WDTalk

BGP is a mixed blend of carriers that allows data center clients to take advantage of that center’s backbone connections to the Internet to reduce latency.

Technically, it’s generally available across all router platforms and relates to the forwarding process in a router. It’s based on standard routing protocols, such as Routing Information Protocol (RIP), RIPv2, Enhanced Interior Gateway Routing Protocol (EIGRP), Open Shortest Path First (OSPF), and Interior Gateway Routing Protocol (IGRP), or derived from statically configured routes and packet forwarding mechanisms. It’s automatically activated if the routing table has multiple paths to a destination.

Gaming applications should always opt for BGP bandwidth to reduce latency.  

Bandwidth Plans Vary by Provider

Generally, providers offer bandwidth in dedicated, shared, metered or unmetered packages. Obviously, dedicated plans are all yours, while shared plans are distributed with other users on their port. Prices can vary widely, so caution is recommended when soliciting requests, to best match a provider’s bandwidth services to a realistic estimation of the bandwidth you’ll require.

Unmetered plans are capped at whatever size bandwidth pipe you’ve provisioned. Typically, in a shared unmetered plan, providers will guarantee a minimum amount of bandwidth, but no maximum. Dedicated unmetered plans are much more expensive.

What is the 95^th Percentile

Billing on the 95^th percentile is a viable option if you anticipate bursting above your allotted bandwidth. For example, you may be paying $15/Mbps with the first 10Mbps included in your plan. Overages would be calculated on a 95% percentile at $15/Mbps. If you used 12Mbps over the month, you’d subtract the 10Mbps in the plan, leaving 2Mbps billable on the 95% at $15/Mbps.

Bandwidth explained 

Bandwidth isn’t as straight-forward as you may think. For some it is the total amount of data, typically measured in Megabytes, Gigabytes or Terabytes, that may be downloaded or uploaded during a given month – also referred to as data transfer. 

In a typical web hosting account, an entire website (with all of its files, databases and images) comprises that websites space (storage). If a website’s hosted plan allows 50GB in storage, and that site’s content takes up 2GB, then there is 48GB of storage left. If that same plan allows 200GB of bandwidth per month, when you upload the site, the host subtracts that 2GB transfer leaving 198GB of bandwidth remaining. 

For simplicity, if 99 visitors viewed every page of your site monthly, they would each transfer 2GB of that bandwidth, leaving your account maxed out. The host would either then cap your account or charge you additional bandwidth fees for the 100th visitor, and so on. But this never happens, and the reality is that only certain pages are viewed – a small fraction of that 2GB. 

Obliviously, the more popular that site becomes, the more bandwidth it uses. Assuming the most popular content viewed is a menu, 50KB in size, downloaded 300 times daily, the transfer attributed to that menu would be 300 downloads x 50KB x 30 days per month = 450,000KB or 450MB (.45GB). 

Here’s the problem. More and more audio and video files are being placed on websites drawing thousands of visitors. A two minute video can easily consume 10MB of transfer. Take the same 300 downloads x 10MB x 30 days per month = 90,000MB or 90GB. Remember that 300 downloads in 30 days only averages 10 per day. What happens when that number reaches 100 per day or 1000? 

Some hosts feature straight data fees. If you use 10MB, you pay for 10MB. If you use 20GB, you pay for 20GB.  Typical bandwidth plans are based on metered, unmetered, burstable and 95th percentile billing. 

Unmetered bandwidth 

Unmetered bandwidth means that the maximum data transfer rate is capped at a specific speed, but the amount of data transfer at that speed is unlimited. The cost for unmetered bandwidth is normally based on a fixed monthly charge for bandwidth consumption payable at the beginning of a monthly cycle. 

There are dedicated and shared unmetered bandwidth plans. 

Dedicated or guaranteed unmetered plans offer bandwidth pipes available to you and you only, that you can max out at will. Some hosts offer burst options for overages on a 95^th percentile. 

Shared unmetered plans means the host shares your pipe with other customers. These types of plans typically provide a guaranteed minimum but not a guaranteed maximum. 

When operating at a speed of 1.54Mbps, a VPS is capable of a maximum 30 day transfer total of (1.54 Megabits per second / 8 bits per byte = .1925 Megabytes per second * 60 seconds = 11.55 Megabytes per minute * 60 minutes = 693 Megabytes per hour * 24 hours = 16,632 Megabytes per day * 30 days = 498,960 Megabytes per month / 1024 bytes = 487 Gigabytes per month. If your requirements exceed 487GB monthly, 5/10/100Mbps plans are normally offered. 

Ten (10Mbps) equates to about 3.3 Terabytes of bandwidth and 100Mbps to about 33.3 Terabytes. It goes without saying that a 100Mbps unmetered dedicated plan is quite a bit more expensive than a shared plan. The shared plan is less expensive, but your server is connected with other servers (2-6) at a switch that is normally connected to a 100Mbps switch port, or in some cases with other servers (10-20) connected to a 1Gbps port. 

Metered plans 

The expense for metered bandwidth is calculated at the end of each monthly billing cycle. Metered essentially means your bandwidth usage is monitored and you’re responsible for any overages. 

Small servers with low bandwidth usage are normally billed at a straight data transfer rate. 

Measuring on the 95^th Percentile 

Another bandwidth plan uses the 95^th percentile method for computing bandwidth expense. For example, a 10Mbps plan billed at the 95^th relates to 10Mbps unmetered but the connection itself may be capable of 100Mbps. This enables your server to reach speeds up to 100Mbps (burstable). At the end of the monthly billing cycle, the top 5% of the speeds are removed, then the 10Mbps is subtracted, leaving any overages. In this case, if your server used 15Mbps over 95% monthly, you would incur an additional bandwidth expense of 5Mbps. 

Consistent traffic with a couple days of bursting results in far less bandwidth expense than, let’s say 10 days of bursting each month. The difference can be as high as ten times more. 

My Recommendation 

Use due diligence to determine your anticipated bandwidth before signing a contract.  If you’re unsure, ask your host if they’re flexible in renegotiating if necessary.

I read an interesting thread recently that continued on for 20+ pages related to bandwidth pricing. The original post was about blended bandwidth being offered for $3.99/Mbps at a 10Mbps commitment.

Is this extremely aggressive bandwidth pricing or some sort of marketing scheme?

To be sure, it’s very aggressive given no one in the industry comes close to $3.99. But, if you weigh the risk that not everyone will use that 10Mbps, then the law of averages leans in your favor.

Let’s say that client one uses 3Mbps (still pays for 10Mbps) and client two uses 8Mbps (still pays for 10Mbps). You collect for 20Mbps ($39.99 x 2 = $79.98). The actual cost to client one for bandwidth is not $3.99/Mbps. It factors out to $39.99 divided by 3 = $13.33/Mbps.

Client two’s actual cost for bandwidth is closer to the norm. Their actual cost would be $39.99 divided by 8 = $5.00/Mbps.  Now that’s a great deal.

But this is just a small part of the big picture.

Everyone in the industry knows the total cost of ownership is the combined cost of the cabinets/racks/power and so on.  Any losses on aggressive bandwidth pricing can be used as loss leaders by augmenting other services.

As a data center, you must evaluate the risks (long term) of marketing campaigns. Can you deliver on your bandwidth deals?  I’m very sure in the coming months, many many posts will circulate the forums about this offering. Stay tuned.

BGP load balancing allows data center clients to take advantage of that center’s backbone connections to the Internet to reduce latency.

Technically, it’s generally available across all router platforms and relates to the forwarding process in a router. It’s based on standard routing protocols, such as Routing Information Protocol (RIP), RIPv2, Enhanced Interior Gateway Routing Protocol (EIGRP), Open Shortest Path First (OSPF), and Interior Gateway Routing Protocol (IGRP), or derived from statically configured routes and packet forwarding mechanisms. It’s automatically activated if the routing table has multiple paths to a destination.

Gaming applications should always opt for BGP load balancing.