Microsoft admits that TCO for developing nation's schools is same for Linux and Windows.
Paw Prints: Writings of the maddog
This is a very long blog entry. I wondered whether it would be better to write it here, or to write an article for Linux Pro Magazine. The editors felt that because of the timeliness of the subject it would be better to have it as a blog entry. For those who do not want to read to the end, the "Executive Summary" is:
Microsoft admits that TCO for education students in developing nations is about the same with their products and Free Software. They further admit that the initial TCO costs of hardware and software are lower with Linux, and that Linux technical people in developing nations command a higher salary than their Microsoft counterparts due to scarcity.
I feel that this mandates greater use of Free Software in universities and high schools to help increase the number of locally employed, highly trained Free Software professionals to decrease that scarcity, and to lower the balance of trade problems that occur by sending software royalty money outside of the country.
Now for the meat of the analysis:
Microsoft recently released a study on "Total Cost of Ownership" (TCO) for schools installing large numbers of computers in developing nations. I have been seeing these "TCO" studies for about twenty years, maybe thirty years, I cannot remember when they first started.
Lately I have seen these reports trying to convince me that proprietary software is somehow less expensive than Free Software. I normally have a big problem with most of these reports in one way or another, as they do not go into the depth that they should for real TCO. I normally pay little attention to them, but this latest report was mentioned in the blog of a nationally syndicated "Microsoft watcher," and after reading some quotes from some Microsoft people in her blog, I decided to take a look.
This latest study, done by a company called "Vital Wave Consulting" (published August 15, 2008, and funded by Microsoft, then posted in Microsoft's "Unlimited Potential" (UP) Blog) was different. They did a great job of justifying the use of computers in teaching, and particularly for developing nations, citing lots of reports. They did a fairly comprehensive study, and had lots of citations on TCO. Microsoft thought so much of this report that in the blog written by James Utzschneider, General Manager of Marketing and Communications for Microsoft's "Unlimited Potential" program, he acknowledged that the initial cost of the Microsoft software and hardware was "slightly higher" than that of Linux and Free Software.
Mr. Utzschneider went on to praise the report, which said that due to the higher salaries of FOSS systems administrators and other support personnel in areas like China and South America, TCO over a five-year period for Window and Linux solutions was "about the same". The only shock registered by Mr. Utzschneider was that the per-computer five-year cost for both Windows and Linux was coming in at about US$ 2,700 per seat, and suggested that this would be an area that might benefit from having Microsoft servers in the schools as well, to help lower the cost of administration.
Now I found this "study" interesting in many ways. First of all, I thought that of all the TCO studies that Microsoft had probably funded, surely they could come up with a better story than "about the same" when it comes to TCO between Microsoft and Linux. Still, it was nice that Microsoft had stated publicly that the TCO was "about the same" and I was really glad that Vital Wave Consulting's report had reached that assertion in an independent fashion, even though the study was funded by Microsoft.
Vital Wave Consulting did squirm a bit in the study with Microsoft's royalty costs, saying that the cost of the software was "only" 2 percent of the TCO. Now 2 percent of $2700 is $54, or about $11 per year. And Vital Wave was right in saying that it is a small number compared to the total amount. However, as I dug into the reports, I thought of that money another way. The royalties were $25 per computer for the Microsoft Starter Edition, and $12.50 per computer for Microsoft Productivity Suite for the five year period. They also mentioned a $6 annual "anti-virus" fee that was needed for Windows, but (they admitted) was not needed for Linux. That is $43.50 per computer that flows out of the country to Microsoft's (and some anti-virus company's) coffers that does not have to be paid by the Free Software users. These charges are most of the reason cited as to why "Linux is cheaper in the initial outlay."
But the report did not mention other add-on programs that students might want to use. Adobe Photoshop? SQL database? Video and audio editing tools? Project management tools? Are these extra, or included in the $43.50? And what happens when the students want to use these tools at home? Are they covered by the school's license? What happens if their parents want to use the student's computer at home? Can they, or is this a violation of the licensing? It would not be a violation if the student were using Free Software both at the school and at their home.
As I said, I do not think that $43.50 is a lot of money, but I also remember all those advertisements by the World Health Agencies telling me how many children could be inoculated by just $1 per month, and the trade off of the $43.50 times the 1600 OLPCs that are being shipped every day with Linux on them in Uruguay, and will be shipped every day for the next year. Pretty soon that $43.50 per computer adds up to "real money."
But when you start to look at the REAL costs, the TCO starts changing rapidly.
As I started to look into the reports, there were some strange assumptions. The report attempted to compare the costs of a new "mainstream" versus a new "Low-cost", "Ultra-low cost," and "Second Hand" computer over the course of five years. But they treated the end of the five years as if everything came to a grinding halt, and after that five year period everything would be purchased new, replaced, and a new five-year period would begin. This is where most of their study falls apart. TCO does not stop after five years.
As an example, they cite the cost of the "second hand" computer as $254 and having a three-year lifespan. At first I could not see whether they got the computer as a donation, or to whom the $254 was paid. In the countries that I have visited, a second-hand computer is typically donated from a bank or another entity that was updating their own systems. In some schools (such as the one in Curitiba, Brazil that I visited), the students pulled the systems apart, sorted the components, and put the systems back together again. Broken disk drives were discarded and replaced with others mined from the extra systems. Bad memory was detected, and machines with lesser amounts of memory had more memory put in them. The students in this school learned almost as much about computers doing this type of work than they did from their studies.
Or, if the students did not do this reconfiguration work, perhaps it was done by a charitable organization as I found in Colombia, busy teaching computer maintenance techniques to their students and using that opportunity to refurbish and restore donated computers for other schools and poorer people in the community. But the hardware certainly did not cost the end user $254.
Reading further, I realize that the $254 cost came from professionally "refurbished" computers. Included in this $254 cost is a "comprehensive, one-year on-site warranty," and while I strain to think of that type of contract for a rural school, which has only five students and no electricity, I should point out that this refurbishing would then typically be done by a local vendor, and not some vendor in a foreign country. Local work spawns local jobs and local wages, important to governments who are funding education. I never found out how old the "second hand computers" were, only that they were characterized as a "Pentium 4."
Secondly, the assumption that the life cycle was three years for the older equipment, and five years for the "mainstream" computer also does not seem right. If you assume that the "second-hand" equipment has a three-year life and is then fully "depreciated" and replaced in kind, then after the five-year TCO study is over, the "second-hand" computer stream still has one year of useful life left to it. Two three-year periods is six years, not five, whereas the "mainstream" computer is fully depreciated after five years by this study, and is assumed to have no future life.
Another assumption is that the ultra-low cost computer has a three-year lifespan. Ultra-low cost computers have not been out very long, but from what I know of them, they seem to have a couple of things in common:
- no fan
- no disk
From my perspective, electronic things that have no moving parts seem to last longer than things that have moving parts. My stereo system, radio, and TV have lasted me twenty years. My computer disks and fans seem to last a much shorter period of time. In salty conditions (aboard a sailboat, or on an island), fans and disks last a much shorter period of time. I am told by a friend of mine that the average lifetime of a fan near the ocean is about one and one-half years....then the fan stops and the CPU burns up.
So I might assume that the life of an ultra-low system might be at least as long as the desktop system (five years), assuming you use the right software. (More on the "software longevity" later.)
Next was the configurations that were listed. The "mainstream" computer was listed as a "client and PC server," which they meant to serve both categories. On the other hand, it was listed with "only" 2GB of main memory and a Intel Pentium dual-core with a 320GB disk. Now you could look at this two ways: not enough memory and disk for Vista, or way too excessive memory and disk for Linux. The "low-cost" computer (albeit new) was still marked as a "client and PC server," but only had 1GB of memory (yet still had the 320GB disk and the Intel Pentium dual-core processor). In fact, the only difference listed in their report between the "mainstream" computer and the "low cost" computer was the 1GB of memory and about $250 in total cost of the hardware systems. I guess it was very expensive RAM at $250 per gigabyte.
On the other hand, the "second hand" computer had a Pentium 4 processor and only 512MB of RAM. But most interesting was that it had only 40-80GB of disk space. Now I can begin to see the real reason why the "second-hand" computer will last only three years. In three years, you will be able to fill up that 40-80GB of disk space with downloaded music (I'm just being sarcastic, folks). Of course if I had to pay $125 for a half-GB of RAM, I might think twice about upgrading my "second hand" computer to hold a full GB of RAM.
One more interesting point is that the "Ultra low-cost" computer is also listed as a "Client and Laptop Server." Strange. It is described as a "mini-laptop," has an x86 running at less than 1GHz, and has only 512MB of memory in it, with a 2GB flash card as its storage. It also uses "wireless mesh networking" while all of the other systems described used Ethernet. Most of the "mini-laptops" I have seen have both wireless and wired networking, but I would not think of using them as a server for anything. I guess I just do not live on the wild side.
There were a couple of other points that I noticed in the report. Although electricity costs were mentioned for running the systems, and the need for air conditioning for the computer labs was mentioned, I think that the report was naive in several areas:
First, most of the rural areas the report talked about will never have air conditioning. Period. I know of sites in various countries that simply do not have electric power, and the plans to deliver computers to them consist of solar panels and/or wind generators. Inveneo, a company in San Francisco that has lots of experience in putting in low-power telephone systems in Africa, often tells the story of how the tribal elders will select a bicycle with an alternator attached instead of a solar panel because "it is cheaper, easier to fix, easier and faster to get spare parts, and we can give a job to the person who pedals the bicycle." Although someone could pedal a bicycle for a couple of hours to charge a few laptops, they would have to have legs of steel to run an air conditioner for a class of 16 "mainstream" computers.
Second, while the report acknowledged that "off-grid" (i.e., electricity generated by generators and other "non-mainstream" electricity) was more expensive than the "grid," they relegated that issue to "remote areas of Africa." I know of lots of schools in various developing nations that are "off the grid" and the generation of electricity is very expensive. I also know of schools in Northern Canada, Western United States, Appalachia, Katrina-hit areas, Baghdad, and Buddhist monasteries (as well as students in cabins, recreational vehicles, and boats) that all use fairly expensive "off grid" electricity. Although I grant that students in cabins, RVs, and boats are not "main-stream", they are people that have to be considered for education.
Third, in the equatorial areas that are on the grid and do have electricity (and air conditioning), you should count the cost of electricity twice: once to run the computers, and once to cool the computers. Without doing this, or without having specially made systems, you will have a greater failure rate because of heat loss.
The report also talks about the supplying of a UPS for the classroom. They went into great detail and cost about the UPS system, its batteries and replacement batteries to supply 1-2 hours of "backup power" for laptops. Now I do not know about your laptop, but my laptop has batteries built into it, which sustains it (on a good day) for about three hours (and I use a three-year-old Thinkpad, which is still going strong, thank you). Adding a UPS to power the laptop is not exactly my first priority to TCO.
The UPS specified by the study cost $831 for a minimal UPS for an urban environment, and if you were in a rural area using "mainstream" computers, you would be spending $3200 on the UPS alone for 16 desktop computers. I guarantee you that the air conditioner mentioned earlier will drain those UPS batteries very quickly.
Laptops do not really need a UPS, but if they did, the UPS in emerging nations would probably be made out of old car batteries operating at 12 volts, and charged with a 12-volt alternator, solar panel, or wind generator.
With all that concentration on something as obscure as UPS systems, you would think that the report would also discuss the fact that most proprietary software does not support a lot of developing nations native languages. Now some vendors are fairly good about supporting the major languages of the world. From Microsoft's own website, they supported 17 languages in Windows 2000 (probably more because they list support for "Western Europe," and I think that area requires more than just one language, as does "Central Europe"). Microsoft listed 91 languages in XP and Windows 2003 (where they seem to have gone in the opposite direction and included in those 91 languages 6 different French languages, 13 types of English, and 19 different Spanish languages). This is not so say that Microsoft is "padding" their language list – I believe the dialects of all of those major languages might warrant different listings, but it is just odd that "Western Europe" seemed to be enough for Windows 2000.
In Ubuntu 7.04 they listed 127 different languages supported. English was only listed once, as was Spanish and French.
Both groups have probably moved on in language support (the Microsoft list was last updated in December of 2007), but I think it was an interesting snapshot.
The real point is that all of the tools necessary to do language support are in the hands of the end user with Free Software. New language support for Microsoft's products are in the hands of Microsoft. At a meeting in August of 2007 in Ethiopia, Microsoft proudly announced that they would have support for Amharic, one of the languages of Ethiopia (where the meeting was hosted) by the end of 2008. I looked on my Ubuntu 7.04 system, and the language was already supported.
The cost of the software in the TCO study was, of course, based on special licenses that Microsoft has granted the school systems. And the report was careful to acknowledge that the licenses on that software was granted in perpetuity to the schools. This is good, since the EULA granted by Microsoft's OEM license for new computers usually does not allow for software to be transferred to computers when they are bought as second-hand.
On the other hand, nothing was said about the issue of having to upgrade the hardware after three years because Microsoft is no longer supporting the software that you were using at the beginning of the five-year TCO study. Let's say you bought XP last year. Your computers that you bought run XP fine, but perhaps you bought the "second hand computers"....the ones with only 512MB of RAM (didn't Bill Gates say that 640KB of RAM should be enough for anybody....I digress). Now you go to get a new "second-hand" computer and you find out that it is not supported by XP. In fact, it is only supported by Vista....and it needs 2GB of memory (at least).
Or halfway through your five-year period Microsoft decides not to support XP any more. No more software updates, no more security patches. Now you do have to switch to Vista. How much will it cost to upgrade our hardware? Can you say "DRM ready"? What about the costs of upgrading, the training needed?
When you do not own your software, you are not in control. Lack of control in your software is one of the "hidden costs" that is never mentioned in the study.
The other assumption made by the report is that at the end of five years your "mainstream" computer has no value. This might be particularly true if you tried to sell it to someone and they had to buy replacement software for it. The cost of commercial, proprietary software would be in the hundreds, if not thousands of dollars, making it impractical for anyone to buy your used equipment, unless they were going to use pirated software...or if they were going to use Free and Open Source Software.
The final technical coup d’état that I have for this study is that it seemed to completely ignore the mainstream "educational lab" masterpiece of Free Software, the Linux Terminal Server Project (LTSP).
LTSP reduces traditional hardware costs by doing most of the work on the server system. The clients can be configured with very minimal memory, and modest CPU power. The clients mostly run diskless (sometimes a small swapping disk is installed in the clients), so the software maintenance is all done back on the server. Because the computing needs and power are put into the shared server, the over-all cost of the hardware drops dramatically, even for new hardware. If the clients are run diskless, even "second hand" hardware has a life much longer than three years.
Now we come to the real differentiators in this study. While the study admits that the start-up cost of Linux is cheaper (even without the lower hardware requirements of the LTSP system, or the re-use of donated hardware), they say that the five-year TCO is higher due to "higher costs of long-term support," a result of the higher wages for the scarcer Linux technical people.
Consistently citing a higher cost of technical support due to a shortage of people trained to do Free Software work, the report manages to struggle through creating a TCO study that shows (by Microsoft's own admission) that the costs are "almost equal."
Here is where the rubber meets the road:
With Free Software, the money paid for the local support stays in the local environment. The royalties paid for Microsoft software, however low, flows out of the local environment and into the hands of Microsoft corporate.
The local money paid as higher wages goes to providing disposable income for the workers, which can be used to buy local food, local housing, and pay local taxes for local people.
The local Free Software professionals can demand a higher salary because they can not only install and integrate the software, but they can change it to meet the customer's needs. They are not dependent on a foreign entity to make the software do what they need.
By working with Free Software, the local Free Software Professionals build their expertise level inside of the country's economy. They produce software as a function of their professional lives, and contribute this back to the Free Software community in general. This software is now available for anyone in the country (or even outside the country) to use, not only the educational community. This leverages other industries and technologies to be developed inside the country.
When students are taught in the elementary schools, high schools, and universities only to use proprietary software to solve their problems, they naturally gravitate toward proprietary software as they graduate. Teachers only know proprietary software, and only teach proprietary software. Naturally the number of software professionals who know FOSS are fewer, and this demands a higher salary level.
But if we break the cycle and start teaching with Free Software today, start teaching how to develop Free Software today, then in a short time (two or three years) we will have the work force that will deliver support without the shortages that drive up costs artificially. Then, as this report sponsored and quoted by Microsoft admits, the TCO will be less for Free Software. Starting this today will require legislators and teachers who are far-seeing, and not just concerned about the next two, three, or four years. Oh, wait: This was a five-year TCO study.
Or else it will take legislators and educators who are willing to admit to the students that they are being trained for lower-paying jobs doing Microsoft installations. It will take legislators and educators to tell employers that they have been teaching topics to students that increase the TCO (by Microsoft's own admission).
Every year I go to Brazil to an event called FISL. There I meet more than 7,000 Free Software Enthusiasts (most in high school and college) who will really love to know that they will get paid more than their Microsoft counterparts when they graduate. I can hardly wait to tell them.
Of course, I also found it interesting that the Vital Wave Consulting report was marked as:
"Vital Wave Consulting, Inc. Confidential and Proprietary. Do not copy or send to third-parties in whole or in part"
and when I went to look for the UP blog entry, it was mysteriously missing. Fortunately, Microsoft still had the report in their download area so I could read and analyze it. "Do not copy or send to third-parties..." – was Microsoft encouraging me to pirate Vital Wave's report? Or are they keeping it private so they can put it directly into the hands of legislators in developing nations, trying to convince them that the cost of their software is insignificant to TCO? Here is the link so you can see it for yourself. Better hurry before Microsoft removes that link too:
Sometime later this week I will tell you why TCO is not what you should be considering for yourself or for your business, and I will tell you why Free Software professionals will always be worth more than closed-source proprietary professionals even if they are paid the same. IT is all about ROI (Return on Investment), and that is where Free Software really wins.
Second hand computers in CambodiaI have just returned from Cambodia, and before I left I was trying to sell my IBM Net Vista Pentium 4 1.7 Ghz machine with 256 mb RAM and a 15" LCD monitor. It ran Ubuntu pretty well.
The computer shops that dealt in second hand computers wouldn't take it. Second hand P4 3 Ghz machines with 512 mb and 17" LCD monitors were selling for around $US250. They are imported in their hundreds from Australia and Japan, where I guess they are considered old.
As to software, it is between $2 - 5 dollars a DVD. For anything. The only software that was difficult to get was Open Source software - and it usually had 'Requires Windows XP or better' on the disk. So proprietary software is becoming the default.
Cambodia has an official policy of favoring Open Source, and there is a lot of work on localization and training. But Microsoft has opened an office and is pushing MS products.
It is hard to believe Microsoft and others could not stop the counterfeit disks being distributed - book publishers who are prepared to make a fuss have had counterfeit copies of their books removed from general circulation. It seems allowing the disks to circulate is the most effective way to hook the user - and one day it will all need to be paid for.
And all the NGO's - UN, UNDP, WHO etc as well as smaller ones - claim to support this policy, but always advertise for MSCE's. NGO's are considered good employers because they pay better and give you the weekend off.
So it is hard to know if Linux will be adopted. Software companies will probably claim the 'lost earnings' as donations and get a tax deduction.
They don't know the worldDear maddog,
You're right about your points and I'm afraid that they don't know the world. When I work in southeast asia, we don't had "regular power", computers was just couple of pieces and the people live with just US$ 80/month. How they can pay to MS? Never, they use FLOSS.
Also, at now some people in East Timor told me that they create an user group to translate the Ubuntu to tetum (one national language, beside portuguese). Microsoft had some idea what is tetum? A language used for 800.000 people? Of course not. And why this people need to learn english? No, they need to use a software in the national language and just the FLOSS can do that.
If the study is right, they don't know the world like I know and like you know.
Jon Maddog Hall's articleJon's article was full of good points.
An addition might be that FOSS administrators are able to administer
many systems, more than administrators tasked with proprietary
Thus, the higher salary costs of the FOSS admins is typically spread
over more systems to the advantage of FOSS.
The company is collaborating with Google and Intel to use Kubernetes as an engine for Fuel
Customers can take a free test drive of SLES for HPC on the Azure Cloud
San Francisco-based chip company announces their first fully open source chip platform.
The whole distro gets rebuilt on glibc 2.3
Ubuntu Vendor tries to solve app packaging and distribution problem across distributions.
Founder of ownCloud launches the Nextcloud project.
Will The Machine change the way future programmers think about memory?
The new Torus distributed storage system is available under an open source license on GitHub
Juries decides Google’s use of Java APIs Was Fair Use