American Internet Services (AIS) colocation facilities are telco and optically-neutral, and each colo facility is fed by two separate fiber vaults.
Fiber enters the San Diego colo facilities via two divergent routes and into a secure fiber pop. Individual carriers then hand-off the circuits and other colocation facilities in a specially designed Meet-Me Room, adjacent to the fiber pop.
Our nine providers give each San Diego colo facility the redundancy and bandwidth needed for any application. We feature gigbit connections from the following providers: AT&T, Global Crossing, Level 3, Qwest, Savvis, Sprint, UUNet, Wiltel and XO. Competitors may say that they have multiple backbones, but they can't actually show them off. When you tour any of our San Diego colocation facilities, your salesperson will take you inside the POP room and show you the backbones entering our San Diego data center.
In addition to our fiber carriers in our San Diego data centers, we have a dedicated private OC48 ring to One Wilshire Blvd. This gives us the ability to connect our San Diego colocation customers to any of 260 different fiber providers and carriers located in Los Angeles.
All this fiber comes into our San Diego data center via the most state-of-the-art equipment available. This includes the biggest and baddest Cisco hardware; redundant Cisco 6500-series switch routers and redundant gigabit Cisco 12000 switch routers With these routers, we have incorporated something that no ISP west of the Mississippi has; the Route Science Box. This $250,000 piece of equipment takes BGP (Border Gateway Protocol) routing to the next level. BGP looks at the least amount of hops it takes in getting traffic from point A to point B. It does not take into account the amount of jitter, packet loss, or latency. Route Science will direct ALL San Diego colo traffic uses parameters such as increased jitter, packet loss and latency to direct your traffic using the best route possible. Additional redundant Cisco 7500 aggregation routers assist in providing full connectivity to our T1, DS3, OC3, and higher bandwidth clients.
Two Major Mistakes When Evaluating a Colocation Facility's Bandwidth and Network
This is always a highly debated topic and everyone will have a slightly different opinion on the subject. I will attempt to point out some facts that will help you make an intelligent decision about choosing your Internet connection.
The Internet is a combination of thousands of small and large networks all connected together with routers making up one large network called the Internet. Many people are under the impression that, if you are connected to the Internet you are connected. That is true. However, when conducting business on the Internet, just being connected is usually not good enough.
Providers of Internet bandwidth come in two basic flavors. "Tier one" providers and "Tier two" providers.
Mistake #1 - Sub-Optimal Internet Backbones
Tier one providers
Tier one providers are companies such as AT&T, Level 3 and Sprint. These companies operate their own national networks, and in most cases, have colocation facilities as well. Their main goal in selling you colocation is to obtain your colocation business (space and power) plus provide your bandwidth. They will typically promote their services with facts about their great network, their low latency from coast to coast and all the redundancy built into their network systems. This is all well and good if your Internet traffic is bound from one point on their network to another point on their network. However, most colocation customers want to get to all points on the Internet in the fastest most reliable manner. Therefore there is no advantage to being on just one large Internet backbone, such as Sprint. When your Internet traffic has to leave that backbone and travel to another network, that is when you can begin to have problems with a "tier one" provider. Of course this will be the case with most of your traffic, because no matter which tier one provider you choose, the chances are that most of your traffic will go to other networks. This type of colocation facility is usually reliable and should provide good levels of service, although even with redundancy built-in, their networks sometimes go down and this creates outages. Their prices will typically be on the high side.
Tier one providers use public peering points where the largest providers exchange bandwidth with other providers at no charge. Everyone there is trying to dump off their traffic to other providers' networks so that they don't have to buy bandwidth. This results in sub-optimal routing, high latency and an overall poor performance when crossing networks.
Tier two providers
Tier two providers are generally smaller, independent ISPs or colocation facilities that do not have their own national backbone. They buy bandwidth from one or two "tier one" providers. If the colocation facility is only connected to one "tier one" provider, check them off your list and move on. If that provider goes down, you will be down, and it could be for hours. Usually tier two providers are connected to two or more backbones, which is good because you get the redundancy of two tier one networks and all the benefits of the two tier one providers instead of just one. These providers use BGP to route traffic across the two or more backbones. However, be aware that those providers often have a primary carrier in which they have made a single large commitment for bandwidth, which provides low cost traffic, and they will tweak BGP controls to modify their routes, driving most traffic across the cheaper backbone, such as Cogent or Yipes. Make sure you ask about their routing techniques. Talk to the head network engineer on this matter as the salesman typically won't have a clue. If you are on one of these networks, and all your traffic goes to one provider, you can be sure that this is what they are doing. Usually, these types of tier two providers will have the lowest prices and the least reliability. If you are simply shopping price, this is the place to start. Like with most things, you will get what you are paying for.
Carrier Neutral Facilities
Some colocation facilities are what are called carrier neutral. This means that they have fiber connections to multiple tier one backbones and you can connect (provided you pay the cross-connect fee) to any backbone available in their facility. This means that the colocation customer will have to provide a router, handle all the BGP and routing issues, and buy bandwidth directly from the tier one providers in the facility. Unless you have a high end network engineer on staff that knows BGP very well, this is not recommended. Focus on your online business and let the backbone experts deal with routing issues. This type of colocation can be the most expensive, the least reliable, and the most headaches for the average online business. Typically, tier two facilities will allow you to be "carrier neutral", but you may have to ask for it because they want to sell their own bandwidth. Cross connect fees don't usually pay the bills, so they will be pushing their own bandwidth, and in most cases you are better off buying bandwidth from the facility, rather than trying to do it yourself.
High performance tier two providers
High-end tier two colocation facilities will offer multiple backbone connections to tier one providers. Anything more than four backbone connections is considered phenomenal and is out of the ordinary. However there are some that offer more than five or more tier one backbones. This is akin to doing business with, and getting the benefits of all 5 of the tier one backbones, and it results in very low latency, low packet loss and a high degree of reliability. A very select few of these tier two providers have implemented a new technology that is known as "performance based routing". This means that they have invested in hardware and software that talks to their routers and makes routing decisions based on the performance of the various networks to which they are connected. Most ISPs use BGP (Border Gateway Protocol) to route traffic according to the least number of AS hops between the various networks. These "performance based routing" hardware and software based systems can cost between $100K and $350K depending on the manufacturer used and the number backbones. Performance based routing hardware can easily deliver as much as a 50% increase in network performance and also create a self healing network. Don't be surprised, if you ask about it, and they say "Performance what?". The technology is new and expensive, therefore most small colocation facilities won't even know what you are talking about and if they do they can't afford it and will tell you all the reasons why it does not help or won't work.
This advanced technology is highly desirable for companies that need the highest levels of reliability and performance from their Internet backbones. This type of provider is typically mid-priced, offers the highest degree of network reliability and the best network performance. Not all areas will have this type of provider, but most of the major metropolitan areas such as Los Angeles, San Francisco, San Diego, Denver, Dallas, Chicago, DC, Atlanta and New York and Miami will have at least one such facility. If you can find this kind of network with a data center that has all of the power redundancy built-in, it can often be your very best choice for colocation. Not only are these facilities highly reliable, they typically offer a high degree of customer service that you won't find at the tier 1 providers and competitive pricing.
Another advantage of a tier two provider is typically better and more personalized customer service. If the provider is locally owned and operated, you can often get the CEO and CTO's cell phone numbers, especially if you are a large client. They typically only have between 100 and 200 customers, so don't be afraid to ask for those numbers if it is not on the escalation matrix.
Mistake #2 - No redundant routers and or switches.
If you want maximum uptime you must find a facility that offers multiple connections to their network. A single Ethernet handoff to a single switch is just asking for downtime. Switches not only have failures, but they often need to have software upgrades. Both of these situations will create downtime. If they have enterprise switches and routers, it could be called HSRP (Hot Standby Routing Protocol), VRRP (Virtual Router Redundancy Protocol) or GLBP (Gateway Load Balancing Protocol). If they have another manufacturer, they may call it something else. But make sure you can get dual Ethernet handoffs from redundant switches or routers.
Make sure those redundant switches are connected to redundant backbone routers that are connected to their redundant tier one backbones. An added feature would be redundant engines within the routers and switches, and redundant power supplies connected to the redundant power grids.
open & close article 
