HomeDigital EditionSys-Con RadioSearch Java Cd
Advanced Java AWT Book Reviews/Excerpts Client Server Corba Editorials Embedded Java Enterprise Java IDE's Industry Watch Integration Interviews Java Applet Java & Databases Java & Web Services Java Fundamentals Java Native Interface Java Servlets Java Beans J2ME Libraries .NET Object Orientation Observations/IMHO Product Reviews Scalability & Performance Security Server Side Source Code Straight Talking Swing Threads Using Java with others Wireless XML

JDJ: Could you describe your company, how it is organized and your position?
CS: My name is Chuck Scifers and I am the Marketing Manager for AboveNet Communications, Inc. AboveNet Communications, Inc. is divided into three product lines: one is WebCondo, another is ISPCondo and the third is MicroWeb Condo. In essence, we provide nonstop, noncongested connectivity to the Internet. This service is co-location tied to a unique business model and with our own proprietary facilities, tools, technology and service. We now have facilities in San Jose, California and Washington, D.C and we're talking about London and Hong Kong. It's just moving forward, it's just going so fast.

JDJ: AboveNet describes itself as being at the network level. Could you expand on that description?
CS: This really deals with our business model and why we are unique in the industry. First, you need to understand the telecommunications OSI seven layer model, the physical layer, the data link layer, the network layer, the transport layer, the session layer, the presentation layer and the application layer. The physical layer relates to the actual lines. The data link is the connection that hooks up those lines. The network ties those lines together. The transport layer moves the information all the way up to the application. Various people use different layers of that OSI model to do their business. Years ago, companies such as AOL, CompuServe and Prodigy would take the top five layers, from the network all the way up to the application, and sell that as a product. Later on, companies such as NetCom came along and would only provide access and the connectivity.

What has happened is equivalent to the computer industry back in the late '70s when you had the Osborns and the Apples. They made every piece that went into their product. They were vertically integrated. Now, anybody who tries to be vertically integrated doesn't stay in the business very long because the whole secret to staying economically viable is to specialize and produce one industry standard that can be used in many different products. Today, you find only a few companies that make chips, a few companies that make mother boards and the same for keyboards and monitors. Other companies will then put the pieces together. It gives the consumer a great product at a low price and also allows people to concentrate on core businesses.

The same thing has happened in the communication part of the Internet. Business is segmenting itself. It is going from a vertical integration model to a horizontal segmentation model. One example of this is companies choosing various layers of this OSI seven layer model. AboveNet, as an example, is not an ISP or a Web hosting company. We deal only with the third layer on the OSI model, the network layer. We provide connectivity and a world class facility for the content providers. We also provide space and connectivity for ISPs to come and put their routes so people can come in and then access content that is across the floor.

That's our model and we've launched what we call the ISX, the Internet Service Exchange. It's a combination of ISP condo and WebCondo. In the same facility, you bring in two hundred ISPs. Since we are not an ISP, we can bring these people in and not have to compete with them. We bring these ISPs regardless of size. There are a lot of small ISPs and they are giving the $19.95 a month access charge to their clients. They might have a thousand or a couple of thousand customers. They don't have the economies of scale to provide all the connectivity, redundancy and fault tolerance. However, they can come to our floor and use our connectivity, redundancy and fault tolerance as part of their facility.

The real beauty of the ISX product, the Internet exchange, is the more access providers we get on our floor, the less congested the Internet becomes. By the end of the year, we will have 30 percent of all traffic that goes through MAE West coming up through our facility. Right now we are at 8 percent and we have been in business for six months. One out of every three clients coming through the Internet will be accessing through our facility.

JDJ: How did AboveNet start?
CS: Our president, Sherman Twan, and the vice president of Marketing and Sales, Wayne Sanders, originally started a company called InterNex about three years ago. There were three people in the company in an office about the size of a closet and right now Internex is the largest digital ISP provider in the United States. They realized that there were certain things you could do in the Internet business that were profitable and could provide good service and there were other things that created an absolutely horrible service situation. After three years of building that business and watching the Internet grow, Sherman Twan had the vision to do what we are doing now. He wanted to get away from the dialups and the people who call in and say, "Hey, I can't get my modem to work" or "My Web Page isn't downloading". Those kinds of things require massive service calls and a massive service staff that is always running and trying to answer 100-300 trouble tickets a day. He built his business model saying, "What we want to do is provide a third OSI layer, network-only situation. We'll provide the facility. We'll provide the redundant connectivity. We'll provide the proprietary tools. We won't enter into the part of the business that is very service heavy and service-oriented."

JDJ: Could you describe the physical characteristics of your Web Condo, Micro Web Condo and ISP Condo.
CS: We are located in San Jose, California in the Fairmont Plaza Building, which is the newest building in downtown San Jose. It is 18 stories high, the highest in San Jose. It's designed with all the seismic requirements that are stipulated by the state and by the county to survive an earthquake, because we are on a fault area. When it was built, fiber was brought into the building. We are about 100 yards from MAE West physically and we have a router there so we are a zero hop to the MAE. MAE West is the Metropolitan Access Exchange. There are four major MAEs in the United States and they were formed when the NSF got out of the Internet. There is one in Washington, DC, one in Chicago, one in Dallas, TX and one in San Jose. MAE West runs consistently at about 750 megabits per second and it handles at least 50 percent of all the traffic from the Rockies west and the Asian territory. This NAP, which is a National Access Point, is where all of your providers, MCI, Sprint, UUNet, GoodNet, link up. When you send out information on a Web site, it goes into the MAE and that is where those packets are exchanged with the carrier or provider that will carry those packets to their destination. It is like a meeting point where all these packets are exchanged.

JDJ: Can you talk about any of your current clients?
CS:We aim our business at the mission-critical Internetcentric businesses. Those are people that make their money on the Web. For example, we have a company called Electronic Arts here and they are a video on-line gaming company. If they are down, they are out of business, they aren't generating any revenue. So, it's critical that they are up all the time, not congested, don't have brownouts. We have commerce sites where people make money off the net. If they are not up, they are not making money. We have other companies where, even if they are not making money directly from the Internet, their Web sites are so important that if they are down, they are in trouble. A lot of our companies are hosting companies. They have clients that they are providing presence for on the Internet. If they are down, their clients are down and consequently their clients are going to be very unhappy with them. So, our clients are Internetcentric. We have a company on our floor called Supernews, which came on our floor about four months ago with three servers. Now they have thirty servers on our floor and they are pumping out about 22 megabits of bandwidth on a constant basis.

JDJ: How do you deal with the bandwidth problem?
CS: Every time a company needs more bandwidth, we just crank them up because we have the tools to do that. If they want another three megabits of bandwidth, we can tune them up in twenty minutes whereas if they were offsite and needed to have Telco bring in a couple of T1s, you may be talking eight weeks. If they want three megs or five megs we can, because of our technology, give them the bandwidth. Another benefit of that capability is for your streaming audio and video. We've got a company, Mediacast, which does the streaming video events. Sun was introducing a new router/server and they did a worldwide streaming video event that required major band width, no latency and clean throughput.

One of the problems for companies like Mediacast is that they may use no bandwidth for the month and then need twenty megabits for a weekend. They can come to us a day ahead of time and say, "I'm going to need twenty megabits of bandwidth over the weekend" and we'll turn them up, let them use it over the weekend, turn it back down afterwards and charge them only for what they have used. There is another event coming up shortly which is going to involve Mediacast. They also will have streaming audio and video with one of the big sneaker manufacturers. They are going to need 200 megabits and that is going to increase the bandwidth through the MAE by about 25 percent. One of the ways we can provide this service is because of the amount of bandwidth that we purchase from various providers. We have redundant multiple providers such as MCI, Sprint, UUNet, GoodNet and multiple NPPs. We've got 90 megabits right into MAE West. We have another DS3 through to Alternet, two to the Palo Alto Internet Exchange and another to the Ames research center. We also have multiple connections so that we are not dependent on any one point of failure and that is very necessary in terms of connectivity.

However, there is a second problem that needs to be addressed and that is whenever you have multiple connections, you need some sort of routing device that will route the packets over those connections. The industry standard is BGP4 or Boarder Gateway Protocol-4 and it's a router to router protocol that will look out over your connectivity and say whether a connection is down. It will then try to send the packet out over other connections by the shortest open route. The problem with BGP4 is that it doesn't recognize if a router is showing packet loss. With the all the streaming video, audio and graphics that come down, you might get a router that starts to jam up and you will start losing packets. Well, the normal protocol will say, "Gee, I haven't heard that the packet went through so I had better send it again." The sending router will keep sending packets until the whole thing just melts down. With BGP4, you can have 90 percent packet loss and the router, the BGP4, will still be sending packets out.

We've developed our own Dynamic Packet Flow Management Protocol and we have the tools that allow us to go out and look at all of our connections to the Internet. If we see a router out on Sprint that starts to show a 20 percent packet loss, we automatically send an e-mail to Sprint saying, "Router number is showing 20 percent packet loss. You've got a problem there." From that point, we automatically monitor it and in twenty minutes, if it doesn't get better or it gets worse, we automatically shut down that connection and route around it. We may route packets out through MAE East and come back through the back door in order to get the packets through.

Granted, the Internet is oversold, pretty fragile and has a lot more traffic than it was designed for, but that doesn't mean that we shrug our shoulders and say, "Oh, the MAE is down or the MAE is congested. Wait twenty minutes and maybe your Web site will be able to download." We don't accept that. Part of our business model is that we want nonstop, noncongested connectivity. Any of the routes that we are using and any of the routes that we are going through is dynamically monitored and managed. We feel that if you use the proper technological tools and proactive management, you can solve a lot of the congestion problems that are out there. I think that most people just accept it because that is what they are used to, but it doesn't have to be that way.

JDJ: You have advertised a very wide range of services from a $99/month introductory offer to the "platinum" group. Are you setting yourself up to be the provider of choice for all Internet transport services, all customers, and if you are, what does a client get for that introductory offer as opposed to your "high rollers"?
CS: The reason that we are offering the $99/month is that we got into an alliance with Cisco systems, who provide 75 percent of the router market. They provide an Internet appliance called the MicroWeb server. We have that on our Web site and it's a new type of network appliance that allows somebody that is nontechnical or currently being hosted to buy a small Web server. It doesn't do a lot, but it will get a Web page up and running with some nifty capabilities and some pretty good throughput. Cisco came to us because the product itself is not complete unless it is hooked up to the Internet. They knew what we were doing and they asked us to help market these small servers. That's why we started MicroWeb condo. The $99 service that you get through MicroWeb Condo includes leasing the MicroWeb server, which holds 100 megs of Web pages, shelf space in our facility, which has our DC power supply, redundant connections, FM fire suppression system and 24 hour security and you get 64K of bandwidth. That bandwidth can be increased at 50 cents a kilobit/month so you could run a T1 through the unit if you want. Basically, all of the other services that apply to our Fortune 2000 companies apply to that $99 month as well.

It really wasn't part of our business model to provide this type of service to small Web servers. We were really going after the mission-critical companies that we had designed our facility and technology for, but we wanted to work with Cisco and this was a way to do it. The major difference between the smaller customer and our larger ones is the amount of bandwidth that they want to purchase. Otherwise, they are actually getting the same facilities and robust connectivity that the person who's coming in and ordering 50 megabits would get.

JDJ: Sounds like a real bargain.
CS: I'll tell you honestly that what we are really getting out of the MicroWeb server situation is our alliance with Cisco. If it weren't for that, we certainly wouldn't be at that end of the marketplace.

JDJ: How do you handle backups and hot sites. If worse came to worse and there was a catastrophe, how would you restore service?
CS: What kind of catastrophe are we talking about?

JDJ: Something on the order of Oklahoma City.
CS: That's an interesting point because we do have video and 24 hour security in our building. Our facility is card key entry only. It's a very secure facility, but if somebody came in and blew up the building it would be an interesting point. We are opening in Washington DC at MAE East, which is the second largest MAP in the world. We will have a duplicate facility that is a mirrored facility there of what we have in California. We've already got clients who say, "I like what you are doing here, but I want to be mirrored on the East coast." As we see the Internet grow, you are going to see people who will want redundancy in their server sites just as they want redundancy in their power supplies, connectivity and servers. There may be a market for people who want to have servers here and back East. One reason might be for a catastrophe like that. You would want to have that same mirrored server giving you closer connectivity on the East Coast and being able to handle the full load if something happens to the California server.

Some major companies want their servers in-house for reasons of security and convenience, but they will mirror a server here so that if, say, the Telco line that goes to their facility goes down, they can say, "OK, switch over to the Web server at AboveNet" and continue operations.

JDJ: Do you do backups to the servers and send tapes offsite too?
CS: One of the services that we provide as an option is to do tape backup daily and send them offsite where they can be archived elsewhere. In fact, we have our AS-UR-HERE service. What we do is make the client feel as if they are right with their server. We have clients from Argentina, Germany, England and Saigon. Obviously, they are too far away to service their server, so we have some proprietary tools that will allow the offsite client to call in and remotely reboot his server, in addition to other remote operations. One of our services is that we ping his server and every five minutes get a report back. We can also probe every program that is executing. If an HTTP, FTP or SMTP program is running and we don't get a ping back every five minutes, we automatically e-mail and page our client, saying, "You've got a problem with your server. Turn on your Web site and make sure everything is running." You can check your bandwidth and observe your performance by logging into your Web site.

This is part of our Ethervalve technology. Our client will know that he has a problem with his server long before a customer does. Ethervalve is part of the proprietary technology that was developed by our CTO and one of the things it does is allow us to provide guaranteed bandwidth. For example, if a person buys one meg of band width, we can guarantee that nobody will get in on their bandwidth. That is important because if you go to any other ISP and ask for a meg of bandwidth, that ISP will take a 10 meg segment of Ethernet bandwidth, put up to 20 or 30 clients on it and sell them each a meg. If everybody wants to use it at the same time, they have a real problem. That's part of the congestion we are trying to solve. We really have a guaranteed channel in there. With our proprietary technology, we can guarantee anywhere from a 64K channel up to a 100 megabit and once the person says, "I want that channel of bandwidth," it's theirs. The client can log into his Web page and actually see the amount of bandwidth that is going across his server.

As an example of how useful this can be, IUMA, the Internet Underground Music Archive, has been doing streaming audio for a long time and they are pretty well known. At their previous provider, they were buying five megabits of bandwidth and they were still getting very sluggish activity and performance. They felt that they had better buy more bandwidth. We convinced them to come in with us and check to see if they were really exceeding their bandwidth requirements. They came in and we gave them five megabits and their system worked fine and got great reviews. When they checked their Web site in terms of their bandwidth being utilized, they found that they had never used over two megabits. So we were able to turn them down to the two megabit level. The neat thing was that they can see exactly what they are using and manage those assets effectively.

JDJ: Do you have competitors for your niche or are you pretty well it?
CS: We have competitors who are colocation, but nobody else had the foresight and the funding that was needed to build up our facility, DC Power supply, FM200 fire suppression, robust connectivity and absolute first class security. Our facility is the best and I can say that honestly. Everybody else has come into colocation from being an ISP and have this legacy of dialup accounts that's killing them. So, that's our competition, but not really our competition.

JDJ: What's it like being one of the AboveNet employees? Do you have any special activities or hobbies that you do together?
CS: If I ever had any free time, I might get together with some of the others, but right now everybody is focused only on the mission. We are a startup and we are trying to do that technological American dream where you build a company and become successful. We don't have many people, but the ones we do have show a synergy that allows us to do a lot more than each of us could do individually. We are working long and hard hours and everybody is just dedicated and focused on it.

Sherman Twan is a bright guy and a visionary and we line up behind him and march forward. I think that we have a really formidable business model. It has a good foundation and it's worked. This industry moves so fast that if you do a business plan and analyze it and then start your business, you are a year behind. What happened here is that they saw where the Internet was going and went way out and planted that stake, hoping that the Internet didn't take a right turn and pass them by. Now, every ISP is trying to put in a colocation company. We know that that part of the business model was right because everybody is following it. I think that the ISX model that has just the transport layer will allow us to bring in other ISPs without competing against them.


All Rights Reserved
Copyright ©  2004 SYS-CON Media, Inc.
  E-mail: [email protected]

Java and Java-based marks are trademarks or registered trademarks of Sun Microsystems, Inc. in the United States and other countries. SYS-CON Publications, Inc. is independent of Sun Microsystems, Inc.