Defining Network Capacity draft-cui-ippm-rfc5136bis-00

1. Defining Network Capacity draft-cui-ippm-rfc5136bis-00 Xiangsong Cui, Xiangsong.Cui@huawei.com

2. Acknowledgements • The author would like to acknowledge Phil Chimento and Joseph Ishac for their great contribution on the item of network capacity. • The author would also like to acknowledge Mark Allman, Patrik Arlos, Matt Mathis, Al Morton, Stanislav Shalunov, and Matt Zekauskas for their contribution to [RFC5136]. Network Capacity

3. Component Definition in RFCs • Earlier documents, such as RFC2460 and RFC2330, define Node, Link, and Path. • RFC5136 defines these terms slightly differently. This difference could impact the definitions of metrics. It might be ambiguous just what is getting measured. • We should reference the terminology found in the earlier documents rather than providing our own, slightly different, definitions. Network Capacity

4. IP-type-P Link Capacity Issue Nominal Physical Link Capacity, NomCap(L), is the theoreticalmaximum amount of data that the link L can support. -- RFC5136, section 2.2 We define the IP-layer link capacity, C(L,T,I), to be the maximum number of IP-layer bits that can be transmitted from the source S and correctly received by the destination D over the link L during the interval [T, T+I], divided by I. -- RFC5136, section 2.3.2 An important distinction between usage and capacity is that Used(L,T,I) is not the maximum number, but rather, the actual number of IP bits sent that are correctly received. -- RFC5136, section 2.3.4 The definitions seem clear enough, but how to understand the ”maximum”? Network Capacity

5. IP-type-P Link Capacity Example H1 H2 Link1 Node H1 and H2 are connected by link1. Link1 is active. • Does the ”IP-layer … maximum … can …” mean that, in physical layer, the nominal physical link capacity of the link is fully utilized? • In case 1, link doesn’t transmit packet not means it can not do that. • We have no actual link capacity number, but the metric is still there. • Is the IP-layer link capacity also a theoretical number? Network Capacity

6. IP-type-P Link Capacity Proposal In theory, the IP-type-P link capacity may be calculated by: C(L,T,I) = [Lp /(Lh + Lp + Lt)] * [1 - BER(T, T+I)] * [1 - BDR(T, T+I)] * P(L) - Lp denotes type P packet length (in IP layer), - Lh denotes link layer protocol overhead length, - Lt denotes link layer protocol overtail length, - BER(T, T+I) denotes Block Error or Lost Rate during the interval [T, T+I], - BDR(T, T+I) denotes Block Duplication Rate during the interval [T, T+I], - P(L) denotes Nominal Physical Link Capacity of the given link. In practice, we can define the IP-layer link capacity, C(L,T,I), to be the maximum number of IP-layer bits (i.e. the Nominal Physical Link Capacity of the link is 100% utilized) that can be transmitted from the source S and correctly received by the destination D over the link L during the interval [T, T+I], divided by I. Network Capacity

7. IP-type-P Router Capacity and Path Capacity • Is it necessary to consider router capacity when we talk about network capacity? • Is it appropriate to implicitly include (or hide) router capacity inside link capacity when we talk about path capacity? • In the scenario where router is fully utilized (e.g. forwarding capacity) while the link is not fully utilized (e.g., physical link capacity), does the number represent our expected IP-layer LINK CAPACITY? • Should we hide router capacity inside link capacity, then what is the relation between router/link/path? Network Capacity

8. IP-type-P Router Capacity Proposal Proposal on definition of the IP-type-P router capacity: C(R,T,I), the maximum number of IP-layer bits (in the formation of type P packet) that can be correctly transferred from the ingress interfaces to the egress interfaces during the interval [T, T+I], divided by I. In practice, the IP-type-P router capacity representation is affected by many implementation factors. Network Capacity

9. IP-type-P Path Capacity Issue Definition in RFC5136, section 2.3.3, Using our definition for IP-layer link capacity, we can then extend this notion to an entire path, such that the IP-layer path capacity simply becomes that of the link with the smallest capacity along that path. C(P,T,I) = min {1..n} {C(Ln,T,I)} Does this clearly show out the IP-type-P path capacity concept? Network Capacity

10. IP-type-P Path Capacity Example 1 • Case 1: links and routers are exclusive for the corresponding path H1 R1 R2 H2 Link1 Link2 Link3 The IP-layer capacity of an exclusive path may be calculated by: C(P,T,I) = min {1..n} {C(Ln,T,I), C(Rn,T,I)} Network Capacity

11. IP-type-P Path Capacity Example 2 • Case 2: links and routers are shared (fully or partly) by paths H1 H2 Link1 Link3 R1 R2 Link2 Link4 Link6 H3 H4 Bit flow from H1 to H2 Bit flow from H3 to H4 Link2 capacity The information transmitted across the link can be generated by any source, including those sources that may not be directly attached to either side of the link. -- RFC5136, section 2.3.4 Network Capacity

12. IP-type-P Path Capacity Example 2 (cont.) • Assume all link capacity in case 2 is 1G bps and all router capacity is 2G bps, • Capacity of path1 <H1, L1, R1, L2, R2, L3, H2>, is 1G bps? • Capacity of path2 < H3, L4, R1, L2, R2, L6, H4>, is 1G bps? • In idle situation, available capacity of path1 and path2 are both 1G bps? • But in fact, we can not get two simultaneous 1G bps path, right? • This means, when we say path1 capacity is 1G bps, then the capacity of link2 is occupied by path1, and available link2 capacity for path 2 is zero … • The IP-layer capacity of the corresponding path depends on not only the IP-layer capacity of the links and the routers but also the "competitive" traffic of other paths which have overlap segment with the corresponding path. • This is to say, the IP-layer capacity of the non-exclusive path is a variable. Network Capacity

13. Next Step • Are the issues trivial or important? • Do we need a new draft or RFC? Network Capacity

14. Questions/Comments and Discussion Thank you for your attention! Network Capacity