400GbE:未来从今天开始起步

2015-08-10 |  作者:戴尔网络事业部以太网传播总负责人 John D’Ambrosia |  来源:独家

摘要在七月召开的最近一次会议上,IEEE P802.3bs 400GbE 工作小组完成了对技术方案的遴选,为接下来制订 IEEE P802.3bs 400GbE 标准草案做好了准备。作为该工作小组的主席,这一进展令我非常激动万分。

以太网联盟主席、戴尔网络事业部以太网传播总负责人 John D’Ambrosia

过去 18 个月以来,以太网的发展突飞猛进,传输速度节节攀升,新的应用和市场不断涌现。这种爆炸式的发展得益于过去 15 年的技术积累,尤其是大量早期的以太网项目所开发出来的技术。例如 IEEE 802.3 以太网工作小组正在开发的 2.5 千兆以太网、5 千兆以太网、25 千兆以太网 (25GbE) 和 40 千兆以太网 (40GbE) 等4 个不同速率的 BASE-T 解决方案,它们所依赖的底层技术基本上都是根据 2006 年获得批准的 10GBASE-T 标准而开发的。再以当前获得业界极大关注的用以实现服务器和交换机之间连接的25GbE 为例,它的研发目标是在电路板印制线、背板、铜线电缆和多模光纤上运行 25 GbE,并充分利用了早期 100 千兆以太网项目所研发的技术。以太网社区在过去确实借鉴了竞争对手的很多技术,但是现在这门技术正在一步步地发掘属于自己的技术瑰宝。

在七月召开的最近一次会议上,IEEE P802.3bs 400GbE 工作小组完成了对技术方案的遴选,为接下来制订 IEEE P802.3bs 400GbE 标准草案做好了准备。作为该工作小组的主席,这一进展令我非常激动万分,因为达成业界共识不是一件容易的事,只有经历过才知道它的艰难。这些决议将从根本上改变电信号和光信号的传输,这不仅会在短期内为整个行业带来两个机遇,还将为未来发展铺平了道路。此次会议还决定把有关以太网未来新速率的讨论付诸行动。

该项目的目标包括开发用于电路板和 400GbE 光模块的 400GbE 电接口,以及在 100米多模光纤 (MMF)、500米单模光纤、2千米单模光纤和 10千米单模光纤 (SMF) 上运行 400GbE。完整解决方案组合包括若干个基于多个 25Gb/s 或 50Gb/s 通道的方案,其中的运行速率主要取决于具体的光纤型号。100米MMF 型号将通过 16 个运行速率为 25Gb/s 的通道解决,即在各个方向上通过 16 条并行光纤以 25Gb/s 的速率传输,同时它也使用基于 16 个 25Gb/s 通道的电接口。该解决方案使用的是当前驱动 25GbE 和 100GbE 的同一 25Gb/s 电信号和光信号传输速度。它基于不归零码 (NRZ) 信号传输这一使用了多年的基本信号传输方式。想象一下基本数字(或称二进制)信号传输方式,这种信号传输由“0”或“1”组成,其中每个值代表一条信息。

它是一套能从根本上改变电信号和光信号传输的是面向单模光纤型号的解决方案组合。我们先来讨论光信号传输。需要着重提醒的是,每种传输型号到最终都会面临绕不开的特殊的技术和经济方面的压力,因此,工作小组在研究如何选择解决方案组合时,不得不为了解决这些不同的压力而作出某些让步。

· 对于 500米的情形,使用并行 SMF 光纤方法的能力一般人认为是可接受的,但有人认为这一能力只是用户希望拥有的,因为并行光纤方法经常用来支持速度较低的break-out)解决方案。因此,工作小组选择了在每个方向上使用 4 根并行光纤的方法,其中每根光纤传输 100Gb/s 的单个光信号。这样,4 根光纤合在一起即可实现 400GbE,并且每个光纤还可以支持 100GbE 。尽管支持 100GbE 并不是该项目的目标,但工作组认识到如果开发出来的解决方案不对其造成阻碍这类实施将具有重大意义。

· 对于 2千米和 10千米的情形,需要采用单个双 工SMF 的方法(+微信关注networkworldweixin),因为使用如此长距离的并行光纤可能会遇到成本较高的问题。对于这些解决方案,将使用单光纤 WDM 的方法来传输 400GbE,该方法使用 8 个光信号 (lambdas),其中每个信号的速度为 50Gb/s。

在这些 SMF 解决方案下,所有的光信号传输将通过 4 个级别的“脉冲幅度调制”(即常说的 PAM4)完成。对于 500米的情形,上面已经提到过,每个光信号都支持 100Gb/s 的运行速率。对于 2千米和 10千米的情形,每个光信号都支持 50Gb/s 的运行速率。因此,投资 4x100Gb/s 并行光纤方法不仅可以省却模块化解决方案的最初成本,也可以为未来可能使用串行 100Gb/s 解决方案打下基础。对 2千米和 10千米的距离使用 8x50Gb/s 光传输方法这一决策,将帮助行业解决需要更高带宽路由器来连接服务供应商网络这一眼前需求。另外,考虑到这些解决方案将使用 50Gb/s 和 100Gb/s 光信号传输,工作小组还决定开发 50Gb/s PAM4 电信号传输,其目的是开发基于 8 通道而非 16 通道的 400Gb/s 接口。

我感觉自己已经看到了以太网为基础的解决方案的光明未来。七年前,当我领导 40GbE 和 100GbE 工作小组的时候,我认为那开启了 25Gb/s 信号传输的起步阶段。如今,我们看到了当时决定的巨大影响力,因为市场已经准备好迎接 10 千兆以太网 (10GbE) 的继任者:25GbE。而对于使用 50Gb/s 信号传输支持 400GbE,我们同样认为这只是未来 50GbE 生态系统的第一步,50GbE 迟早要取代 25GbE。另外,结合单通道和四通道实施的发展历史,在谈到以太网的未来时,人们已经开始讨论有关 200GbE 话题了。

欢迎步入以太网的未来世界!

附:英文原文

400GbE: Tomorrow’s Future Starting Today

By John D’Ambrosia, Chief Ethernet Evangelist, Dell Networking

Over the past 18 months, Ethernet has exploded into an array of new rates, applications, and markets. Fueling this explosion is the wealth of technology developed over the past fifteen years as part of a number of different prior Ethernet projects. For example, four new rate variants (2.5 Gigabit Ethernet, 5 Gigabit Ethernet, 25 Gigabit Ethernet (25GbE) and 40 Gigabit Ethernet (40GbE) of BASE-T solutions are currently being developed within the IEEE 802.3 Ethernet Working Group. All of these solutions are fundamentally based on the underlying technology developed as part of the 10GBASE-T standard that was ratified in 2006. Another example, which is receiving a great amount of attention in the industry right now, is 25GbE for server to switch connections. This effort is targeting 25 GbE operation over circuit board traces, backplanes, copper cables, and multi-mode fiber, and leverages technology developed as part of earlier 100 Gigabit Ethernet projects. While the Ethernet community has unapologetically borrowed from competing technologies in the past, it now finds itself digging into its own treasure trove of technology goodies.

At its last meeting in July, the IEEE P802.3bs 400GbE Task Force completed its selection of technical proposals, which will now enable the creation of the first draft of the standard. As chair of the task force, I am excited, as achieving industry consensus has proven very challenging. These decisions represent a fundamental change in both electrical and optical signaling that will address both near term opportunities for the industry at large, while also laying the basic ground work for future development. These decisions are also setting discussions into motion about future new rates of Ethernet.

The project objectives include the development of 400GbE electrical interfaces on circuit boards, as well as to 400GbE optical modules, operation over 100m of multi-mode fiber (MMF), 500m of single-mode fiber, 2km of single-mode fiber, and 10km of single-mode fiber (SMF). The complete solution set includes schemes based on multiple lanes of either 25Gb/s or 50Gb/s, with the rate of operation dependent on the specific fiber objective. The 100m MMF objective will be solved with 16 lanes of 25Gb/s operation that is transmitted across 16 parallel fibers in each direction, and will also have electrical interfaces that are based on 16 lanes of 25Gb/s. This solution is leveraging the same 25Gb/s electrical and optical signaling that is driving 25GbE and 100GbE today. It is based on non-return-to zero or “NRZ” signaling, which has been a basic form of signaling used for years and years. Think basic digital or binary signaling where you have “ones” or “zeroes” with each value representing a bit of information.

It is the solution set for the single-mode fiber objectives that represents a fundamental change in both electrical and optical signaling. Let’s take a look at the optical signaling first. It is important to note that each reach objective ultimately has different technical and economic pressures to deal with, and the group had to come to terms with these different pressures as it looked to choosing the solution set.

· At 500m the ability to use a parallel SMF fiber approach was deemed acceptable, and some would argue desirable, as parallel fiber approaches are popular to enable break-out solutions for lower speeds. Therefore, the group selected an approach that will use 4 parallel fibers in each direction with each fiber carrying a single optical signal carrying 100Gb/s. Thus, the aggregate 4 fibers will carry 400GbE, with each fiber having the potential ability to also support 100GbE. While supporting 100GbE is outside the objectives of the project, the task force recognizes the importance in developing a solution that won’t preclude such implementations.

· For 2km and 10km a single duplex SMF approach is necessary, as it can be cost prohibitive to run such long runs of parallel fiber. For these solutions a single fiber WDM approach that uses 8 optical signals (lambdas) carrying 50Gb/s each will be used to transmit 400GbE.

For these SMF solutions all optical signaling will be done via 4 levels of “Pulse Amplitude Modulation” or as it commonly known, “PAM4.” At 500m, as noted, each optical signal will be supporting 100Gb/s operation. At 2km and 10km each optical signal will be supporting 50Gb/s operation. Thus, the investment in a 4x100Gb/s parallel fiber approach will reduce the initial cost of the module solution, while also setting up the potential for serial 100Gb/s solutions in the future. The decision to use the 8x50Gb/s optical approach for 2km and 10km will allow the industry to satisfy the nearer term needs for higher bandwidth router connections for connecting to service provider networks. Furthermore, given the 50Gb/s and 100Gb/s optical signaling these solutions will use, it was also decided to develop 50Gb/s PAM4 electrical signaling in order to develop a 400Gb/s interface based on 8 lanes rather than 16.

I find myself contemplating a very exciting future for Ethernet based solutions. Seven years ago when I led the 40GbE and 100GbE Task Force, I watched as the initial steps of 25Gb/s signaling were taken. Today, we are seeing the impact of these decisions, as the market is gearing up for 25GbE to be the next step after 10 Gigabit Ethernet (10GbE). With the adoption of 50Gb/s signaling to support 400GbE, we are seeing the same first steps of the future 50GbE eco-system that will replace 25GbE in the future. In addition, given the history of single lane and four lane implementations, talk of 200GbE is already being discussed in regards to the future of Ethernet.

Welcome to tomorrow’s future for Ethernet!

 

相关文章