Sources of Inefficiencies with FLSM

Fixed-length subnet masking was a tremendous step in the evolution of the IP address architecture. It offered the capability to develop locally significant subnetworks without affecting global routability and gave network administrators the flexibility to create whichever sized subnet suited their needs best. Despite this radical leap forward, the potential for waste remained high. Thus was born the great irony of subnetting. Subnetting, in general, was designed to enable more efficient use of address space by permitting class-based network address blocks to be subdivided into smaller address blocks. Yet the way subnetting was originally implemented was far from efficient.

Sources of waste and inefficiency included the following:

• Subnet 0 and subnet "all 1s" were reserved.

• All-0s and all-1s host addresses were reserved.

• There was one size of mask for all subnets.

The manner in which each of these contributed to inefficiency and waste is further described in the following sections.

Subnet Addresses

I've probably beaten this one to death throughout this chapter, but it remains an important point. The original rules for subnetting in RFC 950 stressed the significance of reserving the subnets whose addresses were constructed of all 0s and all 1s. Depending on the size of the mask being used, this could translate directly into a substantial number of addresses being wasted. Consider a mask of 255.255.255.192 in a Class C network. That size of network has only a mathematically possible 256 host addresses. The subnet mask we've selected would yield four possible subnets, each with 64 mathematically possible host addresses. Reserving subnets 0 and 3 (as per Table 3-4) immediately translates into the loss of 128 of the host addresses. Pragmatically, wasting 128 addresses is better than wasting 384, as you would do if you acquired just two Class C networks and didn't bother subnetting. However, the fact remains that the classical rules are not perfectly efficient.

Host Addresses

A much more subtle form of loss is found in the host addresses themselves. Again, the all-0s and all-1s addresses are reserved, but this time there's a pair within each subnet. Thus, for each subnet formed, two host addresses can't be used. To continue expanding on the example from the preceding section, the all-0s and all-1s addresses in subnets 1 and 2 are as shown in Table 3-13.

Table 3-13. Subnetting with a 6-Bit Mask in a 16-Bit Network

Subnet Number	Binary Value of the Last Octet	All 0s
Subnet 1	01-000000	192.169.125.64
Subnet 1	01-111111	192.169.125.127
Subnet 2	10-000000	192.169.125.128
Subnet 2	10-111111	192.169.125.191

Table 3-13 demonstrates that, in addition to losing all of Subnets 0 and 3 (the all-0s and all-1s subnet addresses), you lose four more addresses. That brings the total of unusable addresses to 132 out of the original 255a more than 50% loss rate.

Sizing the Mask

As you learned earlier in this chapter, there can be only a single mask for each network being subnetted with FLSM. Thus, an additional source of inefficiency lies in the fact that the appropriate size of mask for the largest of the local networks is probably too big for the others. The numbers of addresses wasted in this manner are incalculable, because they vary widely from implementation to implementation. Suffice it to say that the waste is in direct proportion to the size difference between the smallest and largest LANs being serviced with a subnetted network address block.

Given this potential for waste, network administrators are motivated to size their subnets as carefully as possible. A contravening point to consider is that it is often difficult to accurately project growth over time. Thus, a network administrator might be tempted to build in a substantial amount of fluff to accommodate this unspecified future growth requirement. Despite not having good data to work with, a network administrator is the one who would suffer the consequences of an overly conservative subnetting scheme. Users, after all, tend to care more about their own convenience than they do about the global availability of IP addresses. Thus, sizing a network mask often is influenced more by fear than by actual requirements.

Although this might seem trivial, reflect on how radically the base (or first), initial, and broadcast addresses per subnet change with mask size. These were presented in a series of tables in this chapter. Now consider how difficult it would be to change the mask size in an established FLSM network. The vast majority of your endpoints would have to be renumbered. Although tools such as Dynamic Host Configuration Protocol (DHCP) might reduce the burden of renumbering, it still would be a logistical nightmare to plan, and implementing it would cause some downtime. Besides, even the most sophisticated network likely wouldn't be completely supported by DHCP or some other automated tool. Quite simply, many hosts can't afford to have a dynamic address. Servers and network nodes, for example, all but require a static address to function reliably from the perspective of the user community.