????還記得泰坦嗎,？這可不是那支咸魚翻身的橄欖球隊(duì)。這是臺(tái)超級(jí)計(jì)算機(jī),，去年11月,，媒體大幅報(bào)道了泰坦一舉成為全球最快的計(jì)算機(jī)。泰坦的計(jì)算速度非?？植?，處理能力相當(dāng)于50萬(wàn)臺(tái)筆記本電腦。

????不過(guò),，如果沒(méi)有合適的管理軟件,，如此強(qiáng)大的計(jì)算機(jī)也會(huì)英雄無(wú)用武之地,。軟件是未來(lái)高性能計(jì)算機(jī)的關(guān)鍵部分——不過(guò)它經(jīng)常被忽略。隨著越來(lái)越多的產(chǎn)品和服務(wù)遷往云端,，軟件對(duì)各行各業(yè)的影響將更加重要,。自適應(yīng)計(jì)算（Adaptive Computing）公司首席執(zhí)行官羅比?克萊德稱：“高性能計(jì)算和云計(jì)算之間的界限越來(lái)越模糊?！弊赃m應(yīng)計(jì)算是一家軟件企業(yè)，產(chǎn)品能幫助超級(jí)計(jì)算機(jī)和云計(jì)算服務(wù)器提高效率,。1月30日,，該公司展示了自行研發(fā)的軟件，能將泰坦的運(yùn)行效率從70%提升到95%,?？巳R德稱，用來(lái)優(yōu)化泰坦的軟件和用來(lái)優(yōu)化云計(jì)算的軟件非常像,。

????這一點(diǎn)很重要,，因?yàn)殡S著運(yùn)算能力日益強(qiáng)大，最大的挑戰(zhàn)將從提升原始處理能力變成優(yōu)化已有系統(tǒng),?！拔覀兗磳⑦M(jìn)入一個(gè)運(yùn)算幾乎無(wú)止境的世界，”云存儲(chǔ)公司Box的首席執(zhí)行官艾倫?列維稱,?！肮芾磉\(yùn)算的工具正成為下一個(gè)限幅器?！绷芯S稱,，Box公司有數(shù)十名員工每天就在專門思考這一個(gè)問(wèn)題。

????列維表示,，計(jì)算機(jī)管理系統(tǒng)的市場(chǎng)不斷增長(zhǎng),，而且這一趨勢(shì)將持續(xù)。隨著市場(chǎng)開始需要管理計(jì)算機(jī)管理系統(tǒng)的系統(tǒng)等等,，科技公司又會(huì)有新的業(yè)務(wù),。萊維說(shuō)：“我們目前處于這個(gè)問(wèn)題的新層次。而在此之上,，又將有一個(gè)新層次,，直到我們?nèi)祟惐惶蕴,！彼@么說(shuō)并不完全是玩笑話,。

????這類軟件應(yīng)幫助企業(yè)向用戶提供更好的服務(wù)?？巳R德說(shuō),，以一家憑借云端程序提供語(yǔ)音識(shí)別服務(wù)的公司為例,，如果支撐該服務(wù)的軟件能高效地為各任務(wù)確定優(yōu)先級(jí)，該服務(wù)就能更快地進(jìn)行識(shí)別,。而用戶喜歡速度,。

????這種看似遙遠(yuǎn)的軟件還有一大好處：它將提升云端巨大運(yùn)算能力的可訪問(wèn)性。 “三,、四十年前,，必須建造一臺(tái)大型主機(jī)才能進(jìn)行復(fù)雜的運(yùn)算，”列維說(shuō),。這些大型主機(jī)耗資數(shù)千萬(wàn)美元,。“如今,，完全可以在別人的云端進(jìn)行虛擬運(yùn)算,。你只要為你在各個(gè)獨(dú)立的時(shí)間段內(nèi)使用的運(yùn)算付費(fèi)?！倍S著軟件日益改進(jìn),，云管理日益完善，越來(lái)越多的人都可以用上它的運(yùn)算能力,。

????這樣一來(lái),，各公司將能圍繞運(yùn)行復(fù)雜程序得出的發(fā)現(xiàn)進(jìn)行創(chuàng)新，而不至于在這個(gè)過(guò)程中破產(chǎn),。這將反過(guò)來(lái)就能使各公司能提供更好的服務(wù),。直到計(jì)算機(jī)把我們?nèi)宽斕妗＃ㄘ?cái)富中文網(wǎng)）

????譯者：項(xiàng)航

????Remember the Titan? No, not the comeback football team. The supercomputer that generated headlines last November for ranking as the world's fastest. Titan can crunch so many calculations, it has the equivalent processing power of 500,000 laptops.

????All that computing might is for naught without software capable of managing it. Software is a major—if often unsung—factor in the future of high-power computing. It will matter increasingly to businesses of all kinds as more and more products and services move into the cloud. "The line between high performance computing and cloud is blurring," says Rob Clyde, CEO of Adaptive Computing, a company that builds software to increase efficiency in supercomputers and cloud-based servers. Today, Adaptive revealed it developed the software that boosted Titan's efficiency from 70% to about 95%. The software used to make Titan tick is similar, Clyde says, to software that optimizes the cloud.

????That matters because as computing power gets more and more robust, the biggest challenge will switch from increasing raw processing brawn to optimizing the systems that we have. "We're moving to a world where computing is almost sort of indistinguishable from being infinite," says Aaron Levie, CEO of cloud storage company Box. "The tools to manage that computing are becoming the next limiter." Box, Levie says, devotes dozens of employees who spend their days thinking about that single problem.

????There is a growing market for systems that manage computers, Levie says, and that trend will continue. A slice of the tech economy will build out as the market begins to demand systems that manage the systems that manage computers, and so on. "We are now at a new meta layer of this problem. After this, it will be a new meta layer on top that, until we as humans are obsolete," Levie says. He is only partially joking.

????Such software should help companies deliver better services to consumers. For example, says Clyde, take a company that offers a speech-recognition service powered by programs in the cloud. If the company supports the service with software that efficiently prioritizes tasks, the service will find words faster. Consumers love speed.

????This remote-seeming software has another big benefit: It will boost the accessibility of massive computing power in the cloud. "Thirty or forty years ago, you had to basically build out a mainframe to run complex computations," Levie says. Those mainframes cost tens of millions of dollars. "Today, you can just run it in a virtual job in somebody else's cloud. You literally only pay for the amount of computing you use in a very discrete time period." Even better software, better cloud management, will let more and more people use its computational power.

????That will give companies the power to innovate around discoveries made from running involved programs without bankrupting them in the process. That should, in turn, enable companies to deliver better services. That is, of course, until the computers render us useless.