What is Bandwidth?

Written by: Shop For Bandwidth - Oct 15, 2008

Q: What is Bandwidth?
Bandwidth represents the amount of data a network can transport in a certain period of time. For digital devices, bandwidth is usually expressed in bits or bytes per second. For analog devices, bandwidth is expressed in cycles per second, or Hertz (Hz).

Q: Why is there a tendency for small- and medium-sized businesses to buy more bandwidth than they need?
The first reason is that Internet bandwidth is so inexpensive. For less than 5,000//year, a small company (up to 50 employees) can get all the bandwidth it might need. However, if the company does a lot of data transfer, they may need more than one T1 line.

The second reason is that ISPs are glad to sell high-speed connections, especially to companies that don't consume much bandwidth. In fact, the amount of bandwidth a typical customer really uses is a well-kept secret among ISPs. One ISP's chief technology officer revealed that his company provisions its network assuming that the typical client will use less than 64 Kbps of bandwidth, 24 hours a day-a profile that most clients would fit unless they're hosting something like a specialized, high-traffic website.

Q: How do I determine my bandwidth needs?
One approach to determining your bandwidth needs is to conduct the following four-step analysis:

• Determine baseline usage : First of all, measure current network utilization levels and traffic patterns. Network performance management software can monitor and measure network activity levels, report this data in easy-to-interpret formats, and help estimate future demand. However, avoid conducting quick baseline studies that don't measure traffic patterns for a full business cycle-it may be necessary to measure activity for an entire quarter. Without collecting data for a full cycle, you may not capture all the information you need.
• Benchmark competitors : Compare your utilization levels and traffic patterns with your industry peers or companies with similar business patterns.
• Perform modeling/forecast analyses : Using your findings from the baseline and benchmark analyses, determine technology requirements under various scenarios by using modeling programs. These programs enable you to test alternative bandwidth solutions, estimate relative costs, and decide on the best course of action.
• Assess your processes : Process assessment involves determining how to optimize the performance of your network(s) and ensure cost-effective bandwidth utilization. Often, the most important aspect of network and systems management involves process assessment rather than purchasing additional technology.

Q: What if my bandwidth requirements change frequently or don't remain constant?
When you need more capacity, you often have to buy it in the next available increment, which may end up costing more than you want or can afford to spend. Worse yet, with current, "bursty" application requirements, the need for additional bandwidth can often be as temporary as it is critical. By providing flexible capacity, or "bandwidth on demand," service providers can help customers deal with variations in traffic, and with avoiding a high fixed monthly payment. True bandwidth-on-demand services should allow for a wide range in capacity.

Q: What does Quality of Service (QoS) mean in the context of bandwidth management?
Because different applications have varying requirements for bandwidth, delay, and jitter (variances in delay), Quality of Service provides network managers with the mechanisms to monitor and control these performance factors. For example, voice applications have stringent delay requirements and can tolerate minimal packet loss. Alternatively, an FTP file transfer may be minimally affected by delay but very sensitive to packet drops.

Q: What is a service level agreement (SLA)?
A service level agreement is a contract that sets performance goals as well as penalties for failure to meet these goals. In other words, an SLA commits a service provider to sharing responsibility for service reliability. Though SLAs often come with their fair share of hype and fine print, some do credit customers with a refund for every hour or portion of an hour of disrupted service. Providers may also offer credits if they miss a delivery date or don't report an outage promptly.

You may want to incorporate some additional provisions in your SLA, such as:

• Negotiate a more reliable connection that includes alternate routes for your data.
• Create on-call backup circuits.
• Attempt to specify more stringent penalties for service provider failures.
  

Because the terms of an SLA are negotiable and can vary significantly, they should be discussed up front with the provider.

Q: What types of provisions should be addressed in a service level agreement?

• Outage duration : The amount of time in minutes that service is unavailable
• Degraded service : Service which is slower than the performance specified in your contract
• Defects per million : Minutes of downtime per million minutes of service
• Mean time between failures (MTBF) : Average amount of time, typically in minutes or days, between outages. (The target figure depends on your negotiated total-service-availability rate.)
• Mean time to restore (MTTR) : Average amount of time, typically in minutes, required to restore service
• Maximum time between failures/maximum time to restore : A cap on the total number of minutes for restoration of service
• Trouble rate : How often technical support needs to be contacted. (This should be limited to five times before escalation is mandated.)
• Average round-trip latency : The time required for the first transmission to be completed. (The target should be less than 100 msec on the backbone and up to 130 msec longer from end to end, depending on the type of service being contracted for and on application requirements.)
• Average round-trip delay : The time it takes for routine transmission after the first transmission establishes the connection. (The target should be the same as the average round-trip latency.)

Q: What additional guidelines are recommended for negotiating a service level agreement?
Pricing:

• Expect wholesale prices to fall at least 50 percent per year on competitive routes.
• Avoid long-term "IRU" or capital-type leases unless the lease payback is less than three years.
• Include regular benchmark price reviews in contracts longer than one year.
• Look for contracts that allow you to buy network capacity and that offer discounts for larger volumes, but which also allow you to alter the routing later at little or no cost.
• Monitor prices and availability of dark optical fiber (unused fiber-optic cable), especially on terrestrial routes.

Quality and Reliability:

• Avoid multiple, service provider supply chains unless no alternative exists or the financial benefits are substantial.
• Find out to what extent the network fiber is actually owned by the provider. If it's not, request information on its type and performance.
• Inquire about and continue to monitor emerging bandwidth-on-demand products that might help you meet unexpected peaks in demand.
  

Q: Is the cost of bandwidth decreasing?
The market forces that so profoundly affected voice prices and service in the 1980s and 90s are being brought to bear on data and multimedia in this decade. If bandwidth supply continues to outstrip demand, as is currently the case, bandwidth prices will have no place to go but down.

In the past couple of years, the average price for transmitting a megabyte of data over domestic or international lines fell by 25 percent on average according to market data from an online bandwidth exchange vendor. However, in some remote locations, prices have not changed much since the local loop charges have remained constant.

Q: How do I avoid overspending on Internet bandwidth?
The following four basic guidelines will help you avoid overspending on Internet bandwidth:

• Understand how much bandwidth you really use. For instance, use the SNMP (Simple Network Management Protocol) statistics from your router to monitor usage.
• Buy a usage-sensitive connection. Being able to accommodate bursty traffic at times is fine, but if you're only going to use 64 Kbps, then that's all you should be paying for.
• Find an ISP that will offer you a flat service, which you will know the flat fee you're paying instead of the burstable service.
• If possible, instead of having one "large" connection, consider buying two smaller connections for disaster recovery and load balancing reasons.

Q: What tools and technology are available for managing bandwidth?
Aponet Inc., CheckPoint Software Technologies Ltd., and Concord Communications are among the companies that offer bandwidth management solutions. The move to high-bandwidth switching architectures can also bring new management headaches, which has led to the introduction of specialized monitoring and troubleshooting tools such as Fujitsu Software Corp.'s NetPrism and RiverSoft Ltd.'s OpenRiver.

Increasing volumes of network traffic spilling over onto WANs are also causing more Frame Relay and other data networking service customers to monitor their traffic patterns and usage volumes. A variety of mechanisms are available for monitoring traffic:

• Enhanced CSU/DSU (channel service unit/data service unit) products that monitor usage and generate reports
• Bandwidth managers that shape and prioritize traffic
• Carrier monitoring and reporting services
• Carrier customer network management systems (CNMS) that enable customers to make changes to their service on the fly
  

Q: Is managing network usage an option for managing bandwidth?
Yes, and this can be accomplished through usage-based chargeback, an idea gaining favor as organizations realize that as long as bandwidth is perceived as being free, users won't have any particular motivation to conserve it. Systems such as NetCountant enable IT departments to tariff utilization by volume, protocol, and other parameters.