????就在Facebook宣布推出能讓用戶在自己的社交網絡內快速查找信息的功能一天后，該公司就推出了另一項極具創新潛力的產品：一款用于服務器的開源主板，它被親切地稱為“集體擁抱”（Group Hug）。不出所料，比起前一天發布的消息來，這款看起來不怎么美觀的硬件并沒引起多大的轟動。但就改變數據中心的建設方式而言，它可能具有同樣深遠的影響。

????為什么呢？Facebook這款母主板允許不同廠家的芯片裝在同一個插槽中。它意味著一臺服務器就能容納相互競爭的架構，比如英特爾公司（Intel）的x86技術或英國的安謀國際（ARM Holdings）的設計。如果這款“集體擁抱”能被采納，它將在服務器行業激起重大變革。一直以來，這個行業所生產的專有設備總是無法有效兼容，并因此而著稱。而且它可能為ARM芯片廠商開辟一條新的生路，助它們進軍服務器市場。目前這個市場還是英特爾一家獨大。

????“集體擁抱”主板是由Facebook開發的。但這家社交網絡公司是通過“開放計算項目”（Open Compute Project，OCP）將開發工作“開源”給全球進行的。18個月前，為了將更有效的數據中心硬件設計眾包出去，Facebook率先成立了這個項目組。當時，該項目只有一個正式成員——Facebook自己。但本周三，有近2,000人來到加州圣克拉拉市會議中心參加第四屆OCP峰會。目前，該項目已有超過50家成員單位，包括英特爾、EMC、安謀國際、AMD和閃迪（SanDisk）這樣的巨頭。Facebook硬件設計副總裁弗蘭克?弗蘭克沃斯基說：“看到這么多勁敵同處一室真是有意思，這種做法還是有一定爭議的。”

????按照Facebook的宏圖，它絕不只是推出可更換芯片的新主板就了事。通過開放計算的成員單位和它自己的內部研發工作，它意在推動整個數據中心進行創新，涉及機架設計到新存儲技術的方方面面。

????在本周三的開放計算峰會上，Facebook的基礎設施工程部副總杰?帕里克做了一個主題演講。其中談到，目前在數據中心采用閃存——更多用于像智能手機這類較小的設備——已是大勢所趨。實際上，開放計算新成員、企業級固態硬盤廠商Fusion-io公司已與潘多拉公司（Pandora）、Rhapsody公司和Ning公司等簽署合作協議【據大會新聞報道稱，Facebook和蘋果公司（Apple）這類客戶就更不在話下了】，并宣布將通過該項目公開部分硬件設計。

????Facebook有充分的理由推動數據中心技術提高效率和創新。它每天要存儲約3.5億張照片。處理并整理所有這些珍貴的照片——可能還需要用Facebook的新搜索功能篩選它們——對現有的數據中心技術來說實非易事。帕里克對《財富》雜志（Fortune）表示：“我可不能走到馬克（扎克伯格）那兒說，接下來兩年我拿這些數據中心沒轍。我們一貫深謀遠慮，行動迅速。”

????譯者：清遠

????One day after announcing a search feature that lets users quickly find information through their network, Facebook (FB) unveiled another potentially disruptive product: An open-source circuit board for servers affectionately dubbed Group Hug. Not surprisingly the aesthetically challenged piece of hardware generated much less hype than Facebook's earlier announcement. But could have just as much impact by changing the way data centers are built.

????Why? Facebook's mother of all motherboards allows chips from different manufacturers to be inserted into a common slot. That means a single server could house competing architectures, like Intel's (INTC) x86 technology or designs from U.K.-based ARM Holdings (ARMH). If adopted, Group Hug would be a radical shift in an industry that's known for building proprietary machines that don't tend to play nice together. And it could help pave the way for manufacturers of ARM-based chips to finally crack into servers, a market currently dominated by Intel.

????The Group Hug motherboard was developed at Facebook, but the social networking company is "open-sourcing" it to the world through the Open Compute Project (OCP), a group it spearheaded about 18 months ago in an effort to crowdsource more efficient data center hardware designs. Back then, the project had just one official member—Facebook. But on Wednesday, nearly 2,000 people gathered in a Santa Clara, Calif. convention center to participate in the fourth OCP Summit. The project now has over 50 members, including heavyweights like Intel, EMC (EMC), ARM Holdings, AMD (AMD) and SanDisk (SNDK). "It's really interesting to see all of these arch-rivals in the same room," says Frank Frankovsky, VP of hardware design at Facebook. "It's still somewhat contentious."

????Facebook isn't just pushing for new motherboards with interchangeable chips. Through Open Compute members and its own internal research and development, it's trying to drive innovation across the entire data center, from rack designs to new storage technologies.

????In a keynote address at Wednesday's Open Compute Summit, Facebook's VP of infrastructure engineering Jay Parikh said there's a big push to use flash storage—more common in small devices like smartphones—in data centers. Indeed, new Open Compute member Fusion-io (FIO) has signed on customers like Pandora (P), Rhapsody and Ning (not to mention Facebook and Apple (AAPL), according to press reports), and has announced it will open-source some of its hardware designs through the project.

????Facebook has good reason to drive more efficiency and innovation in data center technology. The company stores about 350 million photos per day. Processing and archiving all those baby pics—and presumably sifting through them with Facebook's new search function—is no easy task with current data center technology. "Going to Mark [Zuckerberg] and saying I can't do anything about the data center for the next two years doesn't work," Parikh told Fortune. "We think big and move fast."