BMAS Home | About BMAS | Bandwidth Management | Technology Papers and Guidance | Product Evaluations | Calls for Participation | Enquiries
Bandwidth Management, Techniques and Technologies
3.1 What is Bandwidth Management?
Bandwidth management is a generic term that describes the various techniques, technologies, tools and policies employed by an organisation to enable the most efficient use of its bandwidth resources Efficient use means both the minimisation of unnecessary1 bandwidth consumption and the delivery of the best possible levels of service to users. Bandwidth resources refers to the bandwidth of a network, which might be a local campus network, a Regional Area Network such as Net North West, [1] or a national network such as JANET [2].
Generally speaking, comprehensive and effective application of bandwidth management will involve three broad areas of activity:
deployment, operation and maintenance of a range of bandwidth management technologies;
enforcement of use and access policies based upon organisational Information Technology (IT) strategies and good social behaviour, i.e. netiquette;
monitoring network traffic, applications and services.
Bandwidth management may be applied on a campus network, on specific, heavily used segments within a campus network, on critical links to wide area or regional area networks or on national networks such as JANET. In general terms its application:
reduces traffic levels and improves utilisation efficiency by affecting the flow of traffic within networks and on critical links;
ensures selected applications receive the bandwidth they require to operate to agreed standards and levels of service;
In essence therefore, bandwidth management makes the most of existing network capacity in order to improve and guarantee the quality of user services.
3.2 Why is Bandwidth Management Important to UK education?
Within UK education, university, college and research establishment networks are linked together by regional area networks, that are in turn linked to JANET and thereby the global Internet. This comprehensive network infrastructure supports bandwidth speeds ranging from a relatively low 2Mbit/s linking colleges, through 100Mbit/s plus within universities, to the high-speed multi Gbit/s SuperJANET [3] trunk network.
An individual organisations bandwidth requirements will depend on its teaching and research programme, its development strategy and management policy and the size of its user base. Within all organisations use of the web, voice , video and all forms of multimedia, (in particular streaming media and generally unwanted peer-to-peer (P2P) computing applications such as Kazaa), continues to grow relentlessly. Within the university and research sector there is significant and growing activity in heavily bandwidth-consuming research areas such as Access Grids [4] and high-energy particle physics. Within the college sector the emphasis is on the use of the Internet and the World Wide Web to enhance teaching and the learning experience. New, and soon popular, bandwidth-intensive network applications appear on a regular basis, and, set against this background is the drive within UK education to continue to expand its numbers and activities for some years to come.
In this dynamic and rapidly evolving environment organisations must ensure that their bandwidth capacity meets their users demands, or risk eroding their appeal to prospective students and falling behind their competitors. This is especially important for colleges at the lower end of the bandwidth scale that have to monitor and manage usage very carefully to maintain a good user service. However, universities generally ambitious research and development programmes mean that they too, although generally operating in a bandwidth-abundant environment at the present time, are not exempt from bandwidth considerations and management, and need to closely monitor their bandwidth utilisation and apply bandwidth management when necessary.
It may be concluded that most, if not all, organisations within UK education may need to, or are strongly advised to, implement bandwidth management to a greater or lesser degree, if they wish to deliver the required levels of service to their users now and in the future.
3.3 When Does Bandwidth Management Become Necessary?
In principle, bandwidth management becomes necessary when the cost of adding more bandwidth to deliver required levels of user service becomes prohibitively expensive, and implementing bandwidth management is cheaper. This is, of course, a somewhat rigid condition, and in the real world a more pragmatic approach to delivering service levels is generally adopted.
In practice, most organisations already undertake some form of bandwidth management (typical examples being the use of a web proxy cache and the tracking down, and possibly prohibition, of users of excessive bandwidth) whilst upgrading their networks and associated bandwidth capacities over the course of time, i.e. a dual approach. In fact many bandwidth management components are already an integral part of their overall service.
It is highly advisable to proceed in this way and to begin to implement bandwidth management now, rather than wait for service degradation to appear as usage grows and links become congested.
In this Good Practice Guide organisations are encouraged to continue to implement bandwidth management and also to adopt a more systematic approach by formulating an organisational bandwidth management strategy.
3.4 What are the Components of Bandwidth Management?
The components of bandwidth management may conveniently be classified into three broad categories, corresponding to the areas of activity in Section 3.1, namely:
techniques and technologies known as technology-based bandwidth management;
organisational access and management policies known as policy-based bandwidth management;
monitoring.
Generally speaking techniques and technologies affect the flow of traffic down a communications link. They include:
caching, to store frequently used data locally, (including Web acceleration to cache frequently accessed data by remote users and distribution of content, i.e moving content from a single location to multiple locations nearer the end users);
traffic throttling and shaping, bandwidth prioritization and QoS to deliver bandwidth based on application and guarantee an acceptable level of service;
data compression to reduce the amount of data to be transmitted;
Organisational access and management policies affect the level of traffic down a communications link by the application of mechanisms such as access control and by educating users in good network practice. More specifically this category includes:
content filtering techniques;
preventing unauthorised network access and traffic, based on access control, authority-to-use and acceptable use;
education of users to use the network wisely and with consideration for others, in other words, to use netiquette.
Accurate monitoring of network utilisation, usage and performance profiles, to determine when bandwidth management is required and its effectiveness in the delivery of improved application performance and user services, is of course vital. It involves a range of standard tools and dedicated products both open source and commercial.
3.5 What are the Current Trends in Bandwidth Management?
At the time of writing there are several ongoing trends and developments.
Use of firewalls and filtering techniques (as part of organisational access policies and security, but that are, in effect, a form of bandwidth management) continues to expand.
Web-caching continues to be used by many organisations, and many vendors now incorporate caching facilities in more comprehensive, feature-rich products, which are increasingly favourably priced. Caching remains a dynamic area of growth and development, with vendors continuing to improve cache performance.
Traffic management and bandwidth shaping and throttling technologies are becoming increasingly popular.
QoS technologies are being developed by vendors and their potential is under active investigation at various universities.
The development of in-flow compression technologies is proceeding apace. These locate and eliminate repeated data patterns in traffic flowing down a communications link in real time. Various detection and compression algorithms are employed for this purpose. Compression technologies can potentially deliver significant bandwidth savings, but at a financial and administrative cost.
Content Delivery Infrastructures (CDIs) for media streaming are under active development, with many vendors offering a range of products. Content delivery integrates many approaches, including pre-positioning, caching and streaming.
3.6 How is Bandwidth Management Evolving within UK Education?
Most colleges and universities already employ some form of bandwidth management to a greater or lesser degree. Access controls, in particular on gateway routers, are widely implemented. Content filtering is gaining in popularity especially in FE colleges.
Web proxy caching remains very popular. Caches are already widely deployed in universities and now also in colleges, and most recently in schools and other organisations.
Traffic management technologies such as bandwidth shaping and throttling are attracting much interest, in particular for monitoring and controlling high-bandwidth consuming applications such as Peer-to-Peer (P2P) [5] and are beginning to be deployed in universities. The use of these technologies is expected to continue to expand within the university sector and to make inroads into the college sector.
Trials of QoS [6] between several universities are underway under the auspices of the United Kingdom Education and Research Association (JANET(UK)). There are various complications associated with QoS, in particular in the establishment of a guaranteed end-to-end level of performance across a multi-node network. As a consequence uptake of QoS has been somewhat slow. Nevertheless, it is expected that there may be a gradual implementation over the next few years as these problems are resolved.
A JANET(UK) led CDI initiative between several universities is currently underway [7]. In the immediate future CDIs may be expected to grow as streaming and video become increasingly incorporated in the teaching and learning curriculum.
3.7 An Organisational Bandwidth Management Strategy?
As the popularity and usage of heavily bandwidth consuming applications grows and the number of network users multiplies over the coming years, the need for a concerted and co-ordinated effort to monitor bandwidth utilisation and implement bandwidth management will become increasingly important to ensure excellent service provision. As already stated, it is believed that this effort will be greatly facilitated by the formulation and implementation of an organisational bandwidth management strategy. In Sections 8, 9 and 10 of this guide bandwidth management strategy and its implementation will be looked at in more detail.
To set the scene, a broad review of technology-based and policy-based bandwidth management will be undertaken, followed by an overview of network monitoring. Part 1 concludes with a list of links to more detailed sources of technical information.
4. Technology-Based Bandwidth Management
In Section 3.4 it was stated that bandwidth management components fall into three broad categories: technology-based, policy-based and monitoring. Within the technology-based category there are three primary interdependent components, namely:
caching;
traffic management and QoS;
compression.
A general description of each of these techniques now follows (technical details are provided from the references). The situations in which the use of a technique is likely to result in significant benefit is discussed, as well as those in which this is unlikely. The dependencies among the techniques are also highlighted.
Proxy caching accelerates the delivery of content, thereby reducing retrieval times and increasing the efficiency of bandwidth utilisation in a communications link. Within a typical organisation, this forward caching functionality will be present in web browsers serving individual users and in the form of a local proxy cache servicing requests from many users.
Forward caching operates as follows. A user requests a web page via a proxy cache. If the cache can fulfill the request it sends the page to the user. If it cannot it will send the request to the original site. The original website delivers the page to the cache, which stores it for future requests and delivers it to the user.
In principle a proxy cache will service most of the web protocols (e.g. Hyper Text Transfer Protocol (HTTP), File Transfer Protocol (FTP), Network News Transport Protocol (NNTP) and some streaming protocols). However in practice most caches service only HTTP. This can severely impair caching effectiveness when applied to multi-protocol communications links as we will illustrate below.
Reverse proxy caches intercept and service an incoming request before it reaches a given web server within an organisation. They reduce the loading on the server by storing the most popular content and relieving the server from processing requests. Reverse proxy caches can reduce bandwidth utilisation within a network, but generally caching as a form of bandwidth management refers to proxy caching.
The most meaningful measure of the effectiveness of proxy caching is the hit-rate. This gives an indication of how many user requests were serviced directly by the cache and how many requests had to be forwarded to the original site. The higher the hit-rate the greater the bandwidth saving and the greater the bandwidth utilisation efficiency.
Caching can be made much more effective by the pre-fetching and refreshing of objects during periods when bandwidth is plentiful, for example overnight when network usage is minimal. Materials for the following days lectures may be pre-fetched the night before.
Caching can be highly advantageous in a situation where many users are accessing the same content via a relatively low bandwidth, heavily used link. In these circumstances hit-rates can be as high as 40% or 50% and the bandwidth saving can be quite significant, resulting in tangibly better service quality.
However this happy situation turns out to be highly artificial. It assumes that a very high proportion of network traffic is both web based (e.g. HTTP) and that the content is static as opposed to dynamic [8]. Clearly if one or both of these assumptions is false then caching will not significantly improve bandwidth utilisation efficiency.
A typical communications link supports many protocols and applications. These include video and audio, real-time interactive and delay sensitive streaming applications, and bulk file transfer applications. Moreover an increasing amount of web content is dynamic in nature, such as database transactions. In this environment the added efficiency caching brings is relatively insignificant.
4.1.4 Caching, Traffic Management and Quality of Service
In a multi-protocol environment in which real-time, streaming and interactive applications are significant consumers of bandwidth it is advisable to employ traffic management and QoS techniques in conjunction with caching. These techniques are now described.
4.2 Traffic Management and Quality of Service
4.2.1 Difference between Traffic Management and Quality of Service
A distinction between traffic management and QoS needs to be drawn at this juncture. In general, traffic management refers to the management of bandwidth on an organisational network, (for instance allocating a given bandwidth to an application on a particular communications link within a campus network). In general QoS refers to the allocation of a given or guaranteed amount of bandwidth to an application across multiple sub-networks comprising many nodes and network devices. As might be expected the former is easier to implement and manage. However, development of the latter is proceeding apace and significant progress is expected in the next few years.
4.2.2 Traffic Management and Quality of Service Explained
Traffic management refers to the ability of the network to provide preferential treatment to certain classes of traffic. It is best deployed on heavily loaded communications links that carry significant amounts of time-sensitive, interactive traffic (e.g. video, audio and streaming data). In practice this means limiting the bandwidth that certain applications (e.g. e-mail) receive whilst ensuring that other time sensitive, and possibly business critical applications, such as streaming media are guaranteed the bandwidth they require.
In order that applications are allocated adequate bandwidth for satisfactory operation (20kbit/s, for example) the traffic management technology must be aware of the applications using the network. This often means detailed knowledge of the applications, because many share the same port or hop between ports, and a solution that cannot distinguish between them cannot effectively manage bandwidth. (A common example of a popular port is port 80 used in browsing, streaming media and other applications. Many P2P applications such as Gnutella constantly hop between ports rendering port-based bandwidth management ineffective.)
Another important factor in traffic management is the ability to control both incoming and outgoing traffic. Queuing mechanisms control traffic leaving the network but do not control traffic coming into the network that is often the cause of bottlenecks and congestion. Various techniques under development are available for controlling incoming network traffic. One popular technique is Transmission Control Protocol (TCP) Rate Control [9]. This TCP protocol feature tells remote servers how fast they can send content to the receiving network.
4.2.3 Implementing Traffic Management and Quality of Service
To implement these technologies the following issues need to be addressed:
network traffic should be continually monitored and a network usage profile should be established and maintained;
network traffic should be analysed to determine the traffic patterns of all applications, but especially key and mission critical ones;
bandwidth allocation policies should be formulated based upon usage profile and applications traffic and incorporated into the traffic management system;
services should be monitored for quality and allocation policies amended to take into account shifting traffic patterns;
4.2.4 Traffic Management and Quality of Service Effectiveness
Generally speaking these technologies are state-of-the-art and are only just being deployed in organisations. Significant development and research are being undertaken within both academia and the commercial world, in particular in QoS [10] and their full potential has yet to be realised.
Although caching and compression do provide additional efficiencies and increase available bandwidth, aggressive bandwidth applications will soon consume this resulting in other, less aggressive applications not receiving adequate bandwidth. Active traffic management alleviates this problem by allocating bandwidth to applications according to priority and pre-determined policy.
In this sense traffic management is one of the most critical components of bandwidth management. In a multi-protocol, multi-service environment it is likely to have a significant impact on improving efficient bandwidth utilisation. Very effective bandwidth management may be achieved by combining traffic management with caching and compression.
Compression can be dual or single-sided. The former technique involves the compression of data entering a communications link and its subsequent decompression at the other end. The latter applies specifically to the optimisation of web pages for transmission over a network. Dual-sided compression is of course more generally applicable.
The basic function of compression is to reduce the amount of data to be transmitted over a communications link. Viewed this way compression is very similar to caching: it causes the network to perform as if it had received a one-time bandwidth upgrade. However, as with caching, compression techniques do not solve the problem of poor application performance. As stated previously, the solution to this problem is to implement compression in conjunction with caching and traffic management techniques.
There are several factors to be aware of when considering deploying compression. For our purposes the two most pertinent ones are:
the compression ratio;
increase in latency as a by-product of compression;
Compression ratio, which is a measure of the benefit that compression offers, is a key concept. It is expressed as: x:1, where x is the number of input bytes divided by the number of output bytes. Clearly the larger the compression ratio, the more benefit there is to deploying compression. In tests performed over a variety of file types compression ratios in the range of 1.7:1 to 2:1 were found to add significant benefit.
Compression does improve bandwidth utilisation efficiency. However there are many situations in which compression is either problematic or provides no advantage whatsoever. Little advantage is gained with certain classes of traffic such as Voice over IP (VoIP). This is because VoIP is already inherently compressed and cannot be compressed further, with the exception of the header information. In general, compression offers no advantage with encrypted traffic, since most encryption algorithms produce few repeated sequences and hence cannot be compressed by standard compression algorithms.
An effective bandwidth management solution must address both throughput and latency issues. Compression only applies to throughput and can, under certain circumstances, exacerbate latency. Some traffic management solutions provide latency management and should be combined with compression in a comprehensive bandwidth management strategy.
The very latest generation of compression products have jitter buffers at each end to alleviate the induced jitter caused by variable compressibility of data inherent in a multi-protocol, multi-service environment. Even though these buffers do a fairly good job of eliminating aggregate traffic jitter, they cannot effectively de-jitter any one traffic flow. Thus the overall traffic may flow smoothly but individual flows of traffic may incur more jitter as a result of compression. Traffic management techniques, especially when using TCP Rate Control to manage TCP sessions, mitigate this problem by rate-shaping each traffic flow so that bursty traffic cannot induce jitter into latency-sensitive traffic.
4.3.2 Single-Sided Compression
An alternative approach to dual-sided compression is to deploy a new category of one-sided web acceleration products. These products use a variety of technologies to optimise a web page for delivery, including standards based compression, image conversion and optimisation techniques and caching. The end result is compression ratios that range from 2:1 to 8:1 depending on the product and specific content.
These solutions have several advantages over dual-sided compression solutions and proxy caching solutions. The most obvious is that they are considerably less expensive to deploy and manage. The second benefit is that this technology can be used to optimise bandwidth to dial-in users as well as users at the other end of communications links. Some of these solutions are able to recognize the connection speed and browser type that each user is coming in on and optimise content specifically for that user. A final advantage over proxy caching is that they optimize not only static content, but also dynamic content.
The primary drawback to this technology is that it is only used to accelerate web applications that an organisation controls. Unlike dual-sided compression, which compresses nearly all traffic, and proxy caches, which cache all static web content, single-sided solutions only optimise specified websites.
As web-based traffic continues to grow it may well dominate network traffic within the next few years. In addition to the fact that web-enabled applications are growing as a percentage of the overall number of applications, the bandwidth consumed by web-enabled applications tends to be considerably larger than their client-server predecessors (for example, some Graphical User Interfaces (GUI) can consume over 30 times the traffic of the client-server version).
These trends will make HTTP optimisation (proxy-caching, dual sided or single sided compression) along with traffic management an important component of an organisational bandwidth management strategy in the near future.
There are three primary interdependent components of technology-based bandwidth management, namely: caching, traffic management/QoS, and compression. The relative advantages and disadvantages of these components are summarised in Table 1.
| Traffic Management and QoS | Caching | Compression (dual-sided) | Compression (single-sided) | |
|---|---|---|---|---|
| Protect mission-critical applications | Yes | No | No | Only web applications |
| Reduce Bandwidth | No | Yes | Yes | Limited |
| Optimise Bandwidth | Yes | No | No | No |
| Network Core/Edge Deployment | Core or Edge | Edge | Both Core and Edge require | Core |
| Deployment Cost | Variable | High | High | Low |
Table 1 Comparing Bandwidth Management Techniques
When referring to Table 1, it is helpful to place bandwidth management techniques into two distinct but related classes. One, such as caching and compression, actually reduces the level of traffic. These techniques may be regarded as delivering a one-time increase in bandwidth capacity.
The other class includes traffic management and QoS. These are very important components of an effective bandwidth management strategy. Without them compression and caching will reduce the aggregate amount of traffic but will not protect against aggressive applications, such as music downloads, which may result in mission-critical applications not receiving the bandwidth they require [11].
5. Policy-based Bandwidth Management
The previous chapter reviewed technology-based bandwidth management. This form of bandwidth management may be viewed as affecting the flow of network traffic in one of several ways, to both improve bandwidth utilisation efficiency and deliver adequate bandwidth to services. Put another way, it can be said that traffic already in transit is reduced and re-directed by the bandwidth management technology.
Policy-based bandwidth management involves the allocation of network-based resources and services in accordance with the mission and management policies of an organisation. Implementation of this form of bandwidth management affects the level of network traffic by preventing access by user groups and from given locations, or by preventing access to certain sites or services.
5.1 Defining Policy-based Bandwidth Management
Policy-based management allows an organisation to control which network services and resources are available to its users in accordance with its IT strategy. The following issues will influence the definition of an organisational policy [12]:
should users be free to access any website on the Internet;
which users can access which resources on the network;
should user Internet browsing be closely monitored;
should the use of certain bandwidth-hungry (bandwidth-aggressive) applications be restricted, banned or closely monitored;
should access to certain websites be restricted, if so should this be on an individual user basis, by sub-network or by department;
in a multi-protocol, multi-service environment should traffic be prioritised according to importance, e.g. should mission critical traffic be given top priority, real-time interactive traffic be given a lower priority and screen mode terminal access the lowest priority;
what traffic must have guaranteed delivery;
what traffic is eligible for discard when the network becomes busy and congested.
A policy-based system allows the formulation of rules based on these types of questions and in accordance with the organisational IT strategy and any Service Level Agreements (SLAs) that happen to be in force. These policy rules are then configured into the various network resources, (i.e. actual devices) that then implement and enforce them.
5.2 Implementing Bandwidth Management Policy
Most network devices used to implement technology-based bandwidth management also implement policy-based bandwidth management. (The reverse is not necessarily true. For example a gateway router can implement access restrictions but does not usually cache content or shape traffic.) Devices able to implement policy include caches, traffic managers, QoS resources, gateways, filters and firewalls. Many modern devices combine several of these functions. In recent years network elements such as switches have gained the intelligence to provide quality of service and policy-based decision making.
User access restrictions can be imposed on gateway routers, departmental routers, intelligent switches and within caches. Content filtering can be applied on gateway routers and firewalls, or by channeling traffic through a stand-alone system. Traffic can be prioritised and bandwidth levels guaranteed on most modern routers or by standalone dedicated traffic managers.
5.3 Further Information and Developments
The Internet Engineering Task Force (IETF) policy Framework Working Group (POLICY) [13] has developed a policy management architecture that is considered the best approach for policy management on the Internet. This is a significant area of ongoing development and is beyond the scope of this guide. Further information may be found on the Internet Engineering Task Force (IETF) website:
Network monitoring is of course an integral part of an organisational IT strategy; there are numerous tools available in both the open source and commercial worlds and an ongoing worldwide development programme. It is also a vital component of an effective bandwidth management strategy. Network managers need to be aware of their network traffic daily usage profile and of the teaching and research requirements of their organisations to enable them to configure their bandwidth management resources appropriately. In particular they need to know the following:
sites accessed as part of the organisations teaching and research activities;
applications required for the organisations teaching and research activities;
sites accessed by users browsing the Internet;
protocols used;
incoming and outgoing traffic levels, by protocol type, 24 hours per day;
total incoming and outgoing traffic levels, 24 hours per day;
With this knowledge network resources can be configured to:
permit or deny access to given sites;
allocate and guarantee the necessary bandwidth to mission-critical, streaming and real-time interactive applications;
monitor individual user activity, in particular sites visited, and deny access if necessary;
detect use of banned or security risk protocols;
detect any unusual levels of incoming and outgoing traffic;
Until recently network monitoring tools fell into three categories:
media-focused ;
device-focused;
packet-focussed;
Recently application-focussed tools have been gaining prominence and popularity. In the following a brief description of each category is given, further details may be found in the references.
These monitor the physical structure of a network, highlighting problems such as poor signal quality and breaks in the wiring or cabling. Whilst invaluable for troubleshooting the network, they do not address the issue of application performance over the network.
These are used in the day-to-day operation of a network infrastructure. They are usually specially developed software products incorporating Simple Network Management Protocol (SNMP) [14] designed to monitor the operational status of network components such as switches, routers and servers. Problems, real or potential, are displayed centrally as alerts or warnings.
Device-focused tools facilitate a rapid resolution (or alleviation) of problems with network components and assist in maintaining the continuity of services. In general, however, they do not monitor the operational status of applications and are therefore of limited use for the task of bandwidth management.
These capture data packets as they flow across a link. Commonly called packet analysers or protocol analysers, they are in widespread use. They are indispensable in locating and resolving protocol-related problems, and monitoring traffic levels. These tools do not monitor the operational status of applications.
An application-focused tool differs from traditional network management tools. Instead of providing an analysis of network component behaviour or of the protocols in use, application-focused tools are designed to analyse the flow of applications over a network.
A typical application-focused tool gathers historical data from the network and analyses it to determine which applications on the network were used, by whom and when . In so doing it helps to explain the level of bandwidth utilisation at any time of day throughout the year This type of analysis enables the network management to ascertain which links are reaching capacity and why, and if necessary take remedial action.
7. Bandwidth Management Technologies
It is useful to be aware of the specific technologies currently available. To facilitate this we have included this information in the BMAS bandwidth management review document, that will be available shortly.
1; It is worth mentioning that what counts as unnecessary bandwidth consumption may depend on the observer. For instance the network managers view that certain network traffic consumes unnecessary bandwidth may not coincide with the view of the senior organisational management. Hence it is highly desirable for senior management to agree an organisational bandwidth management strategy, of which more later.
| [Introduction] | [Part 2] |
