【速读摘要】：
原文约 2330 字| 图片 1 张| 建议阅读 13 分钟|
2020年数据中心的9大技术发展及预测
由DCD中国授权DKV编译并发表于DeepKnowledge微信公众号
笔者在梳理完毕曾经对数据中心2019年发展趋势预判的准确性后，现在，是时候展望未知的2020年了。在圣诞节假期某天一次酒会上，我产生了对来年数据中心趋势的“灵感”。接下来，我将系统论述本人对边缘计算、物联网、5G和其他黑科技在未来一年发展的整体看法。
1. 数据中心需要根本性地回归“可持续发展”的主题上
数据中心在制冷方面的能耗不降反升—Thinkstock【提供照片】
我们需要共同努力，使数据中心的发展成为一个“可持续发展”的理念。在关于可再生数据中心头条新闻铺天盖地的报道下，许多人误认为可持续发展的数据中心已经成为基本能力。
2019年12月，Supermicro发布了其第二版数据中心年度及其环境报告。该报告基于对5000余名IT界专家的采访访问完成。报告（数据中心可持续发展实践）呈现的结果令人失望。86%接受采访的专家并不将环保作为数据中心（建设和运维）的重要指标。仅不到15%的受采访者会真正重视能量效率、企业社会责任或环保等问题。
提升数据中心能效最简易的做法之一即是将机房环境设置的温度适当提高。然而，当前的趋势却是反其道而行之。数据显示，2019年，将数据中心进风口温度设置在21℃以下的受访者比例从2018年的12%提升至17%，将进风口温度设置在21~24℃的比例从43%提升至51%，将进风口温度设置在24℃以上的比例则相较2018年有所下降。
随着时间的推移，老旧服务器不能在较高温度运行的关系已经不那么大了。关于这件事，我在2013年和2014年反复提过，因此，温升不会宕机，实在不算什么需要广而告之的新鲜事儿了。无论您是否认可，在Greenpeace的努力下，数据中心业界的头部企业已经完全公开其数据中心的能效指标，并且起到了不错的示范作用。因此，数据中心行业的其他玩家需要承担起相应的社会责任来。
或许当前的情况也反映出，很多企业并没有完全独立和运维自己数据中心的能力。企业通常对变化更容易感到担忧怀疑和犹豫不决。然而，无论如何，在面对数据中心的可持续发展问题上，每一个业内人士都已责无旁贷。
2. 5G 技术在2019年对人们生活的改变微乎其微，在2020年也无需有所期待
去年，我曾作出预测，5G技术并不会深刻影响人们的生活。现在到了2020年，我认为依然不会有大的变化。诚然，市面上出现了更多支持5G的手机（等移动设备），不过进步也基本到这里了。我这么说的原因是基于适配5G的毫米波长高频和传统低频波技术原理区别而出发的。传统的低频波波长较长，其优点是可传输距离长，更容易穿过墙面等阻碍，但缺点是带宽较窄。反之，毫米波波长大大缩短，因此其频谱带宽可大大增加。不过，这也使得毫米波容易受到干扰，在室外经过墙壁、树木和火车时衰减较大。
明确了5G采用的毫米波和传统低频波的原理区别后，我们便可认识到，低频的5G是一个不错的技术小升级。我们平时听到的都是毫米波对5G技术的改变。基于毫米波的技术局限性后，不少评论者认为，5G的应用无法辐射到高密城市以外的区域。若要真正全面铺开毫米波5G，则需花费极大的时间和金钱成本建设大量的相应的基站。因此5G近期能否全面铺开应用，大概率是否定的。不幸的是，这将会直接影响我对数据中心未来趋势接下来的两点判断。
3:工业物联网的宏伟愿景不会在2020年得到实现
在展开这一观点前，首先解释一下工业物联网的概念。概括来说，工业物联网是指利用计算机、互联传感器及其他设备通过数据分析在交通、能源及其他工业领域实现智能化的工业操作。工业物联网已经远远超出了所谓联网冰箱或智能照明等智能家居的范畴。
工业物联网的未来可谓前景无限，但从技术层面，实现工业层级的实时数据采集和分析对带宽有很高的要求。我们不可能给每一个传感器都进行拉线，因此就必须仰赖5G技术作为纽带。5G技术的缺失，并不一定会对工厂本身的部分业务应用造成影响，但它将阻碍更广泛分布式和远程领域的应用。因此，在毫米级5G技术铺开前，我们在工业物联网上只能开展一点皮毛工作。
关于边缘计算一些根本问题的补充分析——成本、连接性、边缘计算的运行方式等。从工业界专家和从DCD 读者等角度对边缘计算的发展方向进行了解。
4. 边缘计算持续处于上升期，但2020年，边缘计算依然不会呈现量的跃升
边缘计算现在已经从一个热词逐渐落地，但仍然处于产业化相当初期的阶段。每一个人对于边缘的定义都不一样，也即，现在存在各种不同规模和形式的边缘计算。边缘计算的发展同样与5G技术和工业物联网深刻绑定。只要三者任何一方成熟，其余二者都会得到相应发展，而这需要时间。因此，边缘计算在2020年不会出现大颗粒的成果。
5. 人才缺口将开始得到填补
数据中心行业以及科技产业的人才缺失问题众所周知。“银发海啸”（指人口老龄化）不断逼近的脚步也使得这一问题不可能短时间得到解决。不过，尽管如此，经过近来一些组织的努力，今年数据中心相关人才的稀缺问题将会得到有效改善。
InfrastructureMasons就是其中一个组织。该组织为一个具有多个委员会的工业协会，致力于将数据中心生态多样化。其中的女性委员会致力于解决人才短缺问题，让更多人有望成为合格应聘者，同时避免无意识的偏见。为此，他们共同发布了一个招聘经理必备技能。该组织的另一个委员会，退伍老兵招聘委员会则致力于让退伍老兵从事数据中心这一行业。此外，该组织的拉丁语委员会致力于在西班牙语系创建一个驱动数字意识的合作生态系统。
InfrastructureMasons 是当前许多致力于解决数据中心人才问题的一个代表。在他们的推动下，数据中心的人才问题将有望在今年得到改善。
6. 另一轮产业联合即将到来
在过去的几年间，数据中心产业经历了大量的产业联合，而这一现象似乎吸引更多资金进入数据中心。部分投资者通过购买现有的数据中心资产，成为主机托管市场的新晋赛道玩家。这些投资者中，可能存在打算长期在数据中国产业发展的，但更多的可能是希望低价买入，高价卖出的短期玩家。今年，这一轮产业联合的势头将有所增大。
7. 数据主权将进入“青春期”的阵痛
2019年10月下旬，德国经济部长Peter Altmaier 宣布了一项欧洲云行动计划，名为“Gaia-X”。他在宣布后不久就下台了。考虑到这一行动计划将会带来的种种麻烦，这样的结果也是情理之中。GDPR的尘埃还未落定，他们这些行为又要把问题搞得更加复杂。
部分人士解读该欧洲云行动计划为与美国云服务提供商进行的较量；另一种观点则认为是对数据主权标准和规则的制定。无论是哪一种，这样的行为至少可能把数据存储的问题进一步复杂化。我从新闻中看到，法国支持了该项行动计划。德国和法国竟然能在一个问题上达成一致，实属难得。无论如何，2020年，或许我们能在数据主权这个问题上看到些有些有意思的事情。
8. 数据的惯性将继续妨碍数字化转型的步伐
2019年时，我预计数字化转型将为信息化数据革新带来大量的利润。不幸的是，我对商业和信息技术对准融合的预期还是过于乐观。回头来看的话，企业对技术变革的保守怀疑和担忧犹豫不是阻碍信息技术商业化的唯一因素。我认为，数据惯性，即数据与业务之间的相互吸引作用，也是其中一个原因。
随着数据体量的增大，它们更难迁移。数据惯性的作用，使其保持数据在其当前数据中心里，防止它移动到云或其他更有用的平台。这样可以防止通过更富有分析能力的工具（如人工智能（AI） 和其他高级分析）来分析数据。这可能是一个缓慢的演变; 因此，数据惯性今年将继续阻碍数字化转型。
9. 液冷技术市场仍将处于早期阶段
经过数年的大肆宣传，数据中心功率密度的飙升终于在2019年成为现实，而这主要来源于服务于人工智能计算的服务器内置芯片的升级。最新的硬件已经将机柜密度提升至60kW，而在普遍认知当中，空冷的有效应用范围为功率密度在40-60kW之间的机柜。显然，数据中心的制冷系统需要发生改变。
我对液冷技术市场（不会全面铺开）的判断并非因为液冷技术本身的成熟度问题。无论是全浸泡式液冷技术，还是芯片液冷技术，都已经得到验证，随时可以上马。我的判断依据是在于上述将功率密度推高的硬件的推广度还不高。随着人工智能技术得到越来越多的应用，机柜功率将会越来越高，液冷技术也会得到更广泛的应用。就目前来说，液冷技术的应用将继续局限在人工智能计算业务聚集的数据中心当中。
时间将验证一切……
我对关于数据中心的未来变化和技术趋势的预测基本就到这里。这些预测正是DCD的首席执行官George Rockett 需要的吗？为何不是10大趋势预测呢？或许，第10条趋势已经在我去夏威夷路上被整日整夜的雨冲掉了……不过，任何时候，我都非常愿意听取您对我的趋势判断的看法，因此请别忘了在网页下面、领英或者推特下给我们留言，感谢！
英文原文：
Top 9 data center and tech predictions for 2020
After reflecting onthe accuracy of my predictions for 2019, it’s time to wade into the unknown of 2020 and unveil the eggnog-inspired divinations that came to me over the holiday season. Read on to see my predictions on edge, IoT, 5G, and the other tech inspired geekery.
1: Time to get backto the fundamentals of sustainability
Data centers are going backwards: getting cooler, not warmer! –Thinkstock
This is aprediction that we all need to work on together to make happen. Given all ofthe headlines about the massive amounts of renewable energy being implemented around the world to power data centers, many incorrectly assumed that the basic blocking and tackling of sustainability was happening in the background.
In December,Supermicro released its second annual Data Centers and the Environment report based on an industry survey of over 5,000 IT professionals. The results are disturbing to say the least. 86 percent of those surveyed don't consider the environmental impact of their facilities as an important factor for their data centers. Less than 15 percent report that energy efficiency, corporate social responsibility, or environmental impact were key considerations for their facilities.
One of the simplest things you can do in a data center to improve energy efficiency, turn up the thermostat, has reversed course. The percentage of respondents that keep their inlet air temperatures at or below 21°C (69.8°F) increased from 12 percent in2018 to 17 percent in 2019. Those that keep inlet air temperatures between 21°Cand 24°C (75.2°F) rose from 43 percent to 51 percent. Operation in all other higher temperature ranges decreased.
The cautionary warnings about legacy equipment not being able to operate at higher temperatures become less relevant as time goes on. I wrote about this in 2013 and again in2014, so this isn’t new or rocket science for crying out loud! Due to the efforts of Greenpeace, whether you approve of their methods or not, the largest players in the industry are transparent about their energy efficiency and they are doing quite well. That puts the onus squarely on the enterprise.
Perhaps this highlights why most enterprises shouldn’t own and operate their own data centers. Enterprises are also more susceptible to FUD (fear, uncertainty,doubt). In any case, change is necessary, and it is incumbent on each of us todo our part.
2: 5G didn’t changeour lives in 2019, and it won’t in 2020 either
Last year, Ipredicted that 5G will change our lives, but not in 2019. Well, it won’t in2020 either. Yes, there are far more 5G enabled mobile devices on the market,so that is improvement. To provide context, let’s first make the distinction between low-band and millimeter wavelength. Low band has a longer wavelength,which allows it to travel longer distances and travel through solid objects like walls more easily, but this comes at the expense of bandwidth. Millimeter wavelength has a much smaller wavelength (thus the name) and much higher bandwidth potential, but is highly affected by walls, trees, and even rain.
With this in mind,low band 5G will be a nice little upgrade. Millimeter wavelength is the lifechanging 5G we have all been hearing about. Given the technological limitations, critics believe that it will never scale beyond dense urban areas.To roll-out millimeter wave length in a serious way, it will take a monumental number of installations, which in turn will take time and a lot of money. Unfortunately,the near-term realities of 5G also impact my next two predictions.
3: The grandpromise of IIoT won’t be realized in 2020
Before we getstarted on this one, let’s define IIoT (Industrial Internet of Things). Very generally speaking, it’s the use of instruments, connected sensors and otherdevices on machinery and vehicles in fields such as transportation, energy, andother industrial sectors. IIoT obviously extends beyond an Internet-connected refrigerator or light bulb.
IIoT has amazing potential, but it takes band width to enable real-time data gathering and analysis on an industrial scale. We can’t run fiber to every single little sensor,so herein lies the tie to 5G. Lack of robust 5G won’t impact certain applications like factories, but it will impact more widely distributed and remote applications. Without widespread millimeter wavelength 5G, we are just scratching the surface of what IIoT can deliver.
A special supplement tackling the fundamental questions surrounding Edge computing - fromthe cost, to the connectivity, to how it is processed, and more. Hear fromindustry experts, and from DCD's readers, to learn where the Edge will live.
4: Edge computingwill continue its upward trajectory, but will remain in its infancy in 2020
Edge computing hasgone from buzzword to reality, yet it’s still just beginning to crawl in termsof maturity. Everyone’s edge is different, which means widespread installationsof varying sizes and shapes. Edge is also very much tied to the success andmaturity of 5G and IoT. As each matures, the effectiveness of the other two will increase. That is going to take some time, which will prevent edge from bearing substantial fruit in 2020.
5: Solutions to thetalent gap will begin to emerge
The staffing challenges of the data center and tech industries are well documented. While those issues aren’t going to go away any time soon with the impending ‘silver tsunami’, there are groups that are working on them and I believe their efforts will begin to bear fruit this year.
One such group isthe Infrastructure Masons, an industry association with multiple committees focused on diversifying the data center community. The Infrastructure Masons Women Committee is focused on addressing “the talent shortage and expanding the qualified applicant pool and avoid unconscious bias by co-publishing a HiringManagers Toolkit.” The Infrastructure Masons Employ Military Veterans Committee focuses on attracting “Military Veterans to pursue a career in this industry.”Finally, the Infrastructure Masons LatinX Community’s goal is “to create acollaborative ecosystem within the Hispanic community to drive digital infrastructure awareness.”
The Infrastructure Masons are but one of many groups tackling these issues and progress shouldbegin to emerge this year.
6: Another wave ofindustry consolidation is coming
There has been substantial industry consolidation over the past few years, but this seems tohave attracted even more money into the sector. Some of these new investorshave purchased existing data center assets and formed new players in the colocation market. There is the possibility that some of these investors are init for the long haul, but that is typically more the exception than the norm.Logic would dictate that many of these companies are being created with theintent of selling them off at a profit as quickly as possible. This new wave of consolidation will gain momentum this year.
7: Data sovereigntyenters its painful prepubescent stage
In late October of2019, Germany's Economic Minister Peter Altmaier announced a European cloudinitiative called Gaia-X. He then promptly fell off of the stage, which isappropriate given how messy this thing is going to be. The dust hasn’t even settled from GDPR and they have to throw another layer of complexity into thering?
It’s been interpreted as a direct competitor to US cloud providers by some and a set of standards and rules by others. Depending on where these lands, it has thepotential to further complicate where data is stored at the very least. Justwhen I thought it couldn’t get any stranger, I read that France backs theinitiative as well. When was the last time Germany and France agreed onanything? In any case, 2020 looks to be an interesting year for datasovereignty.
8: Data gravitycontinues to hinder digital transformation
In 2019, Ipredicted that digital transformation would turn IT into a profit center.Unfortunately, I was a bit premature and/or optimistic in my prediction of alignment between business and IT. In hindsight, FUD was not the only factor atplay. I believe data gravity, the idea that data and applications are attractedto each other, also played a part.
As datasets grow larger, they are harder to move. This data gravity works to keep the data inits current location, preventing it from moving to the cloud or some other more useful place. This prevents the data from being analyzed by more insightful tools such as Artificial Intelligence (AI) and other advanced analytics. Thiswill likely be a slow evolution; thus, data gravity will continue to hinder digital transformation again this year.
9: Market forliquid cooling remains nascent
After years ofhype, power densities finally rose substantially in 2019. This was driven inlarge part by new chip sets aimed at AI. The newest hardware is driving rackdensities as high as 60 kW. The limits of air cooling are thought to be in the40-60 kW/rack range. Simple math would indicate that something has to change.
My contention withthis prediction doesn’t lie in liquid cooling being up to the challenge. The technologies, whether total immersion cooling or liquid to the chip, are provenand ready for prime time. My prediction is simply that the implementation of the hardware that drives these densities is still limited. As AI grows inadoption, it will drive densities even higher and liquid cooling will be more wide spread. For now, liquid cooling will be limited to certain areas of thedata center where these AI compute clusters reside.
Time will tell…
There you have it,my attempt at predicting the future of the ever-changing data center and tech world. Are these the bold predictions DCD CEO George Rockett was looking for? Why couldn’t I come up with an even 10 predictions? Perhaps the non-stop rainon my trip to Hawaii washed away the last one. In any case, I am alway sinterested in hearing your thoughts on my predictions, so please share your feedback below, on LinkedIn, or on Twitter.
翻译：
严睿婷 研发工程师 华为技术有限公司
编辑：
蔡海霞 湖北威那尔机电技术有限公司
联系客服
