立即打开
云计算能耗过高?把服务器搬到太空去!

云计算能耗过高?把服务器搬到太空去!

Kirk Kardashian 2015年02月25日
能耗过高、不具备可持续发展能力,是云计算面临的一大问题。有鉴于此,洛杉矶的初创公司ConnectX试图将储存在云端的企业数据搬到太空,而且正在试验一种非二进制的传输方式。一旦成功,我们储存、传输和分析数据的模式就将发生革命性变化。
    从国际空间站看加拿大,2012年

    当人们谈到云计算时,大家都明白,所谓的“云”不过是一个比喻,象征着一系列远程的服务器网络。但当兰斯•帕克提到“太空计算”这个概念时,他并不是在玩诗意,而是说真的:他想把服务器搬到太空进行运作。帕克是ConnectX公司的CEO,这家位于洛杉矶的初创公司致力于将企业储存于云端的数据搬到太空。一旦成功,它可能会彻底改变我们储存、传输和分析信息的模式。

    “云计算”这个名字虽然好听,但有一定的误导性。云在天上飘,而组成云计算的服务器网络显然是在地面上。云是优美而千变万化的,就像老天爷在空中泼墨一般;数据服务器则笨拙而有形,也没有给人多少解读的空间。不过据帕克称,云计算还有一个大问题,那就是不具备可持续发展能力。他或许说得有道理。从创世之初到2002年以前,全球一共创建了50亿G字节的信息。而今天,我们差不多每隔10分钟就会创建这么多的信息。数据存储技术已经无法跟上呈指数级增长的数据创建速度。另外数据中心也是能耗大户,它们消耗的电力已经达到了全球电力的10%。

    将服务器放到太空或许能够解决能耗问题,因为太空中充足的太阳光辐射可以为它们提供免费的电力。另外,太空环境对旋转式硬盘驱动器非常有利。在零重力下,硬盘驱动器旋转的阻力变小了,而太空中的极寒环境意味着服务器不用担心过热问题,从而运转得更快。

    当然,如果没有一种经济的方法能将服务器发送到太空,那么太空计算的所有好处都是无稽之谈。好在近年来,人造卫星的尺寸越来越小,成本也在逐年降低。比如所谓的“立方体卫星”(CubeSat)只有一个甜瓜那么大,花上不到10万美元的成本,就能将它送到太空轨道上。尺寸更小的“管状卫星”(TubeSat)只需要8000美元就能发射。把这个成本与15万美元的蜂窝基站相比,“小型卫星就变成了一个很明智的选择。”帕克说。

    如果说成本并非太空计算的主要障碍,那么传输效率问题很可能是。目前,服务器和电脑收发信息主要还是依赖电缆或光纤,典型的应用程序的传输速度可以达到每秒100兆左右。而卫星和智能手机则是通过电磁波传输数据。(数据通过二进制进行编码,用一串波峰和波谷来区别数字。一系列数字随后被发送到接收端,译解成可用的信息。比如字母“A”会被显示成8个0和1——也就是8个波形。)这种传输方法的效率要远远低于有线传输。比如,威瑞森电信(Verizon)声称,其LTE无线网络的下载速度可以达到每秒5至12兆。但就算按照这个速率,在太空存储大量数据也是不可行的。

    ConnectX公司就此提出了一个双重解决方案。首先,该公司正在尝试着通过“扭结”无线电波束的方法来提高数据传输速率。这听起来似乎有些天方夜谭,但南加利福尼亚大学电子工程学教授艾伦•威尔纳最近已经开发出一种类似技术,能够使无线电传播速率达到每秒32G——大约比LTE无线传输快了30倍。威尔纳已经在9月16日的《自然通讯》上发表了一篇论文。帕克表示:“看着这个研究成果出来很有意思,因为我们的技术也需要旋转,但我们使用了另一种更加有效的方式。”

    这种不同的方式是指ConnectX公司可能已经实现的一项突破性创新。除了将无线电波束进行扭结以加快传输速度之外,无线电波束还将以一种压缩的、非二进制的符号结构进行传输。这样一来,字母A就不再需要8个波形,而只需要1个波形。甚至像“苹果落在了离树很远的地方”这类句子也只需要一个符号就能完成。帕克表示:“随着研发工作的进一步深入,我们的符号结构甚至可以变得更复杂,一个符号能代表越来越多的信息。”这不仅解决了传输速率的问题,也可以加快数据的分析速度,因为它大大减少了数据梳理过程所占用的时间。

    一家熟悉ConnectX业务的《财富》美国500强企业的数据科学总监表示,ConnectX专有的符号结构可以用一个词来概括,那就是“可视化”。在传统的大数据分析模式中,记录是二维的,因此很难观察到产品和人的关系,或者随着时间的推移而产生的变化。那位数据科学总监表示:“这正是ConnectX想要纠正的,它想把数据描绘成具有关系、特性和运动的物体。这也正是目前市场上所缺少的——一种能够展示分析结果,并帮助企业高管指明公司下一步走向的方法。”

    ConnectX的团队目前拥有13名员工,其中大多数都是工程师和科学家,他们希望填补市场空白。这个团队计划首先在地面上验证符号传输技术的可行性,然后再在太空进行测试。Beta版产品计划于2017年推出。(财富中文网)

    译者:朴成奎

    审校:任文科

    When people talk about cloud computing, it’s usually understood that the cloud is a metaphor for groups of remote, networked servers. But when Lance Parker talks about “space computing,” he’s not taking poetic license. He means it literally: physical servers operating in outer space. Parker is the CEO of ConnectX, a startup company based in Los Angeles that’s working on a way to take corporations’ data out of the cloud and into the final frontier. If his company succeeds, it could revolutionize the way we store, transmit, and analyze information.

    As metaphors go, “cloud computing” sounds nice but is misleading. Clouds are in the sky; server farms are decidedly earthbound. Clouds are aesthetic and amorphous, a Rorschach test in the atmosphere; data warehouses are boxy and don’t leave much room for interpretation. But the bigger problem with cloud computing is that, according to Parker, it’s simply unsustainable. He may have a point. From the beginning of time until 2002, the world created five exabytes (five billion gigabytes) of information; today, we create that much data in about 10 minutes. Data storage technology just isn’t keeping pace with the exponential growth of data creation. Data centers are also energy hogs, using up 10% of the world’s electricity.

    Putting servers in space could potentially remedy the energy problem, since they could be powered by free, plentiful solar radiation. And, Parker adds, the space environment would be advantageous for spinning disk drives. Zero-gravity allows the drives to spin with less resistance, and the extreme cold in space means the servers could process faster without overheating.

    Of course, all of the benefits of space computing are worthless without a cost-effective way to launch a server into space. Fortunately for ConnectX, satellites are getting smaller and cheaper by the year. The CubeSat, which is about the size of a cantaloupe, can be placed into orbit for less than $100,000. The even smaller TubeSat can be launched for $8,000. Compare those costs to, say, a cellular tower at $150,000, and “the small satellite becomes the obvious choice,” Parker says.

    But if cost is not the main barrier to space computing, data transmission rates could well be. Currently, servers and computers send and receive information mostly through cables or fiber optics, which allow a data transfer rate of around 100 megabits per second for typical applications. Satellites and cell phones, on the other hand, beam information using electromagnetic waves. (The data is delivered in a code of ones and zeroes, with the height or the frequency of the wave distinguishing between the numerals. A series of numerals is then translated by the receiver into useable information. For example, the letter “A” is represented by eight zeroes and ones—eight waves.) This method of transmission is much slower than using wires. Verizon claims that its LTE wireless network, for example, downloads at speeds between five and 12 megabits per second. At that rate, storing large amounts of data in space is just not feasible.

    ConnectX’s proposed solution to the transfer rate challenge is twofold. First, the company is working on a way to twist radio beams to increase data transmission rates. This might sound farfetched, but Alan Willner, a professor of electrical engineering at the University of Southern California, has recently developed a similar technique, which achieved a rate of 32 gigabits per second—about 30 times faster than LTE wireless. Willner published an account of the research in Nature Communications on September 16. “It was interesting to see that come out,” Parker says, “because our technology requires spin, but we do it in a different way that’s much more effective.”

    That different approach is where the real blockbuster innovation of ConnectX might be found. In addition to twisting the radio beams for faster transmission, the beams will be sending the data in a condensed, non-binary symbol structure. Instead of the letter “A” requiring eight waves, it could require just one. Or an entire sentence, “The apple fell far from the tree,” could be communicated with a single symbol. “Our symbol structure can get even more complex as we evolve it,” Parker says, “where a symbol can mean more and more information.” This eases not only transfer rates, but could also speed up analytics by reducing the sheer volume of data that must be combed through to find actionable intelligence.

    According to the head of data science at a Fortune 500 company, who is familiar with ConnectX’s work, the value of the company’s proprietary symbol structure can be communicated in one word: visualization. In the traditional approach to big data analytics, records are two dimensional and it’s therefore difficult to see relationships between products and people or changes over time. “That’s what ConnectX is trying to fix,” the data science manager says, “picturing the data as an object with relationships, attributes and movement. That’s what’s missing on the market now: something that can show the insights, and help executives figure out the next steps for the company.”

    ConnectX’s team of 13 employees, most of whom are engineers and scientists, are hoping the company can fill that void. They plan to prove the transmission technology on earth first, then test it in space. The beta launch is slated for 2017.

  • 热读文章
  • 热门视频
活动
扫码打开财富Plus App