ANS Plan for Commercial Services Introduction Executive Summary In December 1990, Advanced Network & Services, Inc. (ANS), working closely with representatives from FARNET, NSF and a group of midlevel networks, began a series of discussions to establish both the mechanisms and a process for supporting commercial services over segments of the Internet. The group's goal was to enhance research and education by broadening connectivity to the network, and its deliberations resulted in a plan for commercial services, which includes new ANS network attachment offerings called "CO+RE Services." Incorporated within CO+RE is a mechanism for funding further development of the network infrastructure. The following plan embodies the ideas and combined efforts of many people, and it will continue to evolve as we gain more experience and insight into building gateways for commercial use. Background The Internet has evolved over several years into a collection of independent, interconnected networks that share some common features. These common features include technical interfaces to the network, application services and in some cases, the policies governing the use of the networks. Government grants often are applied directly to support specific network infrastructures, and this results in policies that restrict the network to those uses related to, or in support of, research and education (R&E). The demand for commercial services from both the R&E and commercial communities has driven a need for change in these restrictive policies, which severely hinder communication between the two groups. Although network service providers have stimulated some experimental commercial participation on the Internet, many commercial institutions still have found it difficult to participate in a broad way due to the current acceptable use policies. These organizations do not want a general purpose tool to be limited artificially to certain kinds of uses. 1 In response to this demand, several pioneering network service providers have emerged to address the specific needs of the commercial community by creating new private network infrastructures. Because these service providers do not receive direct government grants, they are free of the policy restrictions on acceptable use. However, to remain free of these policies, they must maintain some level of isolation from the government-sponsored segments of the Internet, and this limits the connectivity, and therefore, the utility of the network. For the national network to fulfill its potential, it must support pervasive connectivity and access without the technology or policy barriers that isolate the private networks. Each service provider offers a unique set of services and maintains an infrastructure that differs significantly in size, cost and mission. Because of these differences the long-term economic viability and self-sufficiency of the national network will require mechanisms for pricing both the services and the resulting settlements between service providers that would fund the growth of the shared segments of the network's infrastructure. 2 Plan Summary The following summarizes the ANS plan for supporting commercial traffic on the Internet: 1. Establishment of ANS CO+RE Systems, Inc. ("ANS CO+RE"), a wholly owned taxable subsidiary of ANS. This enables ANS to carry commercial or mixed CO and RE across its network. 2. An ANS CO+RE acceptable use policy that supports commercial as well as R&E usage across networks that are interconnected via the ANS CO+RE network. 3. An ANS Internet Connectivity Agreement, which is open to participation by all network service providers. 4. An ANS CO+RE gateway attachment agreement between ANS CO+RE and network service providers that agree to host commercial and R&E institutions. This includes midlevel network service providers for which funding is provided by an NSF grant. The gateway attachment fees include a contribution to an "Infrastructure Pool" funded by the commercial traffic. 5. An ANS CO+RE enterprise attachment agreement between ANS CO+RE and individual institutions that are engaged in commercial activity, but do not provide network services to other institutions. This attachment agreement also includes fees that fund the infrastructure pool. 6. Establishment and deployment of a mechanism for identifying and measuring commercial and R&E traffic at the gateway between two network service providers, in order to support financial settlements. 7. A mechanism for providing funds to improve the network infrastructure, which insures that commercial users contribute to the future of the infrastructure and its development: the "infrastructure pool". 3 The Plan ANS CO+RE Acceptable Use Policy ANS CO+RE has established an acceptable use policy (AUP) for its commercial customers. This policy complements the NSF and ANS AUPs in that it supports the exchange of commercial traffic between networks that are reachable through the ANS network. The CO+RE AUP is comparable to the current ANS and NSF AUP, except that earlier restrictions on use of the ANS-managed networks for commercial purposes are removed. ANS/ANS CO+RE Internet Connectivity (AIC) Agreement ANS/ANS CO+RE has created an Internet Connectivity Agreement as the basis from which all participating network service providers can work together to enhance the network communication between R&E and the commercial communities, as well as among members of the R&E community. The AIC also is designed to strengthen the breadth and quality of information and services available through the networks. By signing the AIC Agreement, ANS and the participating service providers agree to permit the flow of traffic between R&E institutions that they host, and commercial institutions or networks reachable through the ANS Network. IP network numbers are used to determine the commercial or R&E classification of traffic that traverses the ANS network. Organizations providing network service rely on the controls and mechanisms of the attaching organizations to establish and maintain these unique network identifiers. An "Infrastructure Pool" is created as a result of revenues from the ANS CO+RE commercial attachments. Participants in the AIC Agreement, can include organizations that have not agreed to permit direct commercial attachments to their networks. Simply by allowing commercial traffic to be received by their users, those institutions are eligible for an allocation from the pool to enhance their infrastructures. 4 ANS CO+RE Gateway Service Terms and Conditions An ANS/ANS CO+RE Gateway Service attachment permits the exchange of both commercial and research/education traffic. Once a service provider acquires a gateway attachment, it is authorized by ANS CO+RE to provide commercial institutions with attachments to its network, and those institutions may use the ANS/ANS CO+RE network gateway. Institutional Classifications The institution determines the price of a gateway. An institution is defined as an organization that maintains a unique network identifier1 and may be classified as either CO or RE, but not both. The classification of an institution as RE or CO is determined by the ANS AUP or ANS CO+RE AUP that the network service provider agrees to support. ANS depends upon the network service provider to assign all unique network identifiers and classifications (RE or CO) for institutions that use the Gateway. Institutions with activities that can be classified as both RE or CO may choose a CO network identifier for such mixed traffic environments, or may choose to have multiple attachments, one for CO and another for RE. Examples of these institutions might include a university, a commercial laboratory or a not-for-profit research facility2. Network service providers that contract for a Gateway Attachment Service may support both RE and CO institutions. The traffic traversing the Gateway is identified at the gateway by the unique network identifier and is measured in units defined as "COMBits3." For each Gateway, independent COMBit measurements are recorded, and the percentages of total RE usage and CO usage are calculated. The Gateway Service is available to all network service providers, including those midlevels currently being supported by an NSF grant. Under the terms of the gateway agreement, midlevels may provide network connectivity via the Gateway Attachment to other commercial institutions within their service areas. Such commercial institutions also may apply directly to ANS CO+RE for service attachments. ---------- 1 See Attachment A 2 Definitions of institution categories are provided in Attachment C. 3 A COMBit is a Linear COMbination of Bits and is defined in Attachment A. 5 The existing NSF grants allocated to support midlevel gateway attachments will continue to support all R&E classified traffic that traverses the gateways at no additional cost to the midlevel. All commercial institutions that make use of these attachments are priced separately under the Gateway agreement. The fees from these attachments enables ANS to recover the costs associated with the commercial traffic and contribute additional funds to the infrastructure pool. ANS Gateway Pricing Model ANS establishes a price for each Gateway based upon a flat charge for all educational institutions, a surcharge per research or commercial institution, and the relative percentage of total CO traffic generated by the institution. Gateway service fees are based upon several factors unique to each network service provider. They include the bandwidth of the attachment, the type of institution attached (i.e. educational, commercial or research), the unique local exchange carrier circuit costs and the geographic location of the attachment(s). The following are major elements which comprise the Gateway pricing model: - There is a predetermined price for each attachment, exclusive of the local exchange carrier circuit charges. - For service providers hosting commercial institutions, the Gateway price is based on the institutional classification (see Attachment C) and the projected percentage of commercial traffic. - A Gateway attachment customer, such as a midlevel, can connect as many educational institutions4 as if can support without paying any additional fees beyond the base price. The difference between the ANS CO+RE attachment fee resulting from commercial traffic and what would have been charged for an R&E-only attachment, is allocated to the infrastructure pool. ANS, the NSF, participating network service providers and other organizations as appropriate, will have representation on the ANS Resource Allocation Committee (RAC), which will direct the allocation of the infrastructure pool funds. Each gateway service provider establishes its own prices, terms and conditions for the attachments to its gateway. ---------- 4 As defined in Attachment C. 6 Calculation of Gateway Fees The Gateway service fees, which are assessed in advance of actual service, may be based upon historical traffic-type (COMBit) data. During the initial interconnection phase for new institutions where historic COMBit statistics are unavailable, an agreed to COMBit percentage estimate will be used. As indicated in Attachment A, the measurement will be made at the Gateway and the statistics are expected to be auditable. The projections and prices will be revised periodically based upon changing COMBit measurements, as new institutions join or disconnect from the network service provider or as market conditions vary. With this mechanism, the service provider has a fixed Gateway service fee for a specific mix of institutions and traffic distribution over a given period of time. Although this method is imprecise, it does allow for a usage-insensitive gateway price in advance of the service, which helps the service provider budget more accurately, and minimizes the administrative costs for all parties involved. To illustrate this method in another way, there exists a minimum price which is comprised of 100% RE traffic, and a maximum price which is comprised of 100% CO traffic. The COMBit mechanism will determine the actual price charged, which falls between the minimum (100% RE) and the maximum (100% CO). The ANS CO+RE Infrastructure Pool The purpose of the infrastructure pool is to use a portion of the CO revenues to improve the research and education network infrastructure. For independent midlevel networks, the price differential between the ANS CO+RE commercial price and the ANS base price defines the infrastructure pool funds available. For situations where NSF has directly provided grant funding to the backbone supplier in support of a midlevel, the ANS research and education base prices is zero. This mechanism ensures that the commercial use of gateways accessing the Internet funds an appropriate share of the network costs. A portion of the revenue from direct-attached (single institution) CO+RE customers, also funds the infrastructure pool. Funds for the infrastructure pool are sourced only from ANS CO+RE direct attachments or attachment revenues from networks hosting CO institutions. Infrastructure Pool Resource Allocations A Resource Allocation Committee (RAC) will be established to determine the distribution of the funds within the pool. The Committee will be chaired by the ANS Chief Executive Officer, and 7 have representation from the NSF, several network service providers who are participants in the AIC Agreement and other parties as appropriate. The committee will meet quarterly to approve projects and monitor the progress of funded projects. The initial tasks of the committee will be to establish priorities for fund allocation, establish operating procedures and issue specific guidelines for the approval of projects. ANS Gateway Pricing Details and Example This section is to be used in conjunction with Attachment B. It is a detailed explanation of the calculations required to determine the price for the Gateway Service and the determination of the funding to be contributed to the infrastructure pool. The calculation steps are as follows: a) In Example A, a summary of institutions attached to the midlevel is listed and identified by type of institution and size of attachment. b) There is a base gateway charge of $55,000, which includes any and all educational users that are connected. For each additional research or commercial connection, there will be a charge per attachment based on the connection bandwidth. c) The factor used to calculate the CO price is provided for each gateway. The The "FUNDING FACTOR" (FF) determines the funds that will be available to the infrastructure pool. d) The CO COMBit percentage -- the ratio of commercial traffic that traverses the gateway -- is projected. In the example, COMBit percentage is labeled "CO COMBit PERCENTAGE". e) The maximum infrastructure pool amount is the difference between the total BASE PRICE and the total CO PRICE. The CO COMBit percentage is multiplied by this difference to obtain the actual contribution, which is added to the ATTACHMENT PRICE and yields the TOTAL INVOICE amount. f) In Example B, NSF provides funding directly to the backbone provider. The Research & Education attachments are provided at no additional cost. This also shows that the CO portion of the gateway are charged for attachment and for projected COMBit measurement. Thus, commercial use of the RE gateway funded by NSF grants is compensated and, in addition, funds are made available to the infrastructure pool. 8 ANS Direct Connected Enterprise Attachments A similiar methodology presented in this document for Gateway attachments is used for individual direct commercial "CO" enterprise attachments that are attached directly to the ANS backbone. ANS will fund the difference between the CO and the RE price of its attachment fee to the infrastructure fund in a manner similar to that described previously. Local Exchange Carrier Circuits In order to interconnect between the Gateway which resides on the customer premise, and the nearest ANS "point of presence," local exchange carrier facilities (local loops) are required. These costs are added to the price of the ANS/ANS CO+RE gateway service attachment or commercial attachments and represent a pass through of the local exchange carrier's charges. For NSF grant supported sites, these costs are included in the NSF funding. Summary Deliberations among ANS, NSF and participating midlevel network representatives has resulted in an ANS plan for commercial services. The plan includes: 1. A new ANS acceptable use policy. 2. An ANS Internet connectivity agreement among multiple network service providers that carry commercial traffic. 3. ANS/ANS CO+RE attachment agreements with network service providers and direct ANS/ANS CO+RE-attached institutions. This plan includes a mechanism for funding enhancements to the network infrastructure that will benefit the Research and Education community and all users of the network, and it provides a framework for financial settlements between network service providers that interconnect. It is recognized that this plan is a starting point for what will be an evolving set of services and relationships. The ideas embodied in this plan undoubtedly will continue to develop as we gain more experience. 9 Attachment A The COMBit Mechanism A mechanism is needed to identify and measure commercial traffic sent across the ANS backbone by networks with mixed RE and CO traffic. This attachment describes the COMBit mechanism, which is intended to address both TCP/IP and GOSIP/ANSI compliant OSI technologies. Identification & Measurement of COMBits A COMBit is defined as a unit of measure that corresponds to network usage by a direct backbone-attached midlevel network. COMBits are used to calculate the fraction of commercial (CO) usage versus research and education (RE) usage. The COMBit fraction is used to derive the fee for a direct-attached network. In order to measure COMBits, it is necessary to distinguish between institutions that agree to an acceptable use policy for research and education purposes or for commercial purposes. A COMBit quantifies both types of useage. Fees assessed for network transport services are based on COMBit percentages classified as commercial or research/education. In order to identify unique institutions supporting IP protocols that may use the network for commercial purposes, it is necessary that they possess a unique class A, B, or C IP network number. With regard to subnetting, if one subnet of a network is CO, then the entire network is treated as CO. All subnets associated with a designated RE network will be treated as RE. The RE or CO identification of the networks are supplied by the midlevel. In order to identify unique institutions supporting OSI protocols that may use the network for commercial purposes, it is necessary that all unique RE or CO networks have NSAP addresses assigned from either the GOSIP or ANSI spaces, and those networks must be uniquely identifiable by the ORG ID (AAI) and the Domain ID. A-1 COMBit Measurement Objectives The mechanism for the measurement of COMBits has been developed with the following objectives in mind. 1. Minimize the overhead cost of the measurement mechanism. It should not require significant additional hardware, software or labor-intensive processing of information. Network switching capacity should not be adversely affected. 2. Ensure that the COMBit measurement is mutually verifiable. 3. Implement a measurement that encourages the use of applications that make efficient use of the network (e.g. using large packet sizes). COMBit Measurement Statistical Objects for IP In the NSFNET T1 backbone, statistics are collected using the NNStat package on a dedicated IBM RT workstation listening passively on the internal token ring interface. The ANS T3 backbone architecture does not lend itself to a similiar implementation for statistics collection. The ANS T3 backbone architecture consists of a core nodal switching system (CNSS) and an exterior nodal switching system (ENSS). A CNSS interconnects other backbone CNSS routers via multiple T3 circuits. The CNSS also interconnects one or more ENSS routers which typically are situated at customer locations. The ENSS is the demarcation point between a customer Ethernet or FDDI network and the ANS infrastructure. Traffic statistics are collected on transit load at the interface between the ENSS and the customer network. Both packet counts (in and out) and byte counts (in and out) are collected for the following data objects: 1. Network#-to-Network# Matrix 2. IP Protocol Distribution 3. TCP/UDP Port Distribution Statistics are currently only collected in the ENSS routers based upon traffic that traverses the interface between the ENSS and the customer's local area network. A-2 COMBit Measurement Statistical Objects for OSI The OSI protocol suite has features that make packet and byte accounting somewhat more difficult than for TCP/IP. OSI network addresses (Network Service Access Point, or NSAP, addresses) differ from IP addresses in a number of ways. The most obvious is that they are longer (up to 20 octets, as compared to four octets). Another characteristic of OSI that makes accounting more difficult is that very little of the address structure is standardized, and address allocation is distributed. The distributed allocation of address structure is a problem because, in general, it is impossible to look at an NSAP address and determine what part of it is associated with an attaching institution without considerable a priori knowledge. For instance, the US DCC (Data Country Code) address space in the United States is allocated by ANSI and includes a three-octet Organization ID. However, in other countries, the DCC space may not be allocated in the same way. Furthermore, even in the US DCC space, the OID alone might not be enough to isolate the attaching institution, since the OID might simply refer to a midlevel network. The midlevel then subdivides the address space, where it can, in theory, allocate the rest of the space as it pleases. However, in the US, a standard structure for the entire NSAP address has been described both by US GOSIP and by ANSI. Both organizations suggest the following format (the number in parentheses indicates the octet count): FORMAT PREFIX(3), DFI(1), ORG(3), UNUSED(2), DOMAIN(2), DOMAIN-SPECIFIC(9) For GOSIP, the format prefix is 470005; for the US DCC it is 39840F. The DFI is effectively a version number (80 for GOSIP, 01 for ANSI). The organization ID is called an AAI (Administrative Authority Identifier) in GOSIP; this field refers to the organization to which allocation of the rest of the address has been delegated. The domain field refers to the routing domain within the organization; the domain-specific field is further specified, but is irrelevant for this discussion. Given this format, the combination of the ORG ID and the domain will uniquely identify a Routing Domain, which can represent an entity with roughly the same granularity as an attached RE or CO institution with a unique IP Network Number. This is the appropriate granularity for statistics collection (note that the org/domain is qualified by the format prefix and DFI). Thus, statistics could be gathered based upon six octets A-3 (the five octets of ORG ID and DOMAIN, plus an octet to represent the leading portion of the address). Therefore, this method can be used to gather statistics on unique attached RE or CO institutions, however, all unique RE or CO networks must have NSAP addresses assigned from either the GOSIP or ANSI spaces, and those networks must be uniquely identifiable by the ORG ID (AAI) the DOMAIN ID. It should be noted that this does not constrain the address allocation authorities to impose the GOSIP/ANSI address structure; for instance, the institution could assign three-octet domain identifiers subject to the constraint that RE or CO networks have the first two octets be unique. In practice, it is unlikely that very many organizations in the US will stray from the GOSIP/ANSI addressing specification. This scheme works efficiently in the US because of the common address structure. However, internationally, this may be more difficult, since the address structure is not necessarily going to be the same. The collection of forwarding-time statistics in the main forwarding path will become increasingly difficult. It may be possible to customize backbone ENSS collection points in non-US locations so that they gather statistics efficiently, given the local addressing scheme. However, a similar type of constraint needs to be imposed. It also should be noted that, in general, it will be difficult to identify both the source and destination of a particular packet, since only one of the addresses may be structured according to the US scheme, when the other end may be international. In the future, as OSI gains in popularity, it might be necessary to modify the COMBit methodology to support a category of traffic for which one of the source or destination addresses is effectively unknown. Identification of OSI Application Types Identifying the type of application running over a particular OSI CLNP data stream is made difficult for two reasons. One is that the application is addressed through identifiers at multiple layers of the protocol stack (thus requiring that the statistics gatherer be able to pick these apart). The other is that the identifiers are not standardized; they are obtained on a per-connection basis via directory lookup. As such, it is not practical to gather application identification information for OSI traffic. The Measurement Methodology A-4 The measurement methodology described herein represents both the current and proposed implementations for measurement of statistical objects by ANS. It is recognized that the statistical integrity and granularity of these measurements can, and will be improved upon over time. Packets received from backbone-attached networks are sampled at a fixed rate. This rate is currently set to one out of every fifty packets (1/50). Each sampled packet is analyzed, and the appropriate statistical object counters are incremented. Every fifteen minutes, the traffic statistics are relayed to a central collection machine and the statistical object counters are cleared. If the central collection machine fails or cannot reach the ENSS router, the ENSS router will cache several hours worth of fifteen-minute data samples and deliver them all at once when connectivity is re-established. This differs from NNStat on the RT based T1 NSFNET, which effectively increases its collection period and delivers one-large set of objects instead. The centralized collector machine writes out a set of disk files containing the statistics information. The values collected are multiplied by the sampling rate, thereby extrapolating the actual amount of real traffic. This data can be used for the calculation of COMBit fractions, and also can be used for network engineering purposes. The raw data is maintained in a secure repository, and is held in confidence to ensure privacy among ANS customers and the Internet-user community. The raw data is reduced on a daily basis. COMBit Definition and Calculation Figure 1 illustrates two midlevel networks interconnected via the ANS backbone. Each midlevel hosts multiple institutional networks that are categorized as either research and/or education (R11, R21), or commercial (C11, C21). As previously described, the institutional networks are identified and classified via a unique IP network number provided by the midlevel. For each direct backbone-attached midlevel (or institutional network), the raw ENSS collected data is reduced in order to form a [network]x[network] matrix. This reduction is accomplished through the accumulation of packets (in and out) and bytes (in and out) across all autonomous system numbers or unique network numbers, which are reachable through the direct ANS backbone- attached midlevel network. The following table (Table 1) can thereby be constructed as the raw statistical objects summarized for Midlevel network #1. A-5 Midlevel #1 - M1 R11 C11 R21 Packets Packets Bytes Bytes Packets Packets Bytes Bytes In Out In Out In Out In Out C21 Packets Packets Bytes Bytes Packets Packets Bytes Bytes In Out In Out In Out In Out Table 1 - Raw Statistical Objects for Midlevel #1 - M1 COMBits are hereby defined as a linear combination of packets (in and out) and bytes (in and out) and can be expressed as follows : COMBits = [Packets_In + Packets_Out]*300 + Bytes_In + Bytes_Out The linear combination of packets and bytes is weighted to favor larger packets in order to encourage efficient network usage. Users are encouraged to develop applications that use large packet sizes in order to reduce the number of CO or RE COMBits attributed to their Midlevel network. As an example of this, the following table illustrates the COMBit rate generated for a full T1 (1.536 Mbps) attachment at varying average packet sizes: COMBit Rate for Full T1 Average Packet Size COMBits Per Second Rate 100 Bytes 800K COMBits/sec. 200 Bytes 500K COMBits/sec. 800 Bytes 275K COMBits/sec. COMBits also are insensitive to the direction of traffic flow. In other words, the same number of COMBits are attributed to both M1 and M2 for traffic flow between institutions R11 and R21, regardless of which direction the traffic flows. This is necessary in order to ensure conservation of total network usage in the COMBit calculation. This also recognizes that users should not be penalized for being either senders or receivers of information, since the sender or receiver may both initiate and benefit from the transmission. A-6 Using the COMBit definition, table 1 can now be further reduced into a simplified matrix which represents the COMBits attributed to Midlevel #1 due to transmissions between the unique identified network numbers within that Midlevel. This is illustrated in Table 2. Midlevel #1 - M1 R11 C11 R21 COMBits[R11, R21] COMBits[C11, R21] C21 COMBits[R11, C21] COMBits[C11, C21] Table 2 - COMBits Attributed to Midlevel #1 Calculation of COMBit Fractions The simple example illustrated in figure 1 results in COMBits which are either research and/or education (RE), or commercial (CO) that are attributed to both midlevels M1, and M2. The rules for attributing RE versus CO COMBits to any particular midlevel are described as follows: 1. Any COMBits involving the transmission of data between two RE networks across the ANS backbone are counted as RE COMBits and attributed to both of the participating midlevel networks. 2. Any COMBits involving the transmission of data between two CO networks across the ANS backbone are counted as CO COMBits and attributed to both of the participating midlevel networks. 3. Any COMBits involving the transmission of data between a CO network/institution in one midlevel (e.g. M1), and an RE network/institution in the other midlevel are counted as CO COMBits attributed to the midlevel which hosts the CO network, and RE COMBits attributed to the midlevel which hosts the RE network. A-7 Rule #3 ensures that ANS-attached midlevel networks which host an RE institution involved in communication with a CO network in another ANS-reachable midlevel will not be penalized by an increased CO COMBit percentage associated with that communication. In other words, a midlevel which hosts RE institutions only will never be attributed any CO COMBits and will pay the 100% RE fee regardless of whom they communicate with on the ANS network. Table 3 illustrates these rules as they would apply to M1, and M2 in the example illustrated in Figure 1. Midlevel #1 - M1 R11 C11 R21 RE [R11,R21] COMBits CO [C11,R21] COMBits C21 RE [R11,C21] COMBits CO [C11,C21] COMBits Midlevel #2 - M2 R11 C11 R21 RE [R11,R21] COMBits RE [C11,R21] COMBits C21 CO [R11,C21] COMBits CO [C11,C21] COMBits Table 3 - RE/CO COMBits attributed to M1, M2 in Figure 1 Having established these ground rules, it is now possible to express the COMBit fractions for each midlevel (M1, M2) represented in Figure 1 in closed form: COMBits M1 = RE1 [R11,R21 + R11,C21] + CO1 [C11,R21 + C11,C21] COMBits M2 = RE2 [R21,R11 + R21,C11] + CO2 [C21,R11 + C21,C11] A COMBit fraction now can be written for each midlevel (M1, M2). This fraction is used to calculate the fee charged to each midlevel based upon the percentages of RE and CO COMBits, respectively. These percentages can be written as follows : A-8 CO1 M1 CO Percentage = ========= [RE1 + CO1] CO2 M2 CO Percentage = ========= [RE2 + CO2] A-9 Example Calculation of COMBit Fractions In order to illustrate this example, some COMBit values are chosen to correspond to the flow of traffic between M1, and M2 across the ANS backbone. R11 C11 R21 200 COMBits 50 COMBits C21 100 COMBits 50 COMBits The COMBit percentage for M1 can be derived as follows: COMBits M1 = RE1 [R11,R21 + R11,C21] + CO1 [C11,R21 + C11,C21] = RE1 [200 + 100] + CO1 [50 + 50] = RE1 300 COMBits + CO1 100 COMBits CO1 100 M1 CO Percentage = ========= = ======== = 25.0% [RE1 + CO1] 300 + 100 RE1 300 M1 RE Percentage = ========= = ======== = 75.0% [RE1 + CO1] 300 + 100 Likewise, the COMBit percentage for midlevel M2 can be derived as follows : COMBits M2 = RE2 [R21,R11 + R21,C11] + CO2 [C21,R11 + C21,C11] = RE2 [200 + 50] + CO2 [100 + 50] = RE2 250 COMBits + CO2 150 COMBits CO2 150 M2 CO Percentage = ========= = ======== = 37.5% [RE2 + CO2] 250 + 150 RE2 250 M2 RE Percentage = ========= = ======== = 62.5% [RE2 + CO2] 250 + 150 A-10 Closed Form Representation of COMBit Calculation The numeric example derived from Figure 1 has been used to illustrate the COMBit calculation. Figure 2 represents the actual model required for implementation of the COMBit calculation on the ANS network. This model corresponds to the higher order case of the earlier example, where ANS interconnects with many direct-attached midlevel networks (M1...Mk), each midlevel network interconnects with many research and/or education networks (Rk1....Rkn), and many commercial networks (Ck1...Ckn). Figure 2 illustrates this general case and summarizes the derivation of a closed form expression for COMBit calculation for each unique direct-attached midlevel network. Conclusions A mechanism is provided to assist in the transition of the ANS network to service which can support a mix of commercial and research and/or education traffic. The services are distinguished by the two different acceptable use policies that govern them. In order to implement this policy, a mechanism is defined to identify commercial networks by unique IP network number, measure their communication across the ANS network and administer cost recovery in support of commercial service. A COMBit is defined as a unit of measure that corresponds to network usage by a direct backbone-attached midlevel, or are institution that is reachable via the backbone. COMBits are used to calculate the percentage of commercial (CO) usage versus research and education (RE) usage. The measurement of COMBits in the ANS network is implemented as a statistical sample in the ENSS router, which typically resides at the interface between the ANS network and the direct- attached midlevel or institutional network. The statistical sampling technique minimizes overhead cost. The cost of the COMBit measurement is small considering that it represents the the reduction of data which is normally collected in any case (e.g. packets and bytes). The COMBit measurement is verifiable by any direct ANS- attached midlevel or institutional network. An NNStat data collection machine which listens on the ENSS attached local area network at any midlevel or institutional is capable of reproducing the same network x network matrix that is used to derive COMBit percentage for the specific midlevel. ANS will cooperate in the establishment of verification data collection systems, where desired. A-11 Attachment B EXAMPLE A - INDEPENDENT - RE and CO GATEWAY BANDWIDTH ATTACHMENT : T1 ( $ IN THOUSANDS ) TYPE OF TYPE UNIT BASE INSTITUTIONS CODE QTY PRICE PRICE ---------------------------------------------------- BASE ATTACH 00-56 $55,000 $55,000 RESEARCH ORG 10-99 3 200 600 RESEARCH ORG 10-T1 1 4,000 4,000 RESEARCH ORG 10-56 2 2,000 4,000 GOVT ORG 15-56 1 2,000 2,000 COMM ORG 15-99 1 200 200 SERV PROVIDER 50-T1 1 4,000 4,000 SERV PROVIDER 50-56 2 2,000 4,000 ------------------------------ ATTACHMENT PRICE $73,800 FUNDING FACTOR 33% MAXIMUM INFRASTRUCTURE FUNDS $ 24,354 CO COMBit PERCENTAGE 20% FUND CONTRIBUTION $ 4,871 --------- TOTAL INVOICE $78,671 ========== B-1 EXAMPLE B - NSF FUNDED - RE and CO GATEWAY BANDWIDTH ATTACHMENT : T1 ( $ IN THOUSANDS ) TYPE OF TYPE UNIT BASE INSTITUTIONS CODE QTY PRICE PRICE ---------------------------------------------------- BASE ATTACH 00-00 $55,000 $55,000 RESEARCH ORG 10-99 3 200 600 RESEARCH ORG 10-T1 1 4,000 4,000 RESEARCH ORG 10-56 2 2,000 4,000 -------------------------- NSF-FUNDED SUB-TOTAL 63,600 GOVT ORG 15-56 1 2,000 2,000 COMM ORG 15-99 1 200 200 SERV PROVIDER 50-T1 1 4,000 4,000 SERV PROVIDER 50-56 2 2,000 4,000 ------------------------------ ATTACHMENT PRICE 73,800 FUNDING FACTOR 33% MAXIMUM INFRASTRUCTURE FUNDS $ 24,354 CO COMBit PERCENTAGE 20% FUND CONTRIBUTION $ 4,871 --------- TOTAL INVOICE $78,671 NSF CREDIT $63,600 --------- AMOUNT DUE $15,071 ========== B-2 ATTACHMENT C DEFINITIONS OF INSTITUTIONS Education: A public or private organization that is accredited by a federal, state or provincial agency to grant degrees or diplomas for educational studies (including K-12). Research Organization: A for-profit or not-for-profit organization, including a government agency, that engages in research and does not qualify as an educational organization. Commercial Service Provider: A for-profit or not-for-profit organization that uses the network for the delivery of fee-based services or products to any customers. Commercial Organization: A for-profit or not-for-profit organization that does not qualify as an education, research or commercial provider. C-1 SAMPLE CO+RE GATEWAY SERVICE ATTACHMENT MATRIX FOR ILLUSTRATIVE PURPOSES ONLY BASE ATTACHMENT FEE: T1: $55,000 ADDITIONAL ATTACHMENT FEES: ATTACHMENT BANDWIDTH <56 kb 56 kb T1 ------------------------------- -99 -56 -T1 Education: 01 Educational Institution N/C N/C N/C Research and or Commercial: 10 Research Organization $200 $2,000 $4,000 15 Commercial Organization $200 $2,000 $4,000 Commercial Service Provider: 50 Service Supplier $200 $2,000 $4,000 C-2 Acknowledgements ANS would like to acknowledge the invaluable contributions, advice and counsel that were provided by representatives from NSF, FARNET and the many midlevel networks and service providers during the development of this plan.