If nothing else, Arista Networks proves that two people can make more than US$1 billion each building the Internet and still be worried about its reliability.
David Cheriton, a computer science professor at Stanford University known for his skills in software design, and Andreas Bechtolsheim, one of the founders of Sun Microsystems, have committed US$100 million of their money, and spent half that, to shake up the business of connecting computers in the Internet’s big computing centers.
As the Arista founders say, the promise of having access to mammoth amounts of data instantly, anywhere, is matched by the threat of catastrophe. People are creating more data and moving it ever faster on computer networks. The fast networks allow people to pour much more of civilization online, including not just Facebook posts and every book ever written, but all music, live video calls and most of the information technology behind modern business into a worldwide “cloud” of data centers. The networks are designed so it will always be available, via telephone, tablet, personal computer or an increasing array of connected devices.
Statistics dictate that the vastly greater number of transactions among computers in a world 100 times faster than today will lead to a greater number of unpredictable accidents, with less time in between them. Already, Amazon’s cloud for businesses failed for several hours in April when normal computer routines faltered and the system overloaded. Google’s cloud of e-mail and document collaboration software has been interrupted several times.
“We think of the Internet as always there. Just because we’ve become dependent on it, that doesn’t mean it’s true,” Cheriton says.
Bechtolsheim says that because of the Internet’s complexity, the global network is impossible to design without bugs. Very dangerous bugs, as they describe them, capable of halting commerce, destroying financial information or enabling hostile attacks by foreign powers.
Both were among the first investors in Google, which made them billionaires, and before that they created and sold a company to the networking giant Cisco Systems for US$220 million. Wealth and reputations as technology seers give their arguments about the risks of faster networks rare credibility.
More transactions also mean more system attacks. Even though he says there is no turning back on the online society, Cheriton worries most about security hazards.
“I’ve made the claim that the Chinese military can take it down in 30 seconds, no one can prove me wrong,” he said.
By building a new way to run networks in the cloud era, he says, “we have a path to having software that is more sophisticated, can be self-defending and is able to detect more problems, quicker.”
The common connection among computer servers, one gigabit per second, is giving way to 10 gigabit connections, because of improvements in semiconductor design and software. Speeds of 40 gigabits, even 100 gigabits, are now used for specialty purposes, such as consolidating huge data streams among hundreds of thousands of computers across the globe, and that technology is headed into the mainstream. An engineering standard for a terabit per second, 1,000 gigabits, is expected in about seven years.
Arista, which is based in Santa Clara, California, was built with the 10 gigabit world in mind. It now has 250 employees, 167 of them engineers, building a fast data-routing switch that could isolate problems and fix them without ever shutting down the network. It is intended to run on inexpensive mass-produced chips. In terms of software and hardware, it was a big break from the way things had been done in networking for the last quarter of a century.
“Companies like Cisco had to build their own specialty chips to work at high speed for the time,” Bechtolsheim says.
Because of improvements in the quality and capability of the kind of chips used in computers, telephones and cable television boxes, “we could build a network that is a lot more software-enabled, something that is a lot easier to defend and modify,” he says.
For Cheriton, who cuts his own hair despite his great wealth, Arista was an opportunity to work on a new style of software he said he had been thinking about since 1989.
No matter how complex, software is essentially a linear system of commands: Do this and then do that. Sometimes it is divided into “objects” or modules, but these tend to operate sequentially.
From 2004 to 2008, when Arista shipped its first product, Cheriton developed a five million-line system that breaks operations into a series of tasks, which when completed, other parts of the program can check on and pick up if everything seems fine. If it does not, the problem is rapidly isolated and addressed. Bechtolsheim worked with him to make the system operate with chips that were already on the market.
The first products were sold to financial traders looking to shave 100 nanoseconds off their high-frequency trades. Arista has more than 1,000 customers now, including telecommunications companies and university research laboratories.
“They have created something that is architecturally unique in networking, with a lot of value for the industry,” says Nicholas Lippis, who tests and evaluates switching equipment. “They built something fast that has a unique value for the industry.”
Kenneth Duda, another founder, says: “What drives us here is finding a new way to do software.”
Duda also worked with Cheriton and Bechtolsheim at Granite Systems, the company they sold to Cisco.
“The great enemy is complexity, measured in lines of code or interactions,” he says.
In the world of cloud computing, “there is no person alive who can understand 10 percent of the technology involved in my writing and printing out an online shopping list,” he says.
Not surprisingly, Cisco, which dominates the US$5 billion network switching business, disagrees.
“You don’t have to reinvent the Internet,” says Ram Velaga, vice president for product management in Cisco’s core technology group. “These protocols were designed to work even if Washington is taken out. That is in the architecture.”
Still, Cisco’s newest data center switches have rewritten software in a way more like Arista’s. A few products are using so-called merchant silicon, instead of its typical custom chips.
“Andy made a bet that Cisco would never use merchant silicon,” Velaga says.
Cheriton and Bechtolsheim have known each other since 1981, when Cheriton arrived from his native Canada to teach at Stanford. Bechtolsheim, a native of Germany, was studying electrical engineering and building what became Sun’s first product, a computer workstation.
The two became friends and intellectual compatriots, and in 1994 began Granite Networks, which made one of the first gigabit switches. Cisco bought the company two years later.
With no outside investors in Arista, they could take as long as they wanted on the product, Bechtolsheim says.
“Venture capitalists have no patience for a product to develop.” he says. “Pretty soon they want to bring in their best buddy as the CEO. Besides, this looked like a good investment.”
Cheriton says: “Not being venture funded was definitely a competitive advantage. Besides, Andy never told me it would be US$100 million.”
Saudi Arabian largesse is flooding Egypt’s cultural scene, but the reception is mixed. Some welcome new “cooperation” between two regional powerhouses, while others fear a hostile takeover by Riyadh. In Cairo, historically the cultural capital of the Arab world, Egyptian Minister of Culture Nevine al-Kilany recently hosted Saudi Arabian General Entertainment Authority chairman Turki al-Sheikh. The deep-pocketed al-Sheikh has emerged as a Medici-like patron for Egypt’s cultural elite, courted by Cairo’s top talent to produce a slew of forthcoming films. A new three-way agreement between al-Sheikh, Kilany and United Media Services — a multi-media conglomerate linked to state intelligence that owns much of
The US and other countries should take concrete steps to confront the threats from Beijing to avoid war, US Representative Mario Diaz-Balart said in an interview with Voice of America on March 13. The US should use “every diplomatic economic tool at our disposal to treat China as what it is... to avoid war,” Diaz-Balart said. Giving an example of what the US could do, he said that it has to be more aggressive in its military sales to Taiwan. Actions by cross-party US lawmakers in the past few years such as meeting with Taiwanese officials in Washington and Taipei, and
Denmark’s “one China” policy more and more resembles Beijing’s “one China” principle. At least, this is how things appear. In recent interactions with the Danish state, such as applying for residency permits, a Taiwanese’s nationality would be listed as “China.” That designation occurs for a Taiwanese student coming to Denmark or a Danish citizen arriving in Denmark with, for example, their Taiwanese partner. Details of this were published on Sunday in an article in the Danish daily Berlingske written by Alexander Sjoberg and Tobias Reinwald. The pretext for this new practice is that Denmark does not recognize Taiwan as a state under
The Republic of China (ROC) on Taiwan has no official diplomatic allies in the EU. With the exception of the Vatican, it has no official allies in Europe at all. This does not prevent the ROC — Taiwan — from having close relations with EU member states and other European countries. The exact nature of the relationship does bear revisiting, if only to clarify what is a very complicated and sensitive idea, the details of which leave considerable room for misunderstanding, misrepresentation and disagreement. Only this week, President Tsai Ing-wen (蔡英文) received members of the European Parliament’s Delegation for Relations