Tired of users gobbling network bandwidth? Charging for usage of the network can save costs, but IT directors should be aware that it may create resentment among users
There is no such thing as a free lunch unless you are an IT network user. Users can eat up vast amounts of bandwidth downloading MP3s, sending e-mail attachments or videoconferencing and, as an IT manager, it all comes out of your department's pocket.
The irony is that mainframe administrators have been charging users for system resources for years. "It has been feasible in the past - we used to do it all the time with mainframes based on time sharing," said Gordon Haff, senior analyst at research firm Illuminata.
"It tended to fall out of use during the Unix and certainly the Wintel eras."
The move towards distributed systems has made it more difficult to monitor who is using what, said Dan Kusnetzky, vice-president of systems software research at analyst firm IDC.
With user sessions being routed along multiple paths to multiple alternate resources depending on system and network load, it becomes hard to unravel the mess and reach a definitive per-user cost for the use of network resources.
"How many machines were involved in a transaction and how can we come up with a reasonable charge?" Kusnetzky asks. He said that hardware and software suppliers may not have set up their systems to quantify how different components contribute individually. "You need a cultural shift to change that," he added.
This has led even those companies for which resource allocation is a core competency to be conservative in their approach. Bill Pepper is director of security risk management at CSC, which handles multiple customers as part of its outsourcing business.
But even he tends to levy an individual charge per seat, tiered depending on application configuration and storage use. Network traffic does not even enter the equation. "We do not bother with network traffic-based charging because it averages out over time," he said.
However, Forrester Research analyst Colin Rankine has seen a marked interest in distributed systems resource accounting from his clients, who want to charge not only for network time, but for the use of other resources such as storage and server time.
Rankine wrote a report in June which said that although administrators of distributed systems are unlikely to achieve the same accounting standards as those working in mainframe shops, there are still some things that they can do.
Minimising overheads is the prevailing goal for any such system, he said. Data collection must not affect system performance and the best way to avoid this is to keep it simple.
One possible way to minimise complexity is to use statistical sampling rather than trying to monitor every packet that passes over a network. This can work well as long as your environment is not too volatile.
There are suppliers addressing these issues. For example, Allot makes network appliances that shape internet traffic and adjust bandwidth for different application streams to provide users with a quality of service standard. It also provides network traffic accounting information.
Antoine Guy, director of worldwide product marketing for Allot, said that an add-in accounting module uses a MySQL database to store an IP detail record, containing information about application sessions.
"If you consider that an IP session between two addresses is like a telephone call, you can store the sort of information you would see on your phone bill," he said.
"For each connection you see a start time, an end time, the total amount of bytes transmitted, the application used and the protocol, along with the destination and the source."
But user IP addresses can present problems of their own if you use dynamic IP addresses to help make the network more flexible. One possible solution is to allocate groups of dynamic addresses to subnets used by organisational groups. That way, you can identify that a certain application stream was used by the human resources department, for example.
But Bill Pepper is not convinced. "What about hot desking?" he asked. "This is fine for a rigidly structured organisation, but not for dynamic companies."
If and when companies begin applying these disciplines to their networks, they could find themselves in a position to apply extra charges to internal departments which impose a heavy load on the network relative to their peers.
This could be a mixed blessing, according to Michael Winterson, vice-president of European sales and marketing for datacentre hosting company IXEurope. "I hot desk, I travel a lot out of the country and yet people need to reach me all the time," he said.
Winterson called himself a "VIP" within his organisation. But should such VIPs who use higher volumes of bandwidth, or expensive connections such as mobile data links, be charged in the same way as Bob in human resources, who has just downloaded 50 MP3 files from the web and forwarded a four megabyte video file to everyone in the office?
IT managers may need to talk with departments to develop charging policies, creating more work than they anticipated. These policies should be complemented with incentives for operational efficiency, which is something else Rankine recommended.
Chargeback programmes that appear too punitive can affect user behaviour, curtailing valid use of IT resources and creating resentment. Fostering responsibility without preventing users from doing their jobs is a fine line to tread.
The carrot is as important as the stick in such cases, said Winterson. "If you feel as though your employees need to be treated like sheep, you have a bigger, endemic problem. It comes down to something that is more important: creating incentives for employees to do the best for their company."
IXEurope, which spends £100,000 a year on mobile phone bills, will cut 25% from its bill in future thanks to an e-mail asking executives to use landlines wherever they can, Rankine said.
Ultimately, the best solutions will embrace organisational and technical approaches. Certainly, issuing a code of conduct for network usage could help save money without deploying expensive technology, but being able to monitor and charge for network traffic and other resources could boost your efficiency.
The more complex your network, the greater the challenge will be, while the more technologically advanced you become, the more of a cultural shift you will need to make. And as we all know, cultural upgrades are often the hardest of all.
Tools to keep tabs on network costs
- Computer Associates' Unicenter Service Manager includes features for metering usage metrics, and also has a service accounting module so that IT managers can calculate costs based on the usage of IT services.
- CIMS Resource Accounting, Chargeback and Transaction Billing uses data collector modules to gather information from resources including the network, storage, databases and enterprise resource planning systems. They are fed into a central engine linked to a database fronted by Crystal Reports.
- SAS Institute sells a set of IT management products, including SAS IT Charge Management, which lets IT managers allocate resources to different departments and impose different billing schemes based on IT usage. Its Accountview feature handles reporting for IT managers and users.
- Windows System Resource Manager is available in the Enterprise and Datacentre versions of Windows Server 2003, Windows System Resource Manager allocates memory and processor time to different applications. It also includes an accounting feature.
- Allot's Netenforcer appliance includes a Netaccountant add-in module to collect traffic data. It includes traffic reporting, although for financial analysis a flat file produced from MySQL must be manually transferred to an accounting system.
This was first published in September 2004