NAT many subnets to a single pool or each subnet to own pool?

[wdebraal47]

Consider a scenario that looks like: Ten /24 subnets (Red), Ten /24 subnets (Blue), Ten /24 subnets (Green).  Dynamic auto NAT (PAT) Inside to Outside for Internet.

Option 1:  Create 3 large IP pools Red_Pool, Blue_Pool and Green_Pool then NAT RED_Subnets to Red_Pool, Blue to Blue and Green to Green etc.

Option 2:  NAT each individual subnet to its own Global IP(s) Red_Subnet1 to Red_Subnet1_Pool, Red_Subnet2 to Red_Subnet2_Pool … and so on.

In terms of what is going on under the hood on a fairly high end layer 3 device would either of these options win out significantly in terms of performance?

[Rich Uline]

Wdebraal47,

Off the cuff, I would assume no significant difference in performance of the two options. However, I can think some minor points that make option 2 the poorer decision. One point is that the additional ACLs needed to implement option 2 take up extra memory both in NVRAM, as the configuration file, and TCAM, where they are loaded while the system is running. Another point is that the additional configuration complexity of option 2 imposes an unnecessary burden on the O&M staff. Ideally, a more complex configuration should be implemented only to meet some specific requirement, whereas the two presented options appear functionally identical.

[Georg Pauwen · Accepted solution]

Hello,

in addition to Rich's post, generally, NAT doesn't use up a lot of memory, so even low end routers can handle thouands of translations (' Each NAT translation consumes about 312 bytes in DRAM. As a result, 10,000 translations (more than would generally be handled on a single router) consume about 3 MB').

That said, it all depends on how many public IP addresses you have available. If you have more than one, I would NAT each subnet to a separate pool...

[wdebraal47]

I decided to play with NAT'ing each subnet to it's own IP.  Thank you both for your input.