In the B2B space, broadband provision is based on a directory structure. A wholesale broadband provider provides high quality IP services to Internet Providers at wholesale discounts. Businesses (end users) hook up to the networks of Internet Provider (ISP) (i. e. Internet connection wholesale pipes retailers). The wholesale broadband provider, will buy its broadband from carrier partners (Examples of carrier partners are: British Telephony; AT&T (United States); Sprint (United States)) and wholesale it to businesses down the archipelago at discounts. The common service offered by the wholesale broadband provider is that it attaches end-user sites to the networks of the ISPs, but no place else (there is no connection provided to the broader Internet).
Typically, the end users are linked to the ISPs networks via ADSL and SDSL services ultimately offered by Carriers to the wholesale provider and then invoiced about the ISP. These services are delivered to the wholesale provider’s network over connections known as “central pipes” or “centrals”. The wholesale provider then aggregates the data flows to and from these end-user sites and forwards them to the ISPs over a dedicated connection between the two companies’ networks.
An important area of the method of getting broadband is “RADIUS” and proxy “RADIUS”. RADIUS stands for “Remote Authentication Switch In User Service” and is defined by the document “RFC 2865” as modified by RFCs 2868, 3575, and 5080; many RFCs also define related matters. (RFCs are the series of documents which define how the Internet works and are available online. Despite the name “Dial In”, RADIUS is used on DSL connections as well.
The intention of RADIUS is to separate the authentication of connections from the devices handling their traffic. Think about a typical DSL connection to an ISP. There will be a modem at the end-user’s driveway that attaches, over Carrier’s service, to a “gateway” or “LTS” on the ISP’s network. The gateway then forwards the customer’s Internet traffic to various places as appropriate. An ISP will typically have many gateways spread around their network and cannot necessarily predict which gateway a user will hook up to. The vast majority of the work a gateway does is forwarding IP packets and its design is situated around this. These facts yet others mean that it is undesirable for the gateway to also do the work of deciding whether a user is eligible to hook up to the ISP’s service in order to check whether their modem has presented the correct security password. Carrier’s also have the issue of ensuring that the bond is being told her i would the correct ISP in the first place.
Instead, when a new DSL connection is being set up (for example, each time the customer’s modem is turned on), a tool within Carrier’s network called the B-RAS (“Broadband Remote Access Server”) takes the user name and security password offered by the modem, determines which ISP it relates to, and then passes them to another computer in the ISP’s network called a “RADIUS server”. This makes the relevant checks and provide the B-RAS a “go/no-go” decision; it can also provide other instructions to the B-RAS and the ISP’s gateway.
In some situations the ISP’s RADIUS server cannot actually choose directly but must check out of another device, this arrangement is called “proxy RADIUS”.
RADIUS servers are merely involved when a new end-user connection is defined up, not whilst it is active. If the person turns their modem don and doff again, a new connection needs to be set up. If there is a fault in the carriers network causing several connections to be broken and then re-made, all of those connections will cause simultaneous asks from the B-RAS to the RADIUS servers; for example, if the equipment in a particular carriers exchange resets itself as a result of fault, all the DSL users linked to that exchange or that piece of equipment will have to re-establish their connections.
There is no such thing as a 100% reliable service. It is possible to provide an Internet to a active of reliability by, for example, providing several servers where one would do or by copying essential connections through physically diverse avenues, so that damage to one route will not affect the alternate. However, doing this can be expensive, because more equipment must be purchased or hired, the costs for housing machines are higher, as are the electricity bills, and so on. These costs all need to be passed about the customer.
Therefore ISPs, particularly those offering a wholesale service, will choose a level of reliability to offer their customers and design their internal network around that choice. This level is typically termed a portion availability; e. grams. 99. 9% (“three nines”) indicates that the service will be inaccessible no more than 8¾ hours a year, while some telephone systems services are rated at 99. 999% (“five nines”), or no more than 5 minutes 16 seconds unavailability a year. The ISP will offer this service level, defined in some more descriptive manner, and will back it with some kind of guarantee (e. grams. credits at a rate of X pounds for each minute or hour outage beyond the agreed amount). Sometimes either reliability levels emerges with different prices.
Where the ISP’s service makes use of facilities bought in from elsewhere, a commercial choice has to be made. The ISP must exclude the bought-in component from its service guarantee, accept that it sports to pay compensation for outages caused by the bought-in service (and price its service to match), or pay the supplier for an high enough level of reliability for that service. This last case does not mean the same level as it is providing, since this would provide no “failure budget” for its own services but, rather, a higher level. For example, if the ISP is providing a 99. 9% service, those 8¾ hours have to be shared between problems alone network and problems on the bought-in service. If they are shared equally, all the two needs to be 99. 95% reliable.
The test of whether the connection is available involves sending a regular sequence of standard test packets (“ICMP ping”) from some unspecified location within wholesale provider’s network. Such a test would prove only that IP packets are traversing the bond in both directions; it would not demonstrate any other qualities, such as the maximum bandwidth being fed or what fraction of packets was being dangerous. In particular, it would not demonstrate that any end-user was connected through which an ISP. The wholesale provider will need to carry out the test as described.
There is an obvious need for a site level agreement or guarantee of reliability for the differing of the wholesale provider’s network.
Careful consideration needs to be built to the “force majeure” term. Specifically you need to consider carrier’s provision of DSL service to the end users. Is there a SLA on offer for this the main service? Carrier’s ADSL service, in particular, is a low border one and carrier’s SLAs are accordingly low. The standard service aims to fix faults within 40 time hours of being reported, while the “enhanced care” service at extra cost reduces this to 20 hours. In each case compensation is payable if this is not met (though the details do not appear to be publically available). Even if there is no specific SLA on offer, a reasonable wholesale provider would pass fault reports on to carrier’s regularly and keep track of their response, only when to be able to claim contrary to the carrier’s own SLA.
A fault in a service of the kind being provided could have any of several causes:
- A fault at the end customer’s site.
- A problem with the telephone line used to carry the DSL service.
- A fault within carrier’s network.
- A fault on the carrier’s central water line.
- A fault within the wholesale broadband suppliers’ network.
- A fault on the connection between wholesaler and ISP.
- A fault within the ISP’s network.
The first two of these would tend to affect single end-users. The next two are likely to manifest themselves as impacting blocks of end-users simultaneously. A fault within either company’s network could have a range of effects. Finally, a fault on the connection between them would be likely to affect every end-user simultaneously. If there is actually more than one connection, then depending on the technical arrangements, a fault on one connection might have no effect, since all traffic is moved to one of the remaining ones, or might affect a subset of end-users.
Migration access codes
Migration Access Codes, or Mac pcs, were a process set up by Ofcom (UK Telecoms Regulator) to allow end-users to transfer between ISPs while at the same time preventing “slamming” (ISPs transferring end-users to their service without the patron’s permission). Ofcom considers that competition in the ISP market is enhanced by making it straightforward for consumers to transfer between ISPs that both use the same wholesale service. In principle all that is essential for the transfer to occur is for the assorted accounts receivable records in the two ISPs and at carriers’ to be updated, and for the user to be assigned a new username and password relating to the new ISP; no physical action is needed at the telephone exchange. Without such a transfer mechanism, however, the only way to do this is for the customer to terminate their service with the first ISP and then order a new one with the second. This will involve several days without service, to no obvious benefit. There are commercial and in business issues to be addressed. If the “losing” ISP has control over the arrangements, they have no bonus to make them efficient or quick. On the other hand, if the “gaining” ISP can simply order a move, there is a risk an ISP would move a buyer without their agreement (known as “slamming” that customer), potentially causing them significant trouble. Therefore Ofcom introduced the “MAC code” system on 14th February 2007 via General Condition 23, replacing a pre-existing voluntary code.
Under it, an end-user desperate to move their service, or their agent, asks a “MAC code” from their current ISP. The ISP must make a reasonable effort to authenticate the customer (so as to prevent slamming) and then transfer the code from the carrier’s wholesale (or the appropriate provider) and issue it to the customer. The customer can then take this MACINTOSH PERSONAL COMPUTER code to their planned new ISP, who in turn uses it as their authority to achieve the service transferred. Mac pcs are valid for 30 days from issue; once a MACINTOSH PERSONAL COMPUTER has expired, it cannot be used to migrate the customer. However, the customer can request a new MACINTOSH PERSONAL COMPUTER at any time after the previous one expired. On the other hand, the customer does not have to transfer their service simply because they have requested a MACINTOSH PERSONAL COMPUTER: it is purely an enabling tool and expiry of a MACINTOSH PERSONAL COMPUTER is explicitly not grounds to cancel a patron’s service.
ISPs must issue the MACINTOSH PERSONAL COMPUTER within 5 business days on paper or by e-mail (unless the original request was by telephone and a code was issued during the call).
They may only not allow on the grounds that:
- they cannot authenticate the customer;
- the service was already terminated or is in the process of being terminated;
- a still-valid MACINTOSH PERSONAL COMPUTER was already issued (but they must remind the customer of these MAC); or
- they cannot get a MACINTOSH PERSONAL COMPUTER from their own wholesale provider.
The MACINTOSH PERSONAL COMPUTER process is policed and forced by Ofcom under the Communications Act 2003, sections. 94-104. Ofcom has an ongoing enforcement programme in relation to MACINTOSH PERSONAL COMPUTER codes; in respect of an ISP, this normally operates in three levels:
- Ofcom has informal discussions with the ISP over the issue, and may use its powers to request information from the ISP.
- Ofcom informs the ISP that it has reason to believe that it is in breach of the General Condition and sets a timeline for the ISP to remedy the situation. The ISP has to be able to make representations.
- If Ofcom is satisfied that the ISP has failed to remedy the situation and/or to meet the prerequisites in the notification, it can fine the ISP.
A term in a Master Agreement may not prevent the Ofcom regulations overseeing MACINTOSH PERSONAL COMPUTER codes.