|
Mike Cooper是CommScope公司的一名现场应用工程师 由DCD中国授权DKV编译并发表于DeepKnowledge微信公众号。 根据2018年Equinix的全球互联指数，全球连接带宽预计将以48%的年复合增长率增长，达到8200 MMB/秒的带宽。到2021年，为了解决延迟问题，与网络提供商互连的企业将占所有带宽的66%。同时企业以98%的年复合增长率，增长最快的部分是企业互连云和IT服务，以解决网络复杂度的问题。 在下面的一个视图中，很容易理解为什么互联正在爆炸。随着新兴技术和应用程序的延迟需求的减少，强调冗余东西连接的网状拓扑成为支持所需超可靠、低延迟性能的最佳方式。 优雅的拓扑结构从3万英尺的高度看，一个丰富的互连拓扑显得非常优雅。在数据中心的底层，也就是建立物理连接的地方，视图可能会发生巨大的变化。随着进入数据中心的数据量持续飙升，数据中心内的光纤呈指数级增长，管理数千股光纤是一项持续的挑战。 为了管理它们的光纤，大多数数据中心通常混合使用直接连接和互连线缆。顾名思义，直接连接在机架之间点对点运行。数据中心互连——不要与前面提到的网络互连混淆——将线路一直连接到显示面板。对于大型项目，这种策略可能会变得难以管理，因为线路往往会变长，线缆通道也会变得更加拥挤。一旦纤维束的数量开始超过2000或3000股，天平开始倾斜，有利于交叉连接补丁策略。 数据中心交叉链接策略交叉连接布线设备提供了多种好处。最重要的是，它涉及一个专用的修补区域，使移动、添加和更改更易于管理。配线区隔离了任务关键型有源设备，因此在维修配线架时，中断带电电路的风险较小。此外，与运营商和云供应商的交叉连接还可以节省资金，提高可靠性，并为您的网络增加多功能性。 同时，交叉连接需要更多的线缆，交叉连接拓扑中的跳线激增成为一个棘手问题。随着光纤数量的增加，一个良好的线缆管理策略势呼之欲出。 线缆管理策略的重要性在网络的演进和迁移需求的规划和细节设计过程中，许多数据中心管理人员都是非常认真和勤奋。在专注于所需的网络解决方案时，他们往往忽略了对更为常规的明天需求的规划，如移动/添加/更改，这种规划上的欠缺让他们在未来面临问题： 线缆管理策略为光纤设备的管理提供了网络范围的标准，它有助于确保光纤性能，加快移动/添加/更改的速度和准确性，并改进关键指标，如平均解决时间。 精心设计的策略涉及几个关键方面，包括：如何在机柜内和机柜间运行跳线、如何使用线缆槽和其他路径的参数、交叉连接中光纤的标签以及定期“收集”或停用未使用的跳线的最佳做法。 细节中蕴藏着能量与任何运营策略一样，良好的线缆管理策略的价值在于其实际应用的价值。它必须提供每个技术人员都能理解和遵循的工作线路，因此细节的层次等级很重要。 例如，在保持光纤系统的有序和可管理方面，一个常见的问题是如何处理线缆松弛。如果没有储存松弛部分的空间，线缆很快就会溢出线缆通道。因此，线缆管理策略的一个方面可能是指定提供良好线缆容量的接线板和配线架。 另一个经常使整个数据中心的线缆堵塞复杂化的问题是，缺乏规定如何使用架空电缆桥架路径的协议。最佳实践建议将较大的中继电缆与较小的跳线分离：用于较大中继电缆的梯形架，用于跳线的光纤线缆通道。同样，铜和光纤接插线也应该有自己的路径。 这些管理策略不仅旨在保持数据中心电缆设备的可访问性和可服务性，而且还可能对光学性能产生重大影响。一个很好的例子，而且经常性的可能会发生的情况是，当大型干线线缆在为较小的配线光纤设计的线缆通道中运行时，当较大的线缆通过所谓的“瀑布”出口通道时，线缆超过最大允许弯曲半径，从而影响了光学性能。 在制定一个好的线缆管理策略时，魔鬼就在细节上。但是，如果做得好，一个好的电缆管理策略不仅可以保持当前线缆系统的高可用性，还可以提供一个平台，使网络的扩展和升级更快、更可靠。 英文原文 Cable management is key to data center cross connect strategyMike Cooper is a field application engineer at CommScope According to the 2018 Equinix Global Interconnection Index, interconnection bandwidth is expected to grow globally at a 48-percent CAGR to 8,200 terabytes per second of installed bandwidth. By 2021, enterprises interconnecting to network providers in order to address latency will account for 66 percent of all total interconnect bandwidth. With a 98 percent CAGR, the fastest growing segment is enterprises interconnecting to cloud and IT services in order to solve issues of network complexity. In a single visual seen below, it is easy to understand why interconnection is exploding. As latency requirements for emerging technologies and applications shrink, the mesh-type topology with its emphasis on redundant east-west connectivity becomes the best way to support ultra-reliable, low latency performance needed. Elegant topologyWhen viewed from the 30,000-foot level, a topology rich in interconnections appears somewhat elegant. At ground level in the data center, where the physical connections are made, the view can change dramatically. As data volumes coming into the data center continues to soar, the fiber plant inside the data center increases exponentially. Managing thousands of fiber strands is an ongoing challenge. In order to manage their fiber, most data centers typically use a mixture of direct connect and interconnect cabling. As the name implies, a direct connection runs point-to-point between racks. A data center interconnect—not to be confused with the network interconnects mentioned previously—routes patch cords to a presentation panel. For large projects, this strategy can become difficult to manage as patch cords tend to become longer and cable pathways grow more congested. Once the number of fiber strands starts to exceed two or three thousand, the scales begin tipping in favor of a cross connect patching strategy. Cross connects in the data centerA cross connect cabling plant offers a variety of benefits. Most importantly it involves a dedicated patching area that makes moves, adds and changes easier to manage. The patching area isolates mission critical active equipment, so there is less risk of disrupting live circuits while servicing the patch panels. Additionally, cross connects to carriers and cloud providers can also save money, improve reliability, and add versatility to your network. At the same time, cross connects require more cabling. The proliferation of patch cords in a cross connect topology becomes a critical concern. As fiber counts rise, a good cable management strategy is imperative. Importance of a cable management strategyMany data center managers are diligent when it comes to planning and detailing their network’s evolution and migration needs. In focusing on the network solutions needed, they often overlook planning for the more routine Day 2 requirements, such as moves/add/changes. This gap in planning leaves them open to problems down the road: A cable management strategy provides network-wide standards for how the fiber plant is managed. It helps ensure fiber performance, accelerate the speed and accuracy of moves/adds/changes and improve critical metrics such as mean-time-to-resolve. A well-designed strategy addresses key aspects, including: how patch cords should be run within and between cabinets, parameters governing how cable trays and other pathways are to be used, labeling of optical fiber in the cross connects and best practices for periodically “harvesting” or decommissioning unused cords. Power is in the detailsAs with any operational strategy, the value of a good cable management strategy is in its practical application. It must provide a working blueprint that every tech can understand and follow. The level of detail is important.As with any operational strategy, the value of a good cable management strategy is in its practical application. It must provide a working blueprint that every tech can understand and follow. The level of detail is important. Another issue that often compounds congestion throughout the data center is the lack of protocols dictating how overhead cable pathways are used. Best practices suggest segregating larger trunk cables from smaller patch cords: ladder racks for large trunk cables, fiber raceways for the patch cords. Likewise, copper and fiber patch cords should have their own pathways. These management tactics are not only intended to keep the data centers cable plant accessible and easier to service, it can have a significant effect on optical performance as well. A good example is what can happen when large trunk cables are run in raceways designed for smaller patching fibers. As the larger cables exit the raceway via so-called “waterfalls,” it is not uncommon for the cable to exceed the maximum allowable bend radius, affecting optical performance. In developing a good cable management strategy, the devil is in the details. When done right, however, a good cable management strategy not only keeps the current cable plant highly serviceable, it also can provide a template that makes expansion and upgrades to the network faster and more reliable. |
|
|
