Data Center Technology Trends for 2022
NOVEMBER 19, 2021 BY LITTON POWER
The demand for data centers continues to grow, pushed in part by the pandemic workplace, prodded along by internet-enabled TVs and other smart devices, and nudged even further by the changing shape and size of digital information. Slouches don’t last long in the tech industry, and all the players are racing to adapt to a changing landscape. As 2021 comes to a close, we look toward the new year for signs of how the data economy will satisfy its hungry consumers.
A Brief Look at the Present
Emerging Data Center Trends
A Brief Look at the Present
2021 was a rebound year from 2020, when infrastructure spending was stifled by 10% due to an industry-wide cash flow shortage sparked by the pandemic. The demand was there. The money was not.
在2021年一份有关数字基础设施的报告中，不动产巨头CBRE发现美国新的数据基础设施破土动工同比增加了 42%，数据中心服务方今年进行了几项大宗土地购买，为下一阶段发展做好了准备。北弗吉尼亚和达拉斯/ Ft. Worth的储备增长惊人，自2015年以来库存储备增长超过225%。令人惊讶的是，硅谷的增长率翻了一番，很显然当地数据中心不只为满足加利福尼亚湾区的几个社区需求。美国是边缘计算市场增长的热土，使数据中心尽可能接近消费侧。
In a 2021 report on digital infrastructure, real estate analyst CBRE found that new data facilities broke ground at a 42% gain over the year prior, and data center providers made several major land purchases this year, readying themselves for the next phase of development. Northern Virginia and Dallas/Ft. Worth saw spectacular growth in inventory, exceeding 225% inventory expansion since 2015. Surprisingly, this nearly doubled the growth rate in Silicon Valley. However, data centers aren’t just trying to satisfy a few zip codes in Bay Area California. The United States is a hotbed of growth for the edge computing market, pulling data centers as close to consumption as possible.
Emerging Data Center Trends
While some parts of the world have slid back into lockdowns and remote work environments, others have resumed normal operations in person and on campus. The effect of the former is still catapulting change in how data reaches its consumers. Here are some of the ways Big Tech is trying to keep up.
从自动检索存储冷数据磁带的机器人，到在基础设施周边维护安全边界的移动巡检安防机器人，自动化正在进入数据中心 --- 在许多方面它已经存在。疫情产生的一些惯性可远程工作为将机器人技术引入数据中心创造了机会，单只有少量公司能像大型科技巨头那样有能力或有动力在该领域进行试验。
From robots that automatically retrieve cold-storage data tapes, to roving security robots maintaining a safe perimeter around a facility, automation is coming to data centers—in many ways it’s already here. Some of the remote-work inertia generated by the pandemic has created an opening to introduce robotics into data centers, and few companies are as well equipped or motivated to experiment in that field than Big Tech.
Indeed, IBM experimented with Roomba vacuum cleaners, equipping the robot with sensors that would measure data center temperatures and humidity readings. AOL claimed several years ago to have created a small data center that ran completely without staffing. Google put industrial robots to work to automate the destruction of decommissioned hard drives at a rate faster than humans could do so.
As the scale of data centers increases, the total volume of hardware and software that needs to be managed will far outpace the capabilities of human staff. In the near term, robotics can pick up the slack and become another tool in a human’s kit. In the long term, a fully automated server center could operate at higher temperatures, reducing cooling and operational costs.
超大规模数据中心的出现是为了满足大型企业以及 AWS、Microsoft Azure或Google Cloud等云和托管服务提供商的海量数据需求。与零售数据中心不同，超大规模中心通常容纳一个客户，这可解释为什么亚马逊、谷歌和微软占据了整个超大规模市场的50%。通过专注单个客户的需求，可以在基础设施中对数据分层处置，从而可以通过适当的温度控制对具有更密集CPU运算的更高需求数据进行环境分区。
Hyperscale data centers have emerged to satisfy the big data needs of large enterprises, and cloud and hosting providers such as AWS, Microsoft Azure, or Google Cloud. Unlike retail data centers, hyperscale centers typically house one client, which would explain why Amazon, Google, and Microsoft occupy 50% of the total hyperscale market. By focusing on the needs of a single client, data can be allocated across a facility in such a manner that higher-demand data with more intensive CPU activities can be environmentally zoned with appropriate temperature controls.
Workloads can also be more efficiently balanced across servers to displace heat production or more efficiently and reliably distribute energy. Don’t let the name fool you—these centers are not distinguished by their size. Instead, hyperscale data centers are defined by its client maximally using the facility’s floor space and resources.
亚马逊在垂直整合其供应链物流组件时独树一帜，从仓储到卡车运输都与众不同。现在它又在做同样的事情，除了1和0 ——为其 AWS产品构建自用超大规模数据中心。远不止如此，预计到 2022 年，整个行业超大规模建设将出现增长，尤其是在中国。
Amazon made waves when it vertically integrated all the logistical components of its supply chain, from warehousing to trucking. Now it’s doing the same again, except with ones and zeros—building its own hyperscale data centers for its line of AWS products. And it’s far from alone, with a projected industry-wide growth in hyperscale construction through 2022, particularly in China.
“Going green” is more than an industry buzz slogan. Microsoft has set an aggressive sustainability initiative with the ultimate aim of becoming carbon negative by 2030. This will require a multi-faceted approach, examining everything from energy production to water usage. And while Microsoft is attempting to adopt a 100% reliability on renewable energy, that strategy simply cannot work without deep wells of energy storage on-site to hedge against fluctuations in wind and solar. Some data centers are already implementing such measures, by installing massive-capacity experimental batteries from companies such as Tesla. Some data center owners are more transparent than others about their environmental impacts, but many have adopted a forward-stance on working toward reducing harm and “going green.”
边缘计算 Edge Computing
Manufacturers, energy companies, electric vehicles, consumers, among many others are driving the increased demand for edge computing. In contrast to hyperscale data centers, providers will need to become smaller and more decentralized to serve the edge market.
Some companies, such as Dell, are taking this to the extreme, with the advent of Micro-Modular Data Centers (MMDCs). These are small, portable, fully self-contained data centers designed to serve the needs of one customer at a time. They often include their own cooling, power, and backup hardware, making them a turnkey solution to a company’s data center needs. Like any solution, MMDCs occupy a sweet spot of cost effectiveness for their use case, but technologies like these are bringing the edge ever closer.
安全升级 Increased Security
多年来，华盛顿特区一直在警告政府雇员和其他政府有关硬件嵌入式威胁隐私和安全的危险。随着网络威胁越来越普遍，攻击者继续瞄准数据中心—— 名副其实的信息宝藏 —— 这些数据中心正在感受到持续攻击的加剧并做出反应。
For years, Washington D.C. has been warning government employees and other governments about the dangers of hardware-embedded threats to privacy and security. As cyber threats grow more pervasive, and attackers continue to target data centers—veritable treasure troves of information—these centers are feeling the heat and responding to it.
Google introduced chip-level security into its data centers several years ago, then open sourced the project in 2019 to deliver the same types of security measures to the industry at large. Dubbed Project OpenTitan, it aims to create maximal transparency and security at the chip level, along with additional features such as self-testing for memory tampering on every boot of the chip. Google has partnered with data giants such as Western Digital and Seagate to develop the standard and deploy the technology into 2022 and well beyond.
良性竞争 Healthy Competition
Tech is a sink or swim world, and few companies know how to swim as well as Intel. The CPU market for servers has historically belonged to chip manufacturers, but recent deals between AMD and Facebook’s parent company, Meta, give its primary competitor a bigger bite of the pie. The steady advancement of chip technology isn’t the only factor in this deal—much of it has to do with the relative availability of AMD hardware during a global chip shortage. But the AMD and Intel feud is a tale as old as time, and it will surely continue well into the future. Meanwhile, NVIDIA is making its own plays, venturing into the market with its first data center CPU, an ARM-based chip dubbed “Grace.” Of course, we’ll be well past 2022 before we get a sense of how the competition plays out.
深 知 社
DKV（Deep Knowledge Volunteer）计划创始成员