An Analysis of Microsoft's TCO Comparison - Part 2

References

Microsoft's TCO study: Windows NT Server Compared to Solaris on SPARC
Microsoft's Linux Myths

In their Linux Myths web page, Microsoft claims that the TCO of a Linux system should be comparable to that of a Solaris system since both Linux and Solaris are Unix-based operating systems. There are some fundamental flaws in this argument.

The Solaris systems presented in the comparison ran on more expensive SPARC hardware. Linux would run on hardware comparable to the systems in the comparison that were running NT. This would drastically change the initial purchase price of the Linux system.
The Linux operating system and tools necessary to run it can be had for free. Therefore, the server operating system costs go away.
A good argument could be made for reducing the costs for value-added software on Linux systems since much of it is available for free or for costs substantially lower than the equivalent packages for commercial Unix systems.

I took the numbers from Microsoft's TCO study and entered them into a spreadsheet. I copied the Solaris numbers into a column labeled 'Linux'. In this column, I changed just three numbers.

Table 2, line 1 - Hardware Cost
Since Linux runs on the same hardware as NT, I used the NT value of hardware cost in the Linux column.
Table 2, line 2 - Operating System Cost
Since Linux is available for free, I entered 0. Even if I added in the $80 per server for the high-end Red Hat package, this would only affect the per server TCO by $40 (it's a two year analysis) which turns out to be less than 0.1%.
Table 3, line 1 - Hardware Setup
Since Linux is being specified on identical hardware as NT, it should have the same setup cost. Therefore, I used the NT number here.

Since we don't know what additional software was installed on the NT or Solaris systems, it would be difficult to come up with a cost comparison for equivalent software on Linux. Therefore, we will accept the Solaris costs for value-added software as being valid for Linux systems. An actual study into the costs for many of these items would probably produce values much closer to the NT costs.

We also don't know the nature or necessity of many of the other costs listed so for the sake of argument we will assume that the Solaris costs are valid for Linux.

Using all of the costs associated with Solaris except for the changes described above, the TCO for Linux comes to $44,905 compared to $42,130 for NT.

Comparing Linux to Solaris, we find Linux costing about 20% less to own. Therefore, Linux should be expected to cost considerably less to own than Solaris. Microsoft's claim that Linux and Solaris ownership costs should be similar cannot be substantiated.

Comparing NT to Linux, we find that NT costs less than 7% less per year. That small margin should go away quickly after taking a closer look at the rest of the numbers.

If we decide to repeat this exercise but run Linux on the SPARC, we would use all of the Solaris numbers except for the operating system cost. Even with this configuration, Linux comes in with a TCO 10% less than Solaris. NT would only cost 16% less than this Linux configuration.

But, as presented in part 1, no meaningful conclusions can be drawn from Microsoft's report or these analyses because Microsoft didn't give us meaningful descriptions for many of the items presented. In any case, when investing in computers to run a business, both the cost of ownership and the return on investment need to be calculated before the true cost of the system becomes apparent.

$Id: TCO2.html,v 1.2 1999/10/29 23:52:23 pfau Exp $