N-tier challenges and tiered ISP backbone

Challenges of N-tier client/server architecture

N-tier architecture splits an application into separate layers (presentation, application/business logic, data, sometimes more), each running on its own machine or group of machines. The challenges are:

1. Complexity: more tiers, servers and interfaces to design, configure, test and document.
2. Performance and latency: each request passes through several tiers and network hops, so response time increases.
3. Higher cost: more hardware, software licences, and skilled staff are needed.
4. Maintenance and deployment: updates must be coordinated across tiers; version mismatch between tiers causes failures.
5. Security: more network entry points and tier-to-tier connections must be protected (firewalls, authentication, encryption).
6. Reliability: a failure in any tier can stop the whole service, so redundancy and fail-over are needed.
7. Scalability and load balancing: tiers must be scaled separately and need load balancers; session state must be shared.
8. Troubleshooting: finding the faulty tier from logs spread over many servers is difficult.
9. Data consistency: transactions across tiers and databases need careful design.

Architecture of the Internet backbone: tiered ISP network

The Internet is a "network of networks" in which ISPs are organised in tiers, and the backbone is the top tier.

   +------------------ Backbone -------------------+
   |  Tier-1 A <==peer==> Tier-1 B <==> Tier-1 C   |
   +----+-----------------+----------------+-------+
        |                 |                |   transit
     Tier-2 X <--IXP--> Tier-2 Y       Tier-2 Z
      |    |              |               |
   Tier-3 Tier-3       Tier-3          Tier-3
      |      |            |               |
    homes  offices     campus         cyber cafes

• Tier-1 (core/backbone): global networks with fibre and submarine cables. They peer with each other without payment and do not buy transit. Core routers run BGP and carry the traffic at 100 Gbps and above.
• Tier-2 (regional): buy transit from Tier-1 and also peer with other Tier-2s at IXPs. They sell to Tier-3 and large businesses.
• Tier-3 (access): serve end users through DSL, cable, fibre, wireless. They pay upstream for Internet access.
• PoPs and IXPs: each ISP has Points of Presence where customers connect; IXPs let ISPs exchange traffic locally, which lowers cost and delay.
• Routing: each ISP is an Autonomous System; BGP exchanges routes between them, while IGPs (OSPF, IS-IS) are used inside.
• Redundancy: ISPs multihome to several upstream providers so that a failure of one link does not cut service.

Nepali ISPs are mostly Tier-3 networks, buying transit from providers abroad (mainly through India and also China) and exchanging local traffic at the Nepal Internet Exchange (NPIX).