peer networks typically store multiple replicas of the same object to ensure availability. Thus, the routing algorithm maintains minimal information possible about the location of all replicas, and sends requests to the node “live” (ie not faulted) closest that keeps a copy of the relevant object. There are basically two types of overlay networks: structured networks and unstructured networks.

路由通过利用网络中每个节点的知识来定位目标对象，从而确保任何节点都可以访问覆盖网络中的任何对象。 对等网络通常存储同一对象的多个副本以确保可用性。 因此，路由算法将有关所有副本位置的信息保持在最低限度，并将请求发送到最接近“活动”(即没有故障)的节点，该节点保留相关对象的副本。 覆盖网络基本上有两种类型：结构化网络和非结构化网络。

Peer-to-Peer Structured

点对点结构化

In a peer-to-peer structured overlay network is constructed through a deterministic procedure. The procedure most often used by far is to organize the processes through a distributed hash table (DHT). In a system based on DHTs, the data given a random key from a large space of identifiers, typically a handle 128 or 160 bits. The network nodes also receive an identifier of the same space. The great challenge in a DHT-based system is to implement an efficient and deterministic schema that maps only to a key item for the identifier of the node responsible for the item you want. From there, you can return the network address of the node responsible for the item you want, you can be contacted directly.

在对等结构化覆盖网络中，是通过确定性过程构建的。 到目前为止，最常用的过程是通过分布式哈希表(DHT)来组织进程。 在基于DHT的系统中，数据从很大的标识符空间(通常为128位或160位)中获得了随机密钥。 网络节点还接收相同空间的标识符。 在基于DHT的系统中，最大的挑战是实现一种高效且确定性的方案，该方案仅映射到关键项，该关键项是负责所需项的节点的标识符。 从那里，您可以返回负责所需项目的节点的网络地址，可以直接与您联系。

Various implementations and protocols were developed based on DHT. Among them, the most popular are Chord. NodeWiz is a service discovery in grids (GIS) that uses a tree structure to organize your nodes in the overlay network. To this end, each node in the tree is responsible for a subspace of the full space of attributes. Thus, you can perform queries efficiently by ranges of values, contrary to what happens in the DHT, where queries are given by exact values.

基于DHT开发了各种实现和协议。 其中，最受欢迎的是Chord。 NodeWiz是网格(GIS)中的服务发现，它使用树结构来组织覆盖网络中的节点。 为此，树中的每个节点负责属性完整空间的子空间。 因此，您可以按值范围有效地执行查询，这与DHT中发生的情况相反，在DHT中，查询由确切值给出。

Peer-to-Peer Non-Structured

对等非结构化

Peer-to-peer unstructured generally based on randomized algorithms to construct the overlay network. The main idea is that each node maintains a list of neighbors, which is constructed more or less randomly. Likewise, it is assumed that the data is placed randomly on nodes. So when a node needs to locate a specific item, the only thing you can do is to flood the network with a search. A disadvantage of this type of search is that queries can not be answered if the client and host are very far apart in network. This happens because the mechanisms that prevent messages to propagate indefinitely in the network (e.g., TTL). Another disadvantage is that mechanisms usually cause great flood of traffic signs, which often makes this kind of slow research.

对等非结构化通常基于随机算法来构建覆盖网络。 主要思想是每个节点都维护一个邻居列表，该列表或多或少是随机构建的。 同样，假定将数据随机放置在节点上。 因此，当节点需要定位特定项目时，您唯一可以做的就是通过搜索使网络泛滥。 这种搜索的缺点是，如果客户端和主机在网络中相距很远，则无法回答查询。 发生这种情况的原因是，阻止消息在网络中无限传播的机制(例如TTL)。 另一个缺点是，机制通常会导致大量交通标志泛滥，这常常使这种研究进展缓慢。

Many peer-to-peer unstructured construct overlay networks that draw from a random graph. The basic model is that each node maintains a list of neighbors c, where, ideally each of these neighbors is a node chosen randomly among all the nodes “alive”. This list of us can be called a partial view. Among the systems that use P2P unstructured architectures can cite the BitTorrent, which builds its overlay on a random basis, the Content Delivery Networks (CDNs), and Gnutella.

从随机图绘制的许多对等非结构化构造覆盖网络。 基本模型是每个节点维护一个邻居列表c，理想情况下，这些邻居中的每个邻居都是在所有“活动”节点中随机选择的节点。 我们的这份清单可以称为局部视图。 在使用P2P非结构化体系的系统中，可以引用BitTorrent，内容分发网络(CDN)和Gnutella，后者是随机构建覆盖的。

Continued…

继续…

翻译自: https://www.eukhost.com/blog/webhosting/peer-to-peer-part-2/

对等商务