Online data usage in Victorian public libraries: an empirical analysis.
The internet is playing an increasingly important role in the provision of library services. More information has been digitised and more databases are becoming available online. Users are also demanding more immediate access to information, by remote access, or access within a library branch. Victorian research suggests there is 'no evidence to support the proposition that the demand for public access [in public libraries] is plateauing or declining'. (1)
Providing internet access points, and delivering ever greater amounts of data to support this demand, is having an impact on library resources and budgets. It is therefore important for libraries to understand what types of usage drives internet data overheads. It is also important to then determine what funds, as well as policies, are required to support the ongoing increase in online services.
The internet and Victorian public libraries
Since June 2002 a growing number of Victorian public libraries has migrated to one or more of the Vicnet DSL, fibre, or wireless services that constitute Vicnet's eCommunity network. With an increase in bandwidth these libraries have found their use of the internet has risen, in some cases alarmingly so. This not only places a considerable strain on their budgets, it also leads to congested connections. The congestion can occur despite the libraries significantly upgrading bandwidth. Anecdotal information from other Australian states and territories suggests Victorian public libraries are using considerably more data than those interstate. This could be because the Victorian public librarians have migrated their branches to broadband connections earlier than their interstate colleagues. If this is the case, libraries in other parts of Australia may be faced with a similar situation. For this reason it is important to understand what has happened in Victoria. Lessons learnt from the Victorian experience can be applied to the wider library community.
Some Victorian libraries also claim that as the bandwidth increased, usage changed. There is the suggestion that this change in user behaviour increased data usage, and as a result the costs to libraries. Many of the eCommunity sites in this study migrated from the Vicnet shared one way satellite service (initially funded by the federal government's Networking the Nation program and Multimedia Victoria) with ISDN back channels. When connected to this satellite service these libraries 'experienced reasonably stable, or a slight growth, in bandwidth requirements'. (2) However the satellite connections only connected the main library branch. The other branches still used 128K, or more frequently 64K, ISDN connections. There were even some sites that used 64K permanent dialup connections. The issue of whether increases in bandwidth out to branches changes usage, and therefore bandwidth overheads, is intriguing. Understanding these changes should enable libraries to be better informed when making decisions on communication upgrades.
Bill Gates predicted that 'in just the next five years the communications bandwidth available in urban business areas will grow by a factor of 100 as network providers compete to connect concentrations of high user customers'. (3) Although Gates' prediction has not yet occurred for Australian public libraries, the increases have been significant. They have also had an impact on costs. If, as Taylor predicts 'far from building information motorways, we are just about to begin building some A roads and a few B roads ... the computer industry is at about the same stage as the Model T Ford' (4) the imperative for libraries to plan and budget for even greater data increases is of utmost importance. With the introduction of ever more online library services, incorporating data rich multimedia, video on demand, and voice over IP, the library of the future could well look back to today's interact infrastructure in the same way we view vintage cars. What is at stake is public libraries continued participation in the building of the information economy's infrastructure.
If, as Hall notes, (5) this increase in IT infrastructure is part of Schumpeter's process of creative destruction which also necessarily involves bursts of infrastructure, libraries must continue to reposition themselves and the services they offer, to keep up with the current 20 or so year economic and technological cycle. Central to this process of strategic positioning is Porter's view (6) that organisations must ensure their online strategies are an integral, rather than a separate, part of their overall budget and strategic goal setting. Some libraries have already adopted this approach. 'At least one public library in Victoria (Yarra Plenty Regional Library Service) now treats its interact services as another branch and as such internet services are staffed
and budgeted accordingly.' (7)
The e-rate proposal
The Australian Senate's Libraries in the online Environment (8) report of 2003 and the 2000 Report on a policy for public access to the internet for Victoria (9) by Trinitas for Multimedia Victoria, investigate the equity and sustainability of public interact access. The seventh recommendation of the Senate report is that
(a) the Australian government negotiate with telecommunications carriers to establish an e-rate or discount for broadband access to public libraries and that, if negotiations are not successful, consider imposing a requirement on carriers under the Universal Service Obligation; and
(b) that further funds be allocated under an expanded National Broadband Strategy for expanding broadband access in libraries
The e-rate proposal is based on the US precedent. However the Preliminary analysis of public library e-rate data: 1999-2002 shows that public libraries in the US have only ever received between 3-4 per cent of the e-rate subsidy. Oder also reports
because there is a federal cap to e-rate funds of USD$2.25 billion, not all requests are fully awarded. In 2002, for example, 89 per cent of library applicants received e-rate funds, but only 52 per cent of the discount dollars requested were granted. Well over half the e-rate funds goes to telecomm services, while the remainder goes to internet access and internal connections. (10)
Similar funding caps and restrictions applied to an Australian e-rate scheme, would impose a considerable burden on public libraries. This would be particularly true if the Victorian experience of data increases is repeated nationally. During the 12 month study period the Victorian public libraries experienced increases of between 14% and 237% in total data usage, with the average increase being 110%. It is difficult to see Australian governments being willing to subsidise data rates if the data overheads required to support library services double every 12 months. Government may be better served by subsidising circuit costs, as these tend to be more stable.
Whilst not discounting an e-rate program for Australia, libraries should also consider strategic peering options, where the data can be transmitted free of charge. Libraries then need only pay for the circuit costs and the data sourced from outside the peering network. Vicnet has already started this process with its eCommunity network. This network may prove a useful model for libraries other states and territories.
The benefit of peering for governments is that as bandwidth use increases, there is no additional burden to increase subsidies when the data is being drawn from a peering partner. The Western Australian Interact Association and Victorian Interact Exchange or Vix (of which Vicnet is a member) have been proactive in establishing peering relationships in their states.
In Victoria, peering may also play a role in the Victorian government's Telecommunications Purchasing and Management Strategy (Tpams) project. Nationally, the National Office for the Information Economy (Noie) was considering a demand aggregation broker program. Given the increased reliance on internet data, libraries should consider what role these peering bodies or projects will play in the future. Libraries also need to consider how they will align themselves with these emerging projects.
It is also important for libraries to better understand the service elements of various broadband delivery methods and the associated costs. The Dandolopartners report by Adams and Meagher for the Australian Communication Authority identified customer unawareness and lack of standardised information from vendors as one of the main reasons there is so much confusion and disputes relating to broadband service deliverables and costs. It noted 'some consumers, attracted by the promise of speed, seek the best price, without fully understanding the restrictions on speed and downloading'. (11)
As libraries increasingly depend on their internet connections for mission critical applications such as catalogue traffic and staff email, understanding and dealing with the cost of bandwidth capacity will only become more important.
This paper focuses on the delivery and use of interact data by a sample of the connected public libraries in metropolitan Melbourne and rural Victoria. The study looks at what is driving their internet data usage. It also looks at what network and policy changes libraries have adopted in order to take greater control over their data usage. It is hoped that by making this information available to the wider public library community, all library services will be better informed, and better placed, to protect themselves from data and budget blowouts.
Libraries may also be placed to protect themselves against inappropriate, or illegal, use of their internet connections. Furthermore, it is hoped that the libraries will be better placed to implement policies and procedures that provide a high level of online services to users, whilst still keeping control of costs.
For reporting and billing purposes Vicnet uses NetFlow to measure eCommunity DSL, fibre and wireless customers' data usage. Consequently, this study uses NetFlow as the primary method of collecting raw data. It is therefore important to have some understanding of NetFlow, as well as an understanding of the Vicnet eCommunity network.
the measurement base for Cisco's internet and Enterprise Quality of Service (QoS) initiatives ... A network flow is defined as a unidirectional sequence of packets between given source and destination endpoints. Network flows are highly granular; flow endpoints are identified both by IP address as well as by transport layer application port numbers. NetFlow also utilises the IP Protocol type, Type of Service (ToS) and the input interface identifier to uniquely identify flows.
More information on NetFlow is at http://www. cisco.com/warp/public/cc/pd/iosw/ioft/neflct/te ch/napps_wp.htm.
Vicnet uses it to measure daily uploads and downloads per site. In addition, daily and monthly totals are available in real time to all eCommunity broadband customers via Vicnet's Network Operations Centre (Noc) site. The Noc site is located at http://noc.vicnet.net.au.
With NetFlow, Vicnet measures
* free downloads and uploads' this is information sourced from any other eCommunity site irrespective of whether that site is another branch of the library service, another connected public library or a connected government department, NGO or community group
* downloads and uploads' this is information that has been sourced from outside the eCommnnity network. As with the other eCommunity sites, the libraries in this study are charged for downloads outside of the eCommunity network. Vicnet, however, does not charge for uploads
Internet connectivity is supplied to the eCommunity network and the State Library of Victoria via two 100Mbps Ethernet connections to Vicnet's point of aggregation in the third party peering room at the Melbourne stock exchange. Vicnet has also recently added another point of aggregation in the Sydney. For redundancy, two upstream data wholesalers are used.
If a council or associated site eg an aged care site was linked to the library subscription, this site was included in this study. However, none of the other Vicnet customers have been included in this study. Some of the libraries in this study have their own line of sight network, so they only connect to Vicnet via their main site. This study recognises this aggregation when considering the number of terminals per site.
Vicnet's eCommunity network
As all the libraries in this study are linked to the eCommunity network. At its simplest the network uses a number of Vicnet's circuit wholesalers to provide DSL, fibre, or wireless connections to customers. This allows Vicnet to aggregate all the customers' bandwidth. These wholesalers only charge Vicnet for the circuits.
Consequently it does not incur data costs, as long as the data moves between one Vicnet connected customer and another. As a not for profit ISP, Vicnet is then able to waive the charges for data between connected sites, irrespective of whether the sites belong to the same organisation. For example, a connected library can download free data from a connected government agency. If, however, the data comes from outside the Vicnet eCommunity network, the customer is charged for the downloads. As a rule Vicnet does not charge for uploads. The eCommunity network is self funded through the paid DSL, fibre, and wireless subscriptions. As a result it is independent of the uncertainties of government funding processes. One of its strengths is that it is not restricted to public libraries. The eCommunity network connects public libraries to community groups and schools. It also links libraries to state government agencies and departments.
Overview of data usage
The study is in two parts. The first presents the data usage of all 60 sites. These sites support a resident population of almost 2 million people, of which 1.568 million reside in Melbourne or its urban fringe, and 426,545 in regional Victoria. However some of this rural population could be considered as part of Melbourne's urban fringe. This section of the study is interested in macro trends, such as monthly data per site and the overall changes, per library system, of data usage throughout the study period. The second part of the study looks at four of the sites in more detail. These case studies look at two single municipality metropolitan library services, and a metropolitan regional library service supported by two councils. In addition, the case study looks at a rural based regional library service supported by six councils. In the rural library service most branches support their entire town, as well as the community in the surrounding area. This said, three branches of the rural regional library service support one large regional centre.
Using NetFlow, Vicnet commenced monitoring the sample library and council sites in October 2002. For the following 12 months monthly NetFlow upload and download data was gathered from the sites, collard, and then graphed. This part of the study looks at the total monthly uploads and downloads of each library network, as well as the average downloads and total monthly downloads per branch. While upload data was also investigated, Vicnet does not charge for uploads so the main focus of this paper is on download traffic usage. However upload traffic can have an impact, especially on library system traffic between the main library and its branches. This will be explained in more detail.
Some of the sites use the Vicnet proxies, others do not. For consistency this study has focused on the total data usage per site, as well as the total interact data used ie all data coming from outside the eCommunity network. To be included in the study the library site must have had a Vicnet DSL, fibre, or wireless connection for at least six months. To protect the privacy of the participating libraries, each library service was given a letter code (A to N) for identification. Libraries with ISDN and or satellite connections, as well as libraries that migrated to DSL, fibre and wireless in the second half of 2003, have not been included. The exception to the selection criteria is library N. This rural library service was included in the first part of the study because there was less rural library data available.
Growth in data usage
Figure 3 clearly shows that the overall bandwidth required for these sites trebled in the 12 months from October 2002. Whereas in October 2002 Vicnet could support this sample group with just under 200Gb of data downloads a month, by the beginning of October 2003, close to 540Gb of data was required. What this chart does not show is that most of these sites migrated from 128 and 64K ISDN to DSL connections in July and August 2002. As a result, between August and October 2002, the cumulative data usage for these connected sites had already grown dramatically. As mentioned in the introduction, prior to migrating from a shared satellite connection to individual broadband connections, the libraries had experienced relatively stable data usage. As a number of sites use their connections to refresh the proxy servers at each branch, figure 3 also shows the interact only data usage.
The overall increase in data usage is also apparent when considering the findings outlined in table 1. This shows the change in average downloads from the start of the study in October 2002, as well as the end of the study period in October 2003. In October 2002, the vast majority of library branches connected (77%) did less than 5Mb of total data downloads a month. By October 2003 only 44% of connected branches did less than 5Mb of total data downloads a month. By October 2003, not only did the majority of branches (56%) do more than 5Mb of total data downloads a month; 32% did more than 10Mb a month, and 11% did more than 20Mb a month.
Furthermore, in October 2002 there was no noticeable difference between the total data downloads, and downloads from outside of the Vicnet eCommunity network. By October 2003 an increasing amount of data was coming from the wider eCommunity network. A large part of this eCommunity data appears to be coming from within each individual library system. This is especially true where the libraries use their Vicnet connections to refresh the individual branch proxies each night. As the eCommunity network grows, and more sites are connected, it appears that the use of data from other eCommunity sites is also increasing.
Seasonal patterns of data usage
Although the overall trend is for increased data use, it is clear that there were months when use declined, in particular April, June and (even though it falls outside of the study period) November 2003. All of the libraries in the sample held the view that recreational internet use by school students was one of the main drivers of use in public libraries. Most of this student use took place in the afternoons and evening, as well as during school holidays. In 2003 the Victorian school holidays were from 11 April to 28 April, from 27 June to 14 July and from 19 September to 6 October. First term started on 28 January, and the fourth term, which included the end of year exam period, finished on 19 December. If internet use during school holidays increases data use, this may explain the increase in January but it does not explain the decline in April. Library B, however, noted that recreational internet use by students tends to decline near the end of each school term. The view was that students are more focused on completing assignments during this end of term period. This could then explain the declines in June and November, and possibly April. Recreational internet use such as gaming, as well as data streaming and file sharing, appear to have a more profound impact on library bandwidth than more academically focused internet usage. However more research is required to better understand the seasonal nature of user behaviour and internet data use.
Diversity of data usage
The results also show that the amount of data used by each library service varied considerably. For example, library B (2 branches supporting 128,201 residents with 28 user only connected terminals) and library J (8 branches supporting 300,543 residents with 119 user and 106 staff connected terminals) have consistently used considerably more data than the other library services in this sample. Interestingly, the data usage of library B and J is similar. This is despite considerable differences in the number of branches, connected terminals, and the size of their resident populations. Yet when these two very large data consumers are removed from the sample, along with libraries C, D, and N (these other libraries came online after October 2002), the overall trend is still very much upwards. The remaining libraries more than doubled their data downloads in the 12 month period, going from 123Gb in October 2002 to 259Gb in Oct 2003.
A closer investigation of the data highlights other differences. Figure 5 presents the library's download usage but does not cumulate the data. This figure shows that some libraries, especially metropolitan libraries J and B, had also experienced very erratic data usage patterns. These libraries have struggled to contain the blowout in data usage, and over the 12 months had implemented a number of changes to their networks and usage policies. Though less erratic, library H had also implemented a number of changes to the network and policies, as it tried to take control of escalating data charges.
[FIGURE 5 OMITTED]
The case study in part 2 looks at the experience of library B and H in greater detail, in an attempt to shed light on this fluctuating demand.
The overall diversity of data usage is even more remarkable when one considers the findings with additional background information. For example, as mentioned previously, library B is a two branch single municipality metropolitan library with 26 user terminals (staff pcs do not use this DSL connection), yet during the study period it used considerably more data than library H. Library H is a metropolitan regional library with 7 branches with 69 user and 105 staff pcs (not counting opacs that also use the DSL connections). Library H also has a separate aged care site. The second library B branch connects to the main library via the council's line of sight network and both these libraries share the one 2048 / 384 kbps ADSL connection. Library H, on the other hand, uses a 2048 SHDSL connection at the main branch, plus two branches connected to a line of sight network.
Two other large branches are supported with their own 2048/384 kbps ADSL connections and each of the remaining branches uses a 512/128 kbps ADSL connection. Despite having considerably more bandwidth overall, since July 2003 library H has used less than 50% of the data used by library B. From this data it would appear that factors other than the number of branches and internet terminals are the main drivers of data usage.
Irrespective of the size of the site, the connection capacity, or hours of opening, sites that include staff and user terminals have a lower average hourly data download rate per terminal than do sites where the Vicnet connection only supports the user terminals. Therefore, when forecasting future data requirements it is important for libraries to differentiate between the number of user and staff terminals per site.
To further investigate the diversity of experiences between sites, Vicnet needed to develop a way of benchmarking data use that took into account the difference in opening hours, as well as the number of terminals per site. By dividing the monthly downloads by the number of staff and user terminals connected to each site, and then dividing this output by the hours the site is open per month, Vicnet was able to provide data which can be used to benchmark library usage. The hours per month were calculated on the hours open each week multiplied by 52 divided by 12. Over the last 12 months, the libraries changed the number of terminals in their branches, so this section of the study only looked at average data usage back to June 2003.
The Netflow data, the seasonal nature of the data usage, and feedback from the study sites, all seem to support the notion that the difference between user and staff data usage is driven by the amount of user recreational internet usage. Although Vicnet was unable to obtain a set of figures for staff only terminals, it would appear that user terminals have about five times the amount of data downloads as staff terminals.
The second part of the study looks at four of the sites in more detail. The following libraries were chosen because of the diversity of their internet use, as well as diversity of their user profiles. Unlike some of the other libraries in the first part of the study, these libraries all used their connections to support both internet and library system traffic.
Each of the library services was presented with its data usage from October 2002 to September 2003. The libraries were also given the data usage of all 60 sites in the study. The identity of the libraries was masked in order to protect privacy. They were also given a set of questions. The results of the questionnaire were then matched up against the NetFlow data, as well as the number of staff and user terminals using Vicnet's eCommunity circuits. Finally, the results were also matched against the hours of opening per week, as well as the resident population. Population figures were taken from the June 2002 figures used by the Department of Victorian Communities: Local Government Division (DVC).
Library B single municipality 2 branch metropolitan library service with both branches sharing a single 2048 / 384 kbps ADSL link. This library supports a resident population of 128,201. This library also supports a generally lower socioeconomic outer urban population.
During the study period, library B experienced an overall increase in data downloads of 176%. This increase is well above the average increase of 110% experienced by the overall study sites. As with the other case study libraries, library B also experienced periods where the data usage dropped, especially from March through to June 2003. Interestingly, the decline in usage for this service was longer and more pronounced than other case study libraries, especially library H and K. Understandably, library B has been very concerned about the amount of data its users have downloaded. In May and June the library trialed blocking user file sharing with Norton's Internet Security 2000. However, by using the uninstall program freely available from the Norton's Symantec website, a user found a way around this block. The library has since resolved this issue by blocking peer to peer files sharing at the port level, and the data downloads figure has reduced again since October 2003. As a result of the two libraries sharing the one internet feed, Vicnet was unable to separate the traffic of site 1 and 2. Consequently it has not been possible to look closely at the difference between branches belonging to the same library service. However site 1 has more pcs than site 2, and the library service confirmed that site 1 consistently generates 60% of the total bandwidth. The ratio of bandwidth usage therefore appears to correlate with the ratio of user terminals per branch. The survey questions for library, B elicited the following response.
As with the other case study libraries, library B users expect faster internet connections. Library B significantly upgraded the capacity of its link when it migrated from shared satellite with ISDN back channel to DSL. This reduced the level of user complaints about internet speed, but there is still a user expectation for even faster connections.
The library also noted that there has been a change in the user profile with more teenage boys using the library and the internet terminals since the DSL connections were put in place. The library also mentioned that the data usage consumed by filesharing is driven by a small number (less than 20) of users. If this is true, it has important ramifications for other library services. It may only take a few users, over a short period of time, to have a dramatic impact on a library's data usage and costs. Finally, library B noted that it had to devote more staff and resources to monitoring internet usage, implementing policies and procedures, and then police these procedures. These are all issues that concern the other libraries in the study. Library B has been forced to be more proactive as it has faced the financial ramifications of a sudden increase in bandwidth usage. This raises the issue of the skills required by staff to manage networks and support online services. Furthermore, Vicnet and library B have spent considerable time endeavouring to take control and manage the library's data usage.
Closer collaboration with vendors, and sharing lessons learned with other libraries, can only help reduce the resources required to manage internet usage.
Library E is a single municipality 4 branch metropolitan library service that includes 2 aged care centres and the council's server. This service supports a lower socioeconomic population of 61,615, yet as an inner urban municipality parts of it have become increasingly gentrified. In this study the main focus is on library E's public library branches. With the council server traffic removed, library E's data usage over the study period varied from the other three case studies. Unlike these case studies (where there was an overall increase in data usage even though there were months where data usage declined), library E experienced an increase of data from February to May. Downloads were roughly 20Gb a month. Library E then managed to reduce data usage from July to September 2003. During this latter period, the data usage was roughly 16Gb a month. This is despite the fact that the library allows gaming (recognised as one of the larger drivers of data usage by the libraries in the study group), and the proxy server is now only being used for the internet feed to the council servers (site 6). From April onwards, the council server downloads rose from 5Gb to 15 Gb and uploads from 60 Mb to 5.45Gb.
The survey questions for Library E elicited the following response
As with the majority of the other libraries, although bandwidth has increased and this has improved levels of customer satisfaction, it has also raised expectations of even faster connections. The library has had periods of significant growth from January to May 2003, yet the proactive involvement of the council IT has resulted in a reduction in library data. This said, council data usage rose during the study period.
Note that although site 3 is small and only open 60 hours a month, the average hourly usage per terminal figure indicates that when this site is open, it is heavily used. Site 6 did comparable downloads to site 2 from July 2003 to September 2003, but as many more staff terminals use the site 6 connection than user terminals use site 2, the hourly average per terminal for site 6 is much lower than site 2.
Library H is a metropolitan 7 branch regional library service with a relatively homogeneous socioeconomic outer urban population that includes a sizeable concentration of Asian Australians, especially around the main branch (site 2). In addition to the library branches this service also includes an aged care centre. It supports a resident population of 261,263.
The survey questions for Library H elicited the following response
Library H is a large heavily used metropolitan regional library service. During the study this library service's downloads grew from less than 20Gb a month to between 40 and 36Gb a month. For most of the period covered by this research, library H spent considerable time and effort finding out more about user use of the internet, and developing procedures to manage use and expectations. At the same time it tried to avoid escalating data costs. The changes made by library H include blocking chat, data streaming, and filesharing. When users try to do chat or data streaming, they are redirected to a url giving information on the paid public terminals in the branch. Library H has started registering Mac addresses to stop users connecting their own devices to the library network. It is also about to start user session authentication. It appears that if users have to supply their identity in order to commence using a library terminal, they are less likely to try to circumvent the library's internet usage guidelines. Though these changes appear to make a difference in the short term, the longterm trend for library H is for increased data usage.
Also of interest is that during the study period library H was forced to migrate the main branch from an ADSL 6144/640kbps ADSL connection to a 2048/2048kbps SHDSL connection. Although 6144kbps download capacity at the main branch was more than enough to support internet usage and system traffic, the 640kbps up haul was not enough to support internet uploads and the system traffic from their other branches. It was found that a 2048/2048 SHDSL connection overcame the upload bottleneck at the main site.
Library K is a rural 12 branch regional library service with seven branches connected via dedicated ADSL connections per branch. The remaining smaller branches connect to the internet by dialling into the closest ADSL connected branch. The main library and two other branches support a major regional centre, while the other ADSL connected branches support individual country towns and surrounding areas. This service supports a resident population of 156,310.
Although there are periods of declining data usage in February, March and July 2003, during the study period this service experienced a 124% overall increase in downloads.
For the purpose of this study library administration data has been combined with the main branch data, as these two parts of the organisation share the main site's DSL connection. Other than the main site, the DSL connected branches have on average 2.7 user pcs and 2.2 staff pcs per branch. Sites 1, 5 and 6 serve the same regional centre. The other branches each serve their own distinct population centre and surrounding rural district. Site 3 is also a dormitory suburb for Melbourne. After the main branch, site 2 not only has the largest number of pcs and the longest opening hours, it also recently moved into a new branch. This site generates almost twice as much download traffic as site 3, which is the next largest site by number of pcs and hours of opening.
From June 2003 per branch data usage roughly correlated with the number of user terminals per branch and the hours of opening. The exception is site 3 but it is unclear why this is the case. Is it because, as a dormitory suburb for a large metropolitan city, it services a slightly different and larger population to sites 4 and 7? Sites 4 and 7 each support a country town with a single branch. However it was noted by the library service that staff at site 3 have spent more time with the public promoting online services. This, then, could explain this site's usage. Further research is needed in this area.
Despite significant increases in bandwidth across the library service, the user expectation is for even faster connections. Another important issue addressed by library K is the impact on data usage if antiviral software and plugins are not kept up to date. Keeping pc hard drives up date when security such as HDD Sherriff cards have been installed, imposes a significant staff load. This issue has been noted by a number of the other libraries.
One of the more important findings of this study is that it does not take very many users to change their online behaviour before it starts to have an impact on the library's overall data usage and costs. It is also apparent that, in order to support public terminals using broadband connections, libraries are required to devote more attention and considerably more resources to their internet policies. Use can change at any time. To respond effectively the process for setting policies, and configuring networks, has to be fast, adaptive and flexible. When changes are made, libraries need to ensure that the changes are understood by staff, as well as communicated to users. This helps manage expectations. Changes in use include new types of behaviours that the libraries themselves may not have foreseen. For example, the libraries studied did not foresee users would connect their own devices onto the library's own network to upload and or download files. As a result they could not have foreseen the significant impact this behaviour would have on data usage and budgets. That users connect their own devices to a library network also raises legal and risk management issues libraries will have to address.
Vicnet has had discussions with other Australian states and territories about the findings of this study. It appears that the internet data usage of the libraries in this study is significantly greater than that experienced in other parts of Australia. This could be a result of the Victorian libraries implementing broadband internet connections sooner than their interstate colleagues. If this is the case, the findings of this study could have important implications for the roll out and funding of broadband internet connections in other parts of the country. This is especially true if recommendation 7b of the Australian Senate report into libraries in the online environment is adopted. This recommendation advocates that further funds be allocated to expand the national broadband strategy for access in libraries.
In addition to the monetary impact that comes with the increase in internet use, there is bandwidth congestion to consider and the impact this congestion has on service delivery. For example, as mentioned in the library H case study, the library was forced to migrate the main branch from an ADSL to a SHDSL connection to improve catalogue response times. The capacity of the uplink often caused more concern for the libraries. Library F, for example, soon found that with users increasing data use, its own catalogue also started to run more slowly. In fact, with the migration to SHDSL, the library H catalogue ran faster within library F branches than its own catalogue. Subsequently, library F installed a separate SHDSL connection dedicated to its system traffic. As more interactive broadband services become available, and libraries consider implementing services such as video conferencing and voice over IP, symmetrical broadband connections will probably play a much greater role across all branches.
The importance of information
The ability to accurately determine and plan for requirements is dependent on the quality of the information libraries can access about their own usage patterns. It is also dependent on access to readily available benchmark information. The libraries in the study group found that access to an average hourly data use figure helped with benchmarking. Using these benchmark figures removed variables such as the number of terminals and the hours of opening.
This study also shows that it is important to take into consideration the difference between user and staff terminal use of the internet. Because of the study it appears that the level of debate among Victorian libraries on what drives internet data usage has increased. Providing the libraries with the results of their peers and colleagues has established a conduit for information sharing. By learning from the experiences of the other libraries in the study group, these Victorian libraries were better informed, and arguably better positioned, to make proactive decisions on strategies.
Although not part of the case study, Vicnet spent considerable time liaising with library J in order to get data costs down.
One of the strategies the library implemented of its own accord (and this could have benefits for the other libraries) is its uses of the eCommunity network connections to update individual branch proxy servers. As a result, library J uses more data, but because this is branch traffic and remains on the eCommunity network, it is free data.
This, together with other policy changes implemented during the study period, appears to have stabilised the library's internet downloads, helping with speed and congestion. Library J, and the other large libraries, have also migrated to a Vicnet flat data rate. This has been done to take control of the costs.
A number of the libraries also suspected that events such as the Rugby World Cup had a short term impact on internet usage. This study found no conclusive evidence to support this view. For example, the Rugby World Cup was held between 10 October and 22 November 2003. The average hourly download per metro user terminals did increase from 4.48 Mb an hour in September 2003 to 5.24Mb an hour in October and 5.44Mb an hour in November. However the overall internet only downloads for the sample group went from 469,597.13Mb in September to 499,849.21Mb in October but then declined to 427,542.18Mb in November. During this period students were more likely to be focused on their end of year exams and assessments, which seems to have decreased data usage. Further research looking at individual library branch log files for traffic to specific events websites (such as the Athens Olympics) may shed more light on this issue.
On their own, the number of terminals, branches, or hours of opening do not determine internet data usage. Likewise, the capacity of the connection in itself does not. This is apparent when the difference between the average megabyte downloads per terminal, per hour, per branch is considered. There is a considerable diversity between one library service and another. More research is needed, because the Findings demonstrate that internet data usage is driven by a number of factors working in unison. However the main influencing factors between different library services appear to be each library's internet usage policies, and the resources dedicated to supporting and maintaining these policies over time. This conclusion has important ramifications for library policies, budgets and resource allocation.
How libraries respond to these issues will have an important impact on the type of services they offer. It will also impact on the resources they require to deliver those services, and the level of user satisfaction. Although considerable amounts of data usage can be driven by a small group of users, the public in general is demanding more online services. This, together with the reality that online offerings are becoming richer and more varied in their content, means that the amount of data required to support the library services of the future is going to be considerably greater than that required today. Sanders-McMaster notes that
the same strategies that helped libraries get where they are today will be required to move ahead: networking, partnerships, grants, annual upgrades. Those [such as the Victorian public libraries in this study] who are positioned to be the experimenters and early adopters have opportunities and challenges. (12)
Libraries have dealt with change through Experimentation, information sharing, and a collegial approach. This approach will be important for libraries to respond to the wonderful opportunities--and the threats--that the next generation of internet service provides.
Freely sharing internet data sourced from different library services, community groups, NGOs, and government agencies via peering arrangements may become increasingly important. Strategic peering arrangements with commercial data providers may also become more important. Given the increase in data usage, accessing free data through peering, rather than a subsidised community data e-rate, may ultimately prove more valuable for the public library service of the future.
The Victorian eCommunity Network
The benefits of this eCommunity network are that it
Increases broadband connections out to community groups and library branches as many sites have migrated from 64 and 128 ISDN or in some cases 56k dialup connections to a minimum 512 / 128k ADSL connection
Reduces community and library data costs by implementing data aggregation and therefore removing site to site data charges irrespective of whether the sites belong to the same organisation
Allows participating organisations to free data access government and community information hosted on Vicnet at www.vicnet.net.au. Vicnet has offered free community hosting for almost ten years, so this represents a considerable amount of information and is recognised as one of the most successful government sponsored community web publishing programs in the world. Example of participants include: the Victorian Woodworkers' Association, Backpackers Australia, The Australian book review, Young Australian Best Book Award, OzLit Australian Literature, the Dead Persons Society, Friends of the Koalas, and the Motor Neurone Disease Association of Australia. In addition to the free web hosting, Vicnet also offers virtual web services. This is where organisations are hosted at Vicnet but have their own domain name. Examples include Vala, the Melbourne International Comedy Festival, the Royal Australian Ornithological Union, the County Court of Victoria, and the Equal Opportunity Commission Victoria. Access to the Vicnet virtual web server sites does not incur data charges for eCommunity participants. Free hosting and virtual web hosting generated 19,897,157 visits and 395,900,327 hits in the 2002 / 2003 financial year, whereas take up of the free community web hosting is currently growing at 500 community organisations a year
Allows connected sites free access to mc2 resources located http://mc2.vicnet.net.au. mc2 is a Victorian government initiative managed by Vicnet. It offers free and easy to use web based services to community groups so they can publish online, communicate online, and build their own online communities. Mc2 allows the groups to do web email, web publishing, chat, forums, photo galleries, and offers virtual office facilities. The groups participating range from cultural, and sporting groups through to special interest and support and disability groups. Currently there are over 20,000 members in this program
Allows connected sites free data access to information hosted for Victoria's Virtual Library by Vicnet at http://www.libraries.vic.gov.au. This includes the hot topics database (a database of online current affairs issues managed by Victorian public librarians), the Victorian Biography Resource Centre, and the Victorian Library Locater. In addition to the free data, Victoria's Virtual Library provides access to a number of gateways such as Open Road (a gateway to multilingual web resources), Gulliver (full text databases) and Reference Bookshelf (Reference Bookshelf is a directory of web based reference resources)
Allows connected sites free data access to the catalogue and information hosted by the State Library of Victoria This includes information from the State Library's catalogue, online public programs and exhibitions, and the Australian Centre for Youth Literature.
Table 1 Monthly total data downloads by per branch Type of Date Sample Less than Between Downloads Sites 1 Mb 1 and 5 Mb Total Data Oct 2002 52 37% 40% Internet Only Oct 2002 52 37% 40% Total data Oct 2003 63 17% 27% Internet Only Oct 2003 63 19% 25% Type of Between Above Downloads 5 and 10 Mb 10 Mb Total Data 17% 6% Internet Only 17% 6% Total data 24% 32% Internet Only 35% 21% Table 2 Average downloads per terminal per branch and per hour Category Number of sites Jun-03 Jul-03 Metro sites 40 2.07 Mb 2.46 Mb Rural sites 14 0.38 Mb 0.51 Mb Metro user only 8 5.12 Mb 6.04 Mb terminals Metro user only 6 4.30 Mb 4.77 Mb terminals # Metro user and staff 31 1.28 Mb 1.52 Mb terminals Category Aug-03 Sep-03 Oct-03 Nov 03 Metro sites 2.69 Mb 2.59 Mb 2.71 Mb 2.37 Mb Rural sites 0.67 Mb 0.69 Mb 0.78 Mb 0.66 Mb Metro user only 6.84 Mb 6.42 Mb 6.62 Mb 5.68 Mb terminals Metro user only 4.58 Mb 4.48 Mb 5.24 Mb 5.44 Mb terminals # Metro user and staff 1.58 Mb 1.58 Mb 1.66 Mb 1.50 Mb terminals # Metro user only terminals with library B removed. Note that since library B has blocked KaZaa file sharing at a port level, and installed a proxy server, the 8 site metro user data usage is similar to the 6 site metro user only data usage There was not enough data to determine metro staff only terminal usage. Nor was there enough data to determine rural usage by user only, and staff only terminals Table 3 Library B Data downloads results 1 June 2003 to 30 September 2003 Jun data Jul data Aug data Sep data 33,973.63 Mb 49,114.89 Mb 68,017.51 Mb 61,054.27 Mb 22,649.08 Mb 32,743.26 Mb 45,345.00 Mb 40,702.84 Mb User pcs Staff pcs Site Hours open 16 N/A Site 1 72 hours 12 N/A Site 2 72 hours Note site 1 admin data included in Site 1 branch data. Site 2 shares the feed of site 1. According to the library, site 2 takes 40% of the overall traffic. Staff terminals use the council's network to access the internet Table 4 Library B questionnaire results Question Response 1 Drivers of internet Web mail; Lote access to newspapers and usage radio; general browsing; web transactions such as banking, travel bookings, online auctions; access to tertiary institutions by students; school projects, general information searching and research; data streaming; filesharing; employment seeking 2 Changes over last 6 to Changes over the last 6-12 months include 12 months increased access to data streaming; filesharing; internet games have been blocked at the pc level due to policy decision not to allow access to games via the internet as; a) this is not seen as core library service; b) with pcs booked for games use approximately 75% of the time, there was limited access for users wanting to use the internet for other purposes 3 List of staff and user Refer to table 3 terminals 4 Changes to staff / user Added extra site blockings to current internet usage policies Norton's Internet Security Blocking. This had limited success. List for example: Download.com, Download.microsoft.com, KaZaa Downloading websites, and online games websites. Proxy server and firewall have been put in place to block from the port level (ports KaZaa 1214, Napster 6699, 6700, 6701 and Gnutella). Decision was made to block these services due to considerable download costs. The internet conditions of use are under review, as well as guidelines for staff 5 Impact of viral attacks MSBlaster caused problems with Norton's Internet Security program and Windows applications. Another virus created excess traffic on the network and reduced network performance 6 Change in user profile Definite increase in the number of teenage boys using internet for access to datastreaming and filesharing. Anecdotal evidence of new users who have dialup internet connections at home, but prefer to use the library's faster connection for datastreaming and filesharing 7 Change in user Fewer complaints about lack of internet satisfaction access speed 8 Change in user Increased expectations on network speed expectations and overall performance. There is slightly less tolerance of downtime. Downtime has increased recently, with our bandwidth at full capacity during most afternoons 9 Changes to library In recent months one of our branches has staff's work and required more staff surveillance of the responsibility public internet access area to ensure users follow booking guidelines. As the cost of internet access has increased substantially, as well as user expectations regarding reliability and access, staffing resources have been required to review the network set up, reliability, monitoring and management 10 Any other changes No Table 5 Library E Data downloads results 1 June 2003 to 30 September 2003 Jun data Jul data Aug data Sep data 44.98 Mb 101.26 Mb 74.67 Mb 77.14 Mb 17,930.24 Mb 15,504.78 Mb 15,711.93 Mb 17,192.70 Mb 195.56 Mb 432.87 Mb 516.83 Mb 490.18 Mb 0.00 Mb 8.15 Mb 6.40 Mb 45.59 Mb 0.00 Mb 5.16 Mb 6.43 Mb 17.87 Mb 3,838.76 Mb 12,726.29 Mb 12,730.39 Mb 14,878.23 Mb User pcs Staff pcs Site Hours open N/A 2 Site 1 208.0 hr/pm 17 N/A Site 2 260.0 hr/pm 1 2 Site 3 60.7 hr/pm N/A N/A Site 4 N/A N/A N/A Site 5 N/A N/A 146 Site 6 260.0 hr/pm Note Site 2 is the main library branch, site 6 is the council server, sites 4 and 5 are aged care centres. Staff terminals at site 2 are aggregated via site 6. Site 2 includes the aggregation of two other library branches via the council's own line of site network Table 6 Library E questionnaire results Question Response 1 Drivers of internet Main web usage appears to be chat, gaming usage and internet mail (taken from random views of machines 2 Changes over last 6 to There does not appear to have been any 12 months change 3 List of staff and user Refer to table 5. In Oct 2003 ran 296 terminals email users, and around 150 staff internet users. Libraries have approx. 40 public access pcs 4 Changes to staff / user Internet feed has been split in half, one internet usage policies 2Mb feed for staff/mail usage, and one feed to public access (data for each feed is listed separately in figure 8). Proxy server was relocated after feed split, and is currently running on corporate network only. 5 Impact of viral attacks Viral attacks do not appear to have made much difference to internet usage, some regular virus attacks occur each day but are probably not contributing much to internet bandwidth. 6 Change in user profile Unknown 7 Change in user Satisfaction level definitely increased satisfaction 8 Change in user Expectations of availability increased expectations significantly 9 Changes to library No change staff's work and responsibility 10 Any other changes Data pattern has changed (larger downloads from internet), some staff abuse (Virgin Blue/Qantas type sites) and extremely large email attachments Table 7 Library H data downloads 1 June 2003 to 30 September 2003 Jun data Jul data Aug data Sep data 37.73 Mb 50.44 Mb 85.69 Mb 1,700.48 Mb 19,761.53 Mb 24,803.42 Mb 29,624.27 Mb 24,531.21 Mb 1,048.84 Mb 962.09 Mb 1,827.32 Mb 2,032.21 Mb 0.96 Mb 1.36 Mb 375.24 Mb 5,655.19 Mb 4,193.03 Mb 5,197.28 Mb 5,932.49 Mb 2,597.53 Mb 2,293.70 Mb 2,726.74 Mb 2,233.11 Mb 4,207.93 Mb 3,920.98 Mb 4,811.76 Mb 4,123.72 Mb 1,752.79 Mb 2,131.12 Mb 2,149.55 Mb 1,701.21 Mb 122.96 Mb User pcs Staff pcs Site Hours open N/A N/A Site 1 N/A 37 62 Site 2 273.0 hr/pm 8 4 Site 3 156.0 hr/pm N/A N/A Site 4 N/A 9 12 Site 5 242.7 hr/pm 5 11 Site 6 188.5 hr/pm 6 9 Site 7 238.3 hr/pm 4 7 Site 8 208.0 hr/pm Note Site 2 is the main library branch, site 1 is an aged care centre. The staff and user terminals at two additional branches are aggregated via site 2 Table 8 Library H questionnaire results Question Response 1 Drivers of internet The main driver for internet usage within usage our organisation is mostly determined by our clients' usage of the various internet accessible public personal computers available at one of our eight branch libraries. The main internet usage is still internet based mail, messaging and online gaming. 2 Changes over last 6 to We have found that internet based mail, 12 months messaging and online gaming type of usage has grown to such an extent over the last 12 months that free public access personal computers were wholly used for this type of internet access, and other types of web based usage has been suffering because of this. 3 List of staff and user Refer to table 7 terminals 4 Changes to staff/ user The organisation has tried a passive internet usage policies implementation to curb the proliferation of internet based mail, messaging and online gaming and is continuing to deliver internal web page based links via subscriptions to online data bases and information pages, and thus attempting to drive our clients' online usage of the internet. Furthermore we have segmented half of our personal computers at branch libraries to be for internal web page based links via subscriptions to online data bases and information pages, and have blocked access to web mail, messaging and chat sessions on these personal computers 5 Impact of viral attacks Due to the type of server operating systems and our approach to viral protection we have not suffered any viral attacks to our internal system over the past six months; however access to other sites has limited usage slightly 6 Change in user profile Initially our user profile skewed towards the younger after school person using the internet for gaming and mail access. However since we split the public internet access personal computers into mail, messaging and gaming (a pay for use service) and allowed only information searching on all other public access personal computers (a free use service) our user profile has shown a return to a more balanced use based on age and sex. Our proxy statistics have shown that the types of information gathering by library users has changed as we moved from a satellite with 64 Kb ISDN back channel connection to DSL from each branch. Intensive data streaming sites are being accessed that include video, radio and voice. This include sites such as, such as the ABC (News Radio) and other multicultural sites, such as Sina.com, an Asian news sight with links to intensive data streaming 7 Change in user With the increase in public access booking satisfaction across all branches up dynamically, we feel that general satisfaction is use as a result to the implementation of the DSL. 8 Change in user We have not done any surveying in this expectations area so I am unable to comment. 9 Changes to library With the implementation of DSL we staff's work and installed proxy servers at all branches responsibility with the ability to allow blocking of sites specific to a personal computer and thus the libraries' work load to police internet usage has decrease greatly. 10 Any other changes One large change that has occurred is the increased cost of providing access to the internet. Our download costs have doubled and we foresee this occurrence to continue in the same way over the next two years as more products go line and websites become more interactive with the inclusion of audio, video streaming and flash product. This phenomenon of increased download cost presents our organisation with a grave problem if we are to continue to provide public access to the internet at our current level of service. Our diminishing financial resources will greatly limit our ability in the future to provide as many public access terminals as we now have, therefore we see an impending need to be downgrading our service to be even able to provide a public internet access service Table 9 Library K data downloads results 1 June 2003 to 30 September 2003 Jun data Jul data Aug data Sep data Note 1 Note 1 Note 1 Note 1 9,499.05 Mb 6,303.12 Mb 6,891.87 Mb 8,196.47 Mb 1,509.54 Mb 1,957.49 Mb 2,626.10 Mb 2,534.13 Mb 666.17 Mb 1,029.73 Mb 1,293.21 Mb 1,346.75 Mb 736.70 Mb 931.84 Mb 1,244.32 Mb 1,190.46 Mb 361.65 Mb 583.18 Mb 708.30 Mb 657.29 Mb N/A 70.21 Mb 316.68 Mb 368.02 Mb 148.60 Mb 168.45 Mb 205.42 Mb 273.52 Mb User pcs Staff pcs Site Hours open N/A 10 Site 1 admin N/A 9 10 Site 1 branch 46 hr/pm 4 3 Site 2 40 hr/pm 3 2 Site 3 35 hr/pm 2 4 Site 6 38 hr/pm 3 2 Site 4 39 hr/pm 3 1 Site 7 21 hr/pm 1 1 Site 5 14 hr/pm Note Site 1 admin data included in site 1 branch data Table 10 Library K questionnaire results Question Response 1 Drivers of internet usage Email is the main driver, however some gaming and games. Also accessing Gulliver consortium full text databases 2 Changes over last 6 to 12 months None 3 List of staff and user terminals Refer to table 9 4 Changes to staff / user internet usage policies 5 Impact of viral attacks Internet pcs in the main branch have had attacks from viruses that pull down large amounts of data. These pes have a Sherriff card in them so the updates from Vet are not kept once the machines are rebooted 6 Change in user profile None 7 Change in user satisfaction Still get complaints about speed of connections 8 Change in user expectations Want internet to be faster 9 Changes to library staff's work None, bookings are still done by and responsibility staff; all pcs are booked out 10 Any other changes Figure 1 Sample Netflow Data (last 2 days of August plus monthly summary) Site Date Free Down Downloads Total Down Site 1 Aug 29 2003 4.06 92.25 96.31 Site 2 Aug 29 2003 10.16 51.44 61.61 Site 3 Aug 29 2003 19.12 867.57 886.69 Site 1 Aug 30 2003 5.63 55.05 60.68 Site 3 Aug 30 2003 6.08 13.76 19.84 Site 4 Aug 30 2003 6.05 100.85 106.90 Site Free Down Downloads Total Down Free Up Site 1 492.64 23,079.86 23,572.50 525.78 Site 2 318.99 1,093.05 1,412.03 256.63 Site 3 257.63 4,735.82 4,993.46 245.08 MONTH TOTAL 1,674.28 38,280.94 39,955.22 1,613.02 Site Free Up Up Total Up Site 1 3.77 10.4 14.17 Site 2 1.44 0.74 2.18 Site 3 25.34 129.66 155.01 Site 1 6.14 3.32 9.46 Site 3 3.64 0.44 4.08 Site 4 8.28 22.31 30.59 Site Uploads Total Up Kbps Site 1 3,282.48 3,808.26 66.43 kbps Site 2 142.60 399.23 3.98 kbps Site 3 393.20 638.28 14.07 kbps MONTH TOTAL 4,436.02 6,049.68
(1) Hardy G and Johanson G Characteristics and choices of public access internet users in Victorian public libraries Centre for Community Networking Research, Monash University 2002 p24
(2) Feighan, D and Schmidt, P Gee I didn't think it was going to be that much: a report on the issues and implications of technically sustainable and affordable bandwidth for Australian libraries e-volving information futures, proceedings of the Vala national conference on library automation Melbourne 6-8 February 2002 2 p611
(3) Gates, W The road ahead London, Viking 1995 p136
(4) Taylor, P The networked home: domestication of information Journal of the Royal Society of the Arts April 1995 pp42-44
(5) Hall, P Cities in civilization: culture, innovation, and urban order London, Weidenfeld & Nicolson 1998
(6) Porter, M Strategy and the internet Harvard business review 79(3) March 2001 p62
(7) Feighan and Schmidt op cit p604
(8) Libraries in the online environment Australian Senate Environment, Communications, Information Technology and the Arts Reference Committee 2003 http://www.aph.gov.au/senate/committee/ecita_ctte/on line_libraries/report/index.htm
(9) Report on a policy for public access to the internet for Victoria Hobart, Trinitas 2000
(10) Oder, N Libraries get 4% of e-rate funds Library journal 128(16) 2003 p20
(11) Adams, B and Meagher, B Broadband quality of service issues: consumer perspectives a report for the Australian Communication Authority by Dandolo-partners July 2003 p8
(12) Sanders-McMaster, L Internet 2: an overview of the next generation of the internet Computers in libraries 17 (3) March 1997 p57
David Feighan is the ISP Manager for Vicnet. Address: Vicnet State Library of Melbourne Swanston Street Melbourne Vic 3000 tel(03)86647001 email email@example.com
David Feighan ISP Manager Vicnet State Library of Victoria
|Printer friendly Cite/link Email Feedback|
|Title Annotation:||Vicnet eCommunity network|
|Publication:||Australasian Public Libraries and Information Services|
|Date:||Sep 1, 2004|
|Previous Article:||User satisfaction surveys.|
|Next Article:||Beyond four walls: adult literacy services in Queensland public libraries.|