The mind is a terrible thing to waste. However, we then get the type advance can come along once in 25 years. Marc Andreessen had already shown us that he was capable of creating the type of advance that comes along once in 25 years. Few expected him to do so repeatedly, but many now feel he is doing that.
You can see a measure of Andreessen’s greatness in this Wikipedia capsule by Bob Metcalfe, the inventor of Ethernet networking.
In the Web's first generation (1990), Tim Berners-Lee launched the Uniform Resource Locator (URL), Hypertext Transfer Protocol (HTTP), and HTML standards with prototype Unix-based servers and browsers. A few people noticed that the Web might be better than Gopher.
In the second generation (1993), Marc Andreessen and Eric Bina developed NCSA Mosaic at the University of Illinois. Several million then suddenly noticed that the Web might be better than sex.
In the third generation (1994), Andreessen and Bina left NCSA to found Netscape...”
The Netscape web browser had 90% of web users for many years. This scared Microsoft so much that it illegally tied its browser to Windows, as detailed in antitrust litigation. Intuit founder Scott Cook testified that Microsoft paid to make QuickBooks use Internet Explorer. Despite this, Andreessen got AOL to buy Netscape for $4.2 billion.
Andreessen has since been a key to many important startups. He is now on the board of Facebook, eBay, HP and others. Some now say that his latest startup, Nicira (pronounced nice era), involves the type of advance that comes once in 25 years. The Nicira program can replace many changes in the latest Cisco and Juniper high-end network hardware.
One computer can use VMWare software to act like 10 or more computers at once. Each “virtual machine” has the properties of a physical computer. This ensures uses each machine in the most efficient and cost-effective way. Customers using cloud services can quickly create new virtual machines as needed, usually within minutes.
Networking generally has not kept up. The pipes through which bits go in and out of data centers have gotten faster, but not smarter. Cloud servers are flexible, precise and easy to manage, but networks seem like blunt instruments. Meeting new demand means adding new capacity, which usually means new hardware, which usually takes weeks or longer.
Nicira wants to “spin up” virtual networks as readily as you do virtual machines. It does this not by adding hardware, but by adding software to existing servers. You do not even need especially advanced network hardware. This gives big companies new flexibility in managing network infrastructure, particularly as needs and demand peaks and drop, by the day, season or year. Of course, some of the big companies will make this available to companies of all sizes.
Nicira calls its product a NVP (network virtualization platform). Some say such an advance comes once in 25 years. The argument for this may hold water. Where you deploy an application, in a data center, is as much a function of how much networking capacity you have as it is one of computing capacity.
Virtualization of servers lets you spread one app over as many physical machines as you need. However, if the network connecting the machines is inadequate, you can add new routers and switches or live with it. This means you cannot as flexible in deploying apps as you would like and some machines can be substantially under-utilized. You have buy more servers and pay to run and cool them.
Nicira “decouples” a virtual network from physical network hardware. You apply the intelligence, control and services in virtual space. A dumb networking pipe, carrying bits to two virtual machines, can now act in according to virtual rules. In the same way that one computer turns into a dozen, one network can act like dozens of networks. You also can cobble several networks together to act as one. You can create these virtual network on the fly, in minutes, just as you can create virtual machines.
This is already producing around 50% network cost savings for companies like AT&T, eBay, Fidelity Investments, Rackspace and Japanese telecom giant NTT.
I have everal issues with this once in 25 years idea. Metcalfe internet generations took as little as one year. It also has been only 17 years since the once every 25 year creation of the web. This also is the second once every 25 idea from Andreessen, who is one of seven billion of us who might have such ideas. There also are those at VMware, Citrix and elsewhere, who constantly update virtual products. Nicira will compete with many such products.
In fact, it seems like many will invent this more of less simultaneously. It is only one more example of why Kurzweil extended Moore’s Law (the accelerating increase of computer power) to the Law of Accelerating Returns. In 2001 Kurzweil said that. “An analysis of the history of technology shows that technological change is exponential… we won’t experience 100 years of progress in the 21st century — it will be more like 20,000 years of progress (at today’s rate).”
This means our collective once in 25 years ideas will occur 8 times in every calendar a year. So hang on for the glorious ride, while you keep trying to innovate and adopt. By the way, you also are likely to have many productive years to see this happen. Babies born today could easily see 30,000 years of progress at year 2000 rates.