The concept of contention ratio comes from terrestrial networks where typically, when they were once switched, they referred to the number of available ports for a fixed population of users. For example, a telephone exchange had 10 ports for 100 users therefore the contention was 10:1. The 11th caller on a such a network would get a “busy” tone, suggesting the called recipient was engaged. They were not, the exchange was!
Contention ratios exist because the cost of providing a 1:1 or dedicated service to individuals leads to a price point that is unacceptably high. Therefore the concept of resource sharing is introduced; the trade-off is therefore: price of service versus quality of service.
This same concept followed through into shared terrestrial data networks (eg. ADSL) and was used to define the amount of bandwidth available for a population of users (since each user had a dedicated DSLAM at the exchange the contention or concurrency was only seen between the DSLAM and the core network - known as the "middle mile"). So for example, for a fixed middle mile of 500kbps and a population of 10 ADSL customers (each with max 500kbps) the contention is 10:1. The difference being that the first customer online gets 500kbps, the second concurrent customer gets 256kbps (and then so does the first) and so on. You see the degradation of performance according to the number of concurrent users, so on a typical business network it is slow during the day and speeds up at night. Telcos will typically size their middle mile (or even last mile) according to the % fill they see in their core network. Anything approaching 80% and they throw more bandwidth at it – because it is relatively cheap – thus obscuring the true effects of an over-subscribed or over-utilized network. Contention is therefore only relevant for the middle mile for big telcos offering terrestrial bandwidth across national networks.
For satellite operators the rules are different. The satellite network is the first, last and middle mile, therefore there is no cheap core network to push the effects of contention to the edge of the network where published figures of 10:1 and 20:1 are common. Satellite bandwidth is very expensive and so contention ratios are therefore much higher on large shared satellite networks than they are on terrestrial networks for an equivalent service; because of this, satellite bandwidth is managed in a more creative way. Additionally, satellite networks can either be sized with a CIR (constant information rate) and/or a BIR (burst information rate), the latter suggesting that the burst rate is fixed but the minimum rate (CIR) is not. These are the top-level parameters for determining the performance of any given site in a satellite network.
Let's take the CIR network first (where contention is referred to as “over booking”). A fixed amount (say 500kbps) is allocated to a network and 10 VSATs (with max. BIR 500kbps) are installed. The contention is 10:1 since an allocation of 500kbps is shared between 10 VSATs. As each VSAT uses the network concurrently the actual performance does not degrade linearly with each concurrent use. This is because satellite access schemes (whether CIR or BIR are defined or not) are not the same as terrestrial partly because of the inherent delay in satellite communications that does not exist in terrestrial, and partly because of the history behind the schemes developed for satellite communications.
Firstly, the outbound direction (from the hub to the remote site) is multiplexed into frames (e.g. using DVB-S) much like terrestrial Frame Relay, however, due to the expense of satellite bandwidth compared to terrestrial, the encoding and compression schemes deployed on broadcast satellite bandwidth have a much greater efficiency in comparison with terrestrial schemes. On the inbound (from the remote site to the hub) the network topology is classically a star configuration therefore each site is competing with the other for a “share” of the time and frequency allocated to the network as a whole. At its most congested (10 concurrent VSATs active all sending/receiving in a pool of 500kbps) the CIR for each site would be 50kbps. If this were a customer network they may assume that the performance will fall to around 50kbps minimum when fully loaded. However, what the customer sees at its most congested may be higher or lower than 50kbps according to the type of applications & protocols being used and the behaviour of the access scheme in place. This is because some applications (file transfer for example) will “grab” an allocation of bandwidth for the duration of the its session and may not use it efficiently for 100% of the time. Similarly, applications that are very “chatty” will use the available bandwidth inefficiently as much of the data transmitted is control information and the volume of “chat” leads to a significant increase in collisions and retries (as a consequence of competing for the time and frequency allocated to the network) which are inherently inefficient in their use of bandwidth. Contention is therefore applied in a meaningless way to the inbound as it can never be fully controlled nor predicted.
On a network where the BIR is fixed (but not the CIR) the same effects are at play but there is no minimum committed information rate. As these networks are subject (increasingly) to abuse by a few individuals who use all the bandwidth, reputable operators, such as Satlynx, implement complex Quality of Service (QoS) tools and Fair Usage Policies to ensure the network is shared fairly amongst all users. Therefore, on the Satlynx network the concept of Contention Ratio is meaningless because the QoS is dynamically controlled to allocate sufficient bandwidth to each user with a performance commensurate with the service class they are using and their recorded utilisation of the network over a defined period. Therefore, Satlynx dynamically manages the performance of individual users according to the time of day and network loading so that the actual contention is managed to avoid the typical network congestion associated with contended shared networks.
In other words, for a population of 10 VSATs sharing a fixed allocation of bandwidth, if one VSAT is not restricted under the terms of the Fair Usage Policy and the other 9 are restricted (at different levels of restrictions) the Contention Ratio will actually be higher than 10:1 since the demand on the network from the restricted sites will be lower than the site that is not restricted. Contention ratio is therefore indeterminate.
Typically, when a network is designed and dimensioned contention can be calculated as (Peak/Allocated per site) kbps. This leads to an artificially high contention ratio but in reality the Fair Usage Policies ensure that this worst-case is never seen. The more important value is the real contention ratio in the network, which varies constantly and can be calculated as (Peak/Actual Performance) kbps.
In practise, utilising QoS tools, the actual contention on the satellite network is often much less than values published by ADSL service providers.
If Satlynx did not control QoS and then published contention ratios then these ratios would be false. This is also attributable to the way in which such a network starts its life and grows. There is always idle capacity on such a network at the beginning (or any subsequent upgrade to it). This “over allocation” means that the actual performance of VSATs is very good when new bandwidth is added to such a network (equivalent to 2:1 or 5:1) but this strategy mismanages expectations of end users and ultimately leads to a large population of very unhappy customers when the network becomes loaded to its published contention ratio. Even at a point where the network is fully loaded the contention ratio would still not be true because of the effects of the satellite access schemes which, compared to terrestrial, are non-linear. There are classic examples of this in the form of service providers who once made promises of contention ratios and had happy customers at the beginning but hit problems as their networks became loaded, these service providers are sadly no longer in business.
Therefore, Satlynx does not publish contention ratios on its shared networks.
Send an email to Contact Us
or call us now on +353 (0)27 30040 if you still have questions.
