Under what conditions will distributing a file by a client-server architecture be faster than distributing the file by a peer to peer architecture.
Try an answer before revealing the guidance below.
Key Concepts
- D_cs = max(NF/u_s, F/d_min)
- D_p2p lower bound
- Peer upload capacity
- Number of peers
Answer Approach
- Write both distribution-time formulas side by side.
- Compare each term of the P2P bound with the client-server bound.
- Consider whether the models assume ideal peers, and what real-world factors they ignore.
Full Answer
Answer status: Draft answer (unofficial). Revision notes, not an official marking scheme.
Using the textbook (K&R) distribution-time formulas: in the ideal model client-server is never faster. F/us ≤ NF/us and NF/(us + Σui) ≤ NF/us, so D_p2p ≤ D_cs; the two are equal only when N = 1, when peers contribute no upload capacity (Σui = 0), or when the slowest download F/dmin dominates both. So client-server can only be faster under conditions the model ignores: very few clients (so P2P peer discovery and coordination overhead outweighs the gain), peers with little or no usable upload capacity (asymmetric access links, NAT, free-riding), or peers leaving before they have uploaded (churn).
D_cs ≥ max(NF/u_s, F/d_min)D_p2p ≥ max(F/u_s, F/d_min, NF/(u_s + Σu_i))Shortcuts: K concepts · A approach · F answer · R reviewed · B bookmark · ← / → previous / next