|
Changing Landscape Of Data Centers 5 数据中心在不断的演变发展 5 Addressing disruptive change in the data center can be a complicated process and challenging to predict. As rack density trends upwards and cooling becomes a greater challenge, there is more potential for both overcooling and undercooling of IT equipment. 处理数据中心中的破坏性变化可能是一个复杂的过程，而且很难预测。因为机架密度呈上升趋势，冷却成为一个更大的挑战，无论是过冷和欠冷的IT设备都有很多的问题需要解决。 Simultaneously, IT workloads are trending towards being more on demand, and thus less predictable. The speed at which loads can be increased or shed, in order to meet demand, may be the fastest that it has ever been, and it is happening more frequently. 与此同时，IT的工作负载趋向于响应需求侧， 因此更难以预测。为了满足需求，负载增加或减少的速度可能是有史以来最快的。而且这种情况正在更加频繁的发生着。 因此，了解数据中心内每小时、每天或每个月可能发生的更改非常重要。同样重要的是，理解（如果有必要的话，调整）数据中心基础设施是如何运作的。 The previous columns in this series have discussed IT equipment, its thermal design, its interactions with the facility, and current and future sources of change within the IT space. This column will discuss how to identify, analyze, and predict change by applying the practical methodology of a case study to existing data centers. 这个系列的前几篇文章已经讨论了IT设备，它的热设计，它与其他设施的交互， 还有在IT空间中现在和未来资源的变化。这篇文章将会通过在实际案例研究中的实用方法，讨论如何辨别，分析和预测在现有数据中心内将要发生的改变。 High Density Data Centers—Case Studies & Best Practices 高密度数据中心－－案例研究&最佳实践 The case study mentioned here can be found in the Book 7 of the ASHRAE Datacom series, entitled High Density Data Centers— Case Studies and Best Practice. Published nearly 10 years ago, this book takes an empirical approach to understanding the state of data center cooling in order to provide guidance for the high-density environments to come. 案例分析中提到的内容可以在ASHRAE Datacom 系列的第七卷《高密度数据中心－－案例研究和最佳实践》中找到。这本书出版于大约10年前， 这本书采用了经验主义的方法去理解数据中心的冷却状态，以便为即将到来的高密度数据中心环境的研究提供指引。 Today, we see high-density data centers around every corner, the majority of which use the very same cooling approaches that were investigated in this book. There have been many improvements across the past decade, but the fundamentals remain the same. Further, the techniques employed throughout the book can still be applied today in order to study nearly any data center. 今天，高密度数据中心充斥在每一个角落，其中大多数使用了和本书中相似的冷却方法。过去10年，有许多的改善和提高，但是基本方法还是一样的。此外，本书中使用的技术仍然能够在今天使用，以便于研究近乎任何一个数据中心。 The Case Study Process In a typical case study of a data center there are four steps: -Defining the study; -Designing the study; -Data collection;and -Analysis. 案例分析 在一个典型的数据中心案例中，需要四个步骤： －定义这个研究； －设计这个研究； －数据的收集； －分析。 The combination of these steps can provide immense value; skip a step, and get significantly less value. The first step of a case study is to define the questions that are going to be asked and hopefully answered. This is probably the most frequently skipped step, but possibly the most important. Understanding the questions to be studied is necessary to ensure that the subsequent steps are done completely and without bias. 结合四个步骤能够得到非常有用的信息；跳过任何一个步骤，只能得到很少的信息。 第一步是定义一个将要被询问的问题，也希望被解答。这可能是最经常被忽略的问题，但是可能是最重要的问题。理解要研究的问题是很有必要的，为了确保后续步骤能被完成和没有任何的偏见。 The range of potential questions for investigation is nearly limitless. One might investigate the cause of hot spots in a small section of the data center, or attempt to determine why the efficiency of one data center is exceeding the efficiency of another. It could be to investigate the cause of systemic equipment failure, or to simply understand whether the current data center can handle a refresh with higher density IT equipment. 潜在问题的研究范围是没有限制的。可能研究一个热源在一个小范围中的起因，或者试图找出为什么一个数据中心的效率会超过另一个。可能会研究系统设备的故障原因，或者简单的理解是否现在的数据中心可以处理高密度设备的刷新。 The second step is to design the study. Armed with the clear goal in mind, it should be determined what data, measurements and observations will be necessary. This may include some form of spatial, power and cooling measurement, data regarding the installed IT and facility equipment, or observations of the qualitative conditions within the data center. 第二步是设计研究。带有清晰的目标决定了什么样的数据、方法和观察是必要的。这些可能包括一些空间、能源和冷却测量方法的形式，安装的设备数据或者在数据中心内定性条件下的观察 At this point, the scope of the study should be considered. If one were armed with all the data imaginable, there would be few questions that could not be answered. However, gathering data is no instantaneous task, and can easily get out of hand given the scale of many data centers. The sampling size or frequency may be scaled back, yet still offer sufficient data, if scaled back correctly. 基于上述一点，需要考虑研究的范围。如果一个人可以拥有可以想象到的所有数据，那么几乎没有什么问题是不能回答的。然而，收集数据不可能是瞬间完成的任务，考虑到众多的数据中心，收集时有可能会脱离掌控。采样的大小或频率可以缩小，但是仍然能够提供充足的数据，如果缩小正确的话。 Fortunately, as data centers have evolved, the number of installed sensors has grown dramatically, and typically offers up data at a scale that was previously unheard of. The wealth of available data may require modern analysis tools in order to process, but the speed and frequency at which this data can be collected can be invaluable... or it can cause analysis paralysis if managed incorrectly! 幸运的是，随着数据中心的发展，安装的传感器也在急剧增加，而且通常以前所未有的速度提供数据。大量的可利用数据需要现代的数据分析工具来处理， 但是 以何种速度和频率来收集是无价的。。。或者会引起分析瘫痪，如果管理不正确的话。 The third step is to proceed with the data collection as determined in Step 2. While this sounds simple enough, the organized, methodical and error-free collection of data should not be taken too lightly. 第三步是分析第二步中收集的数据。虽然这个听起来很简单，但是这些有组织、有系统和无错误的数据不应该被轻视。最简单的错误能够轻易的污染数据导致得出错误的结论。举个例子，如果要测量整个冷通道的气流，在变化高度或者机架前端距离下进行测量会导致数据产生巨大的变化。图1阐述了一些参考图，这些有助于保持测量的一致并正确地记录数据。
An enabler of consistency and thorough documentation is the proper organization and recording of data. This is especially important for gathering data in the field, as it is easy to get “in the weeds” and make a mistake. Having predesigned input forms, spreadsheets, and graphical tools (annotated plans, key plans, floor grids, etc.) greatly increase the chance of accurate data, not only at the time of collection, but potentially days or weeks later when revisiting the gathered data. 适当的组织和记录数据能够保持一致性和文档的完整性。这对于在田野中收集到的数据很重要，因为很容易“陷入杂草中”并且犯错误。预先设计好的输入表单，电子表格和图形工具（带注释的计划，关键计划，楼层网格等）大大增加了准确数据的机会，不仅收集数据的时候如此，而且可能在几天或几周后重新访问收集的数据也是如此。 Another easily overlooked consideration is a clear and thought out nomenclature. While it may seem simple, consider that even the basic orientation of left and right can reverse when observing both the front and back of a rack. Sloppy nomenclature could easily muddy field observations that could be important. Clear nomenclature also makes it easier to correlate measurements to drawings, making it easier to pinpoint where measurements were made. 另一个容易被忽视的是一个清晰和深思熟虑的命名法。虽然看起来很简单，但是当你去观察一个机架的正面和背面时，基本的左右方向都会反转。草率的命名法也会轻易的混淆重要的野外观 测。清晰的命名也使测量与图纸关联更容易，更容易查明测量的位置。 The final step is analysis. Here, the effort organized into the first three steps culminates, into some well qualified answers. It would be difficult to cover the depth and breadth of analysis techniques that exist, and so we leave it up to the reader to research and apply the tools at their disposal. 最后一步就是分析。这里，经过前面三步的努力，现在能得到一些很有资格的答案。这篇文章很难达到现在已有分析技术的深度和宽度，因此我们将其留给读者去研究和应用他们使用的工具。 However, analysis does not need to be complicated. Sometimes the more straightforward path the best answer. For example, if wanting to understand the uniformity of air distribution throughout an area of perforated floor tiles, little more than observation of the measurements may be necessary to find areas of high and low ow. Add in knowledge of the layout within that space and it might be enough to form a reasonable hypothesis for the uniformity or lack thereof. 然而，分析不需要复杂，有时更直接的路径是最好的答案。例如，如果想要理解整个穿孔地砖区域的空气分布均匀性，只需一点观察测量数据就可以找到区域内的高低。再加上对空间内结构的了解，就足够对其均匀性和非均匀性形成一个合理的假设。 On the other hand, the analysis may be complicated involving a sensitive process with many variables and many field measurements combined with the data from thousands of sensors. Ideally, a well thought out formulation of the study will offer a clear path for analysis, rather than chaotically wading through mountains of data. 另一方面，分析处理是一个敏感又复杂的过程，基于大量传感器提供数据的情况下，要处理许多可变的参数和许多户外现场中的测量。理想情况下，一份经过深思熟虑的研究报告将会提供一条清晰的路径而不是堆积如山的混乱数据。 Case Study Review 案例分析回顾 Understanding the case study as a process is an extremely valuable tool for investigation, not only within a single data center but potentially across many data centers. To provide a more practical understanding, the following will review an actual case study from the High Density Data Centers book. 将案例分析理解成作为一个极其有价值的研究工具，不仅在个别单一的数据中心中而且可能跨越许多数据中心。为了提供更真实的理解，接下来将会回顾一个从一本高密度数据中心书籍中找到的真实案例。 Defining the Question 定义问题 This particular case study deviates from the norm in the sense that it was for a book, and toward that end, the question being asked was simply about understanding the overall thermal profile of the data center. In a broader sense, the question could also have been to understand which of all of the data centers being compared was most efficient, and why. 这本特殊的案例从某种意义上而言背离了正常的标准，只是为了这本书而设计的，直到最后，被提及的问题也是关于数据中心温控技术的理解。从更广的意义上看，问题的重点还是在比较所有的数据中心哪一个更高效，以及为什么会高效。 Designing the Study 设计案例分析流程 The design of the case study began by understanding the general nature of the data center in order to identify the measurements and data that would be collected. This case study was conducted at the National Center for Environmental Prediction. The data center consisted of approximately 6000 ft^2 (550 m^2) of raised floor space, cooled by perimeter com- puter room air conditioner (CRAC) units. 为了能辨别测量方法和收集的数据，设计的开始是理解整个数据中心的大致情况。这个案例分析是在国家环境预测中心里进行的。这个数据中心由大约6000平方英尺（500平方米）的上升空间组成，由周边的计算机机房空调（CRAC）单元冷却。 Located on the raised floor were 51 racks of servers in five rows, with four of them arranged in two hot aisles. The majority of racks were low density, with a few racks with four times greater density. Perimeter power distribution units supplied power to the racks from beneath the raised floor. 在凸起的地板上有51个服务器机架按照5排组成，其中四排被安排在两个热通道里。大部分机架密度较低，少数机架密度为大部分的四倍。周边的配电单元从凸起的地板下向机架供电。
Figure 2 shows the general floorplan of the data center. To provide an efficient system for capturing data, the raised floor tiles were used to create a coordinate system, using letters along one axis and numbers along the other. In this way, each raised floor tile within the data center has a simple and unique identifier that helps to correlate a measurement to its location. The beauty of this simple technique is that the grid is physically visible while in the data center. 图2显示了整个数据中心的平面布置图。为了提供一个有效的系统来捕获数据，凸起的地砖被用来创造一个坐标轴，使用字母沿着一条轴线，使用数字沿着另一条轴线。通过这种方法，每一个凸起的地砖都会有一个简单且唯一的辨别符，这个有助于测量时将它的位置关联。这个简单做法的美妙之处在于网格在数据中心里是物理可见的。 Physical data to be collected or verified included: · The height of the raised floor; · The type and location of the perforated tiles;and · The flow through them. 需要收集或者验证的物理数据包括： －凸起地板的高度； －在地板下的拥挤程度； －设备的位置和方向； －穿孔瓷砖的类型和位置； －通过它们的气流。 Power measurements of as much of the equipment as possible, including lighting, would aid in understanding the total heat load throughout the space. Air temperatures were measured at the CRAC supply, the server inlets, and the CRAC return to understand the temperature throughout the space. 尽可能多得测量设备的功率，包括照明，将有助于理解整个空间的总热负荷。在CRAC供应端、服务器入口、和CRAC返回端测量空调温度，以便于了解空间的温度。 Collecting Data 收集数据 Armed with a plan for what was to be measured For airflow measurements, a properly calibrated velometer was used, which measures the flow of air through its hood. The velometer’s hood was exactly the size of the perforated tiles, making it an efficient tool for the job. 制定一个计划具体什么被测量和如何去测量将会增加数据收集的平稳性和降低错误率。这个计划包括使用恰当的工具为了必要的测量和记录。 在进行空气流的测量的时候，使用了一个恰当的速度校准计，在空气流经过速度校准排气管的时候测量。速度计的排气管尺寸正好和穿孔瓷砖的尺寸一样，这使得它成为一种高效的工具。 For power and temperature measurements, high quality hand-held tools were used including a voltmeter, clamp-on current meter, and digital thermometer. Additional power data was gathered directly from servers using a laptop and specialized software. 在功率和温度测量方面，使用了高质量的手持工具包括一个电压表，夹紧式电流计和数字测温计。使用笔记本和专业的软件直接从 服务器收集额外的电力数据。 These tools were the right ones for the job given the size of the data center and available manpower. However, methodology is just as important. Standards were observed where applicable, such as the ASHRAE guideline for measuring inlet air temperature 2 in. (50 mm) from the inlet. 根据数据中心的尺寸和可利用的人力，上述这些工具非常适合做那些工作。然而，方法论也同样重要。标准会在合适的时候被使用，比如ASHRAE 导则，用于测量进气管从入口深入50毫米的温度。 We leave it to the reader to examine this case study in full for a thorough description of the applied measurement techniques and methodologies. Suffice it to say that it is highly detailed and methodical. Further, many techniques have advanced since this case study and it is recommended the reader carefully examine new approaches, tools, etc. 我们让读者全面检查这个方案分析，主要是那些应用的测量技术和方法。只要说，这是很详细和有条理就够了。此外，自从这个案例分析以来，许多技术得到了提升，建议读者小心仔细检查新的方法和工具等。 Analysis 分析 A great takeaway from this case study was that the analysis did not produce some of the expected results. For example, expected correlations between the airflow and air temperature rise at the racks were not found. However, other data yielded unexpected insight that airflow through the cable cut-outs was amounting to one-third of the total airflow, no doubt a large source of cooling inefficiency. 从这个案例研究中得出的重要结论是，分析没有得到预期的结果。举个例子，没有发现机架上的气流与空气温度上升的的之间的预期关系。然而，其它的数据却出人意料的揭示了通过电缆的断流器的电流是总数的三分之一，这无疑是冷却效率低下的一大原因。 An important step in the analysis was validation of the results through an energy balance. An overall data center flow rate was calculated using the total dissipated power and differential temperature across the CRAC, and compared to the measured CRAC airflow. The compared values were within 5% of each other, providing a reasonable confidence factor in the validity of the measurement and analysis. 一个重要的步骤是通过能量平衡来验证结果。利用总耗电量和穿过CRAC的温差来计算数据中心的流量，结果与CRAC的空气流比较。比较值在5%以内，为测量和分析提供一个合理的置信因子。 Undoubtedly, this case study revealed some detailed results about the data center’s thermal profile that would not have been the same as the day the data center came online. The obvious variable like perforated floor file locations, rack locations, and servers within the racks affect the performance of the data center, 毫无疑问，这个案例分析说明了数据中心的温度剖面在上线后会和之前有很大的区别。明显的变量比如多孔层文件位置，机架位置，和在机架内的服务器影响数据中心的性能， 但是经常忽视了其它因素。在这种情况下，通过CRAC的流量低于预期，部分原因是空气过滤器脏，这表明可能没有进行定期维护。 Conclusion 结论 Change, in today’s data center environment, is unavoidable. Starting from the day that a data center is commissioned, changes will occur that create differences between its intended and actual operation. Not only does the facility equipment change over time, but the IT loads are changing more rapidly than ever. 在今天的数据中心环境中，改变是不可避免的。从数据中心启用的那天开始，改变就会发生，使得预期和实际操作中产生差异。不仅设备会随着时间变化而变化，而且IT负荷的变化速度也会变的比以往要更快。 The case study process is a helpful way to take back control of the changes happening to the data center. This does not mean that suddenly you can control the change, but rather understand how those changes are propagating throughout the data center and adjust accordingly. 案例研究过程是重新控制数据中心发生的更改的一种有用方法。 这并不意味着您可以瞬间控制更改，而是理解这些更改如何在整个数据中心中传播并相应地进行调整。 Whatever the case, understanding the causes of change is as important as knowing how to mitigate it. By fully understanding the environment, the course of the data center’s evolution can be altered in subtle or dramatic ways that can increase its performance, stability, or reliability. 无论哪种情况，理解变化的原因与知道如何减轻变化同样重要。通过充分理解环境，可以以微妙或戏剧性的方式改变数据中心的发展过程，从而提高其性能、稳定性或可靠性。 翻译： 翻译： 邓秋实 同济设计院电气设计师 |
|
|
