How the FB server poses a threat to HP Cisco

[Guide] hardware vendors are free to use this Open-source hardware to drive down the price of new server systems. A test rack in the Facebook hardware Power Test Lab (Tencent Technology map) Beijing Time July 29 News, the U.S. technology website Arctechnica recently published articles about how Facebook is using its "no flashy" server to HP , Cisco and other large companies in the IT field. The article points out that the reason why open source hardware can rapidly expand its scope of influence from the initial audience to more users, because the hardware manufacturers can be very free to adopt this hardware, thereby driving down the price of the new server system. The following is the full text of the article: the 17th Building at Facebook headquarters is located in the original campus of Sun Microsystems, dubbed the "Sun Quentin", which resembles California State Quentin Prison (San Quentin). Now, working in the building is a Facebook engineer team at the Facebook Power Lab. In every working day, they are pushing the company's ideas about how data center hardware should be built. These engineers are constantly testing the design of Home-grown server hardware--basically, Facebook's Home-grown server hardware is quite different from the hardware we normally know. Arctechnica recently visited Facebook's headquarters campus and toured the company's server labs under the leadership of Matt Caudri, a senior manager of Facebook's hardware engineering, and the head of the server hardware design team, Matt Corddry. What is happening in the Facebook lab is not just an impact on the company's data center, it is also one of the contributions that Facebook has made to open computing (Compute project). The purpose of this project is to bring open source design to the Global data center server and storage hardware, infrastructure and management interface. Facebook, Amazon, and Google are all picky about their server hardware, which is mainly about making their own hardware by buying parts and then developing them in their own way. Frank Frankovsky, vice president of Facebook's hardware design and supply chain operations, has Frank Frankowski the open Computing project because he recognizes the waste that large cloud companies will have in the process of re-creation of what they can share. Frankowski feels that introducing the Open-source principles that Facebook has been following in the software world into the hardware sector will save a lot of money for the company and other companies-a savings that is reflected in two areas, a direct hardware cost, and a maintenanceand power costs. Just as Raspberry Pi and Arduino open source microcontrollers have inspired the imagination of small hardware hackers, the goal of the Open Computing project is to make DIY easier, more efficient and more flexible at the macro level. In the end, what Facebook and open computing projects are doing to datacenter hardware may not be the risk of extinction for the entire hardware industry, but it will certainly have a certain impact on the industry. Yes, open source, commercialized motherboards and other subsystems used for Facebook were originally designed for the "Super scale" data center world, such as Facebook, Rackspace and other data centers for cloud computing services providers; These designs can easily find roads to other DIY hardware environments, or paths to "vanity free" server systems that are sold not only to small companies but also to large enterprises. The reason why open source hardware can quickly expand its reach and move from its initial audience to more users is because hardware vendors are free to adopt this hardware, driving down the price of new server systems. This may not be good news for HP, Cisco and other large companies in the IT sector. "No exaggeration", open source design of the server system is likely to promote the rapid development of innovation activities, so that HP and other companies have been on the development of the entire model has a devastating impact. Eliminate all the complexities. To be exact, an open computing project cannot go all the way to the CPU along the open source, even Raspberry Pi is not based on open source hardware, because none of the chips in the open design chip have enough power (for economic reasons, Manufacturers will not be willing to produce such chips. The hardware design for open computing projects is "open" at a higher level, and anyone can use standards-based components to create motherboards, bases, racks, rack parts, and other server parts. "We focus on the simplest possible design." Caudri told me. "Our focus is on very high scalability, eliminating all the complexities and abandoning any magnificent or grandiose design." "This means that the hardware created in this way is easy to maintain, production costs and prices are low, and can be handled relatively simply when problems arise." In addition, it is simpler to create something based on this design, which is helpful to Facebook and other companies that agree on the philosophy of open source computing projects. This can be demonstrated by Facebook's "Open Source computing project hardware Hacker Marathon" (OCP Hardware hackathon) competition at the headquarters campus on June 18, and some great ideas emerged from the contest. So far, open source computingThe main users of the project's hardware are Facebook and Rackspace, but that will change soon, since the momentum for open source hardware is starting to take off. As Intel, AMD and others have started to create more parts that meet the specifications of open source computing projects and are contributing more intellectual property to the project, some people believe that the open Source computing project community will soon become bigger and larger. "I think our industry probably realized about 10 years ago – when Linux was taking over UNIX sites – openness was a very good thing for consumers in the world of suppliers and even large computing." Said Delhi. "Linux did not overwhelm the data center industry or the operating system industry. I think we're seeing the same pattern today, and seeing open hardware doesn't mean the death of hardware, but it probably means the rebirth of hardware. In the process of rebirth, what we see is that the speed of innovation will be greatly accelerated, because we will not always waste effort to do some repetitive work. "Facebook's senior manager for hardware engineering, Matt Caudri (Tencent technology map) deconstruction server The core of the design philosophy of Facebook data center is" deconstruction "--that is, dismantling things that traditionally make up a" server ", A hardware group that makes it a specific purpose, which is largely connected to the network hardware. To some extent, such a thing happened in the Sun's park with a little irony. In Sun's heyday, the company's slogan was "The Internet is a computer," and today computers are networks, both conceptually and physically. The design principles behind Facebook's hardware come from direct hands-on experience. Caudri says all of his engineers have spent time working as a technician in the Facebook data center, "so everyone has hands-on experience and a sense of how the hardware Works". The approach of Facebook and open computing is a "postmodern deconstruction" of the data center, which is to disassemble the parts that normally make up a server, making it a part that functions, minimizing its complexity and maximizing its productivity. In the Facebook data Center architecture, the real meaning of the so-called "server" is very few, replaced by many of the various functional "sled" rack. "That will be a pattern that you will see from us for the next two years." "He showed us some sled designs in the hardware lab," says the test. "Many of our hardware designs will focus on a category of issues." "In the latest Facebook data center, the company has taken aMethod。 In these data centers, the server racks are full of systems, storage sled (high-density disk arrays), and "Memory sled" (systems with large amounts of RAM and Low-power processors), consisting of a variety of multi-purpose computing sled (motherboards equipped with CPUs, memory, and PCI cards to handle specific tasks). This processor is designed to handle in-memory indexes and databases. "We're not going to use this on the site. "He refers to a Web site that Facebook uses to connect to a large internet site," says Delhi. "But we will use it in all of our data center facilities, including the Virginia State Ashburn (Ashburn) data center, where the data center environment is more Classic mode." In fact, we have designed a series of initial open racks. "All of these decomposed hardware modules have a common feature that they are completely independent and can be disassembled, repaired, or replaced in a very different way." Caudri pointed to a calculation on the lab bench. Sled says: "If this thing needs fixing, then all you have to do is grab the handle and pull it out, because there's no screws and no screwdriver." "Windmill" based on Facebook computing sled (Tencent technology map)