In my previous blog - An Network Designer's Thought Process Part 1, I presented a simplified design example and my preliminary analysis. Now I’m going to explain how I would approach it, and my final design solution.
Develop network designs
What do I consider when developing network design options? I believe that OSPF area design could affect OSPF stability, scalability, convergence, optimal routing, interaction with other features, and ease of operation, maintenance, and troubleshooting. What design options do I have?
- Option 1 – a single OSPF area in the whole OSPF domain
- Option 2 – multiple areas in the OSPF domain: If this option is chosen, the next step is to design the areas – define the standard area modules, define area boundaries, identify the ABRs, consider stub area, NSSA etc. I’ll skip the detail multi-area design in this article.
Single area pros
- Optimal OSPF paths are always chosen because all routers have the full topology and routing info of the whole OSPF domain.
- Optimal routing likely minimizes traffic latency.
- Single area configuration is simpler than multi-area.
- Single area likely has less of a chance to encounter software defects and human error.
- It’s easier to troubleshoot.
- It’s simpler to deploy MPLS TE and FRR in a single area.
Single area cons
- It requires more powerful router CPU and more router memory because a router will have to process more LSAs, bigger OSPF topology, and more routing info. This is not a big problem for the modern ISP routing platforms.
- It could also slow down OSPF convergence, although the iSPF feature can help to improve the OSPF convergence.
- It slows down routing RIB/FIB table updates due to bigger table sizes.
- The number of OSPF routes is insignificant compared to a much larger BGP table in PE routers
The multi-area’s pros and cons are generally the opposite of a single area. With proper route summarization design on the ABRs, optimal routing between PE routers is achievable.
Compare the design options
Can the design options meet all of the requirements? After analyzing each option, I believe both options would meet all of the requirements. Here comes the difficult part (or some people say the fun part) of the network designer’s job – choose the best design option from a few feasible options.
|Single Area OSPF||Multi-area OSPF|
|99.999% of network availability||Achievable with proper resiliency design with 200 number routers||Achievable with proper resiliency design with more than 200 routers|
|Minimize traffic latency|
Achievable with proper bandwidth capacity planning
Always uses the optimal traffic path
|Achievable with proper bandwidth capacity planning and proper route summarization design|
|Traffic convergence in less than 100 milliseconds||The PE-to-PE convergence relies on MPLS FRR and/or IP FRR||The PE-to-PE convergence relies on MPLS FRR and/or IP FRR|
|Financial investment||The routers have already been purchased so the design doesn’t change the financial investment||The routers have already been purchased so the design doesn’t change the financial investment|
|Supports network growth|
I have seen a few OSPF backbone areas in ISP networks with more than 200 routers and they run just fine. OSPF is not the scale-gating factor (considering a bigger picture including BGP and all other processes)
The effect of the single area design on the growth is insignificant
|Can support an OSPF domain much bigger than 200 routers|
|Simple to implement||It’s much simpler to implement MPLS TE/FRR in the single area design, and with less limitations|
Inter-area MPLS TE is more complex to implement
It has more limitations
|Simple to operate, maintain, and troubleshoot||Very simple||More complex|
|Proven in production networks|
Yes, some ISPs use single area design, but it’s much more popular to use a single level design in IS-IS networks
|Yes, most of the large Tier-1 and Tier-2 ISPs use multi-area design due to the large network scale|
Select the most suitable design option
Based on the above comparison, I believe the single area design option is more attractive due to its simplicity. The simpler MPLS TE implementation is the tiebreaker to me.
To summarize, a network designer will need to analyze the given info and requirements, identify any missing information, obtain the missing info by asking the right questions, develop a small number of design options that meet all the requirements, compare the options, make the right trade-offs and choose the most suitable option to achieve the design goals.
About the Author
Cary Chen is CCDE #20130038 and double CCIE #14263 (SP and SP Operations), and a manager in Cisco Advanced Services. Cary supported ISP customers, and developed ISP routing platforms in the last 15 years. He enjoys working with ISP customers to design next generation networks.
Here are a few additional ways for us to engage and keep the conversation going:
- Cisco Learning Network CCDE Study Group
- Connect on Twitter too
- CCDE study materials for the Written and Practical exams
- Related Unleashing CCDE blogs: A Network Designer’s Thought Process Part 1 by Cary Chen, Commercial Solutions for Classified (CSfC) with Joe Galimi, Network Function Virtualization in Enterprise by Stephen Lynn, All Smoke and Mirrors by Michael Kowal – Part 1, IWAN Part 1: PfRv3 Design Considerations by Dmytro Muzychko