나는 누구인가먼저 이 글을 쓰고 있는 자신에 대한 소개를 하는 것이 예의일 것 같다. 나의 이름은 Manuel Soriano이고 컨설턴트로 일하고 있으며 리눅스를 좋아한다.
나는 늘 컨설턴트를 한 것은 아니었다. 예전에는 스위스에 있는 Polytechnic of Geneva 에서 공부를 했었다. 그당시에는 리눅스나 빌게이츠, 인텔등등이 없던 시기였다는 것을 말해두어야지... 사실 한층을 차지하고 있던 컴퓨터가 쓰여지고 있었다. 32K의 메모리램(놀랍게도 32K였다.K는 Kilo이다.)을 가진 사람은 잘 사는 사람이었고 아주 특별한 사람들만이 20MB짜리 하드디스크와 한두개의 마그네틱 테이프를 가지고 있었다.
이런 신기한 환경에서 약 200명이나 되는 우리들은 컴퓨터센터에 있는 컴퓨터를 사용하기 위하여 손에는 천공카드(punched card)에 기록된 우리들의 프로그램(FORTRAN)을 가지고 기다릴 수 밖에 없었다. 분위기는 거의 환각상태에 가까웠다. 우리가 카드리더를 사용할 수 있게되면 그것은 마법을 부릴 수 있는 시간이었다. 천공카드에 기록된 프로그램을 읽기시작한 후 5~6분이 지나면 프린터는 결과를 출력해냈다...그러나 대부분의 결과는 잘못된 것이 다반사였다! 그러면 잘못된 문장을 고쳐서 줄의 제일 뒤에 가서 다시 기다려야 했다.
리눅스를 발견하다아마 내 전문경력에 두번의 큰 일이 있었다. 첫번째는 ATT에서 만든 유닉스가 아닌 유닉스(3.0)가 설치된 Z80 멀티프로세서머신이었으며 6년이 지난 후 두번째일이 있었는데....바로 리눅스였다.
I was coming from huge and powerful machines (Control Data Cyber) with zillions of megas of hard disk, tons of memory, innumerable users thirsty for CPU time. One day someone comes and shows me a CROMEMCO (my first micro). Surprise! It had an operating system whose name was: Unix, what is that?
I kept it. Later in time I learned about other things, VMS, VM, AIX, PROTEUS, MS.DOS (and its sequel), UNIX ATT 4.2, etc...
For my spare time I acquired a 386 and installed on it a UNIX ATT 4.2
with X11(R3) and Motif(1.0), but in certain magazines there were already
advertisements of a version 5 of X11, Motif 1.2.4 and an operating system
(UNIX) that cost 50 US dollars. Well, those of you with working experience
in SCO or any other, will know what their licenses cost. I invested 250
US dollars on the purchase of that OS and Motif. It arrived after a month.
It was a Yggdrasyl with kernel 0.95. I gave it a whole hard disk
first, later all the computer, and finally my professional life (and I
do not regret it).
Evaluating Linux"Yes man it works!", "Believe me!, with X11R5 and MOTIF 1.2.4, it doesn't hang up and my RESET button is accumulating dust due to lack of use". This is what I told a Swiss friend of mine when he called me to Spain. He was exploiting Solaris, X11R4 and Motif 1.1 for PC's, each license cost him no-more-no-less than 8,000 US dollars and he was looking for something "more economic". I mentioned Linux to him. He did not hang up the phone on me because of our friendship but he indeed alluded to something about "being drunk". Nevertheless, he bought it.
This good man has a company (DAPSYS SA.) devoted to healthcare informatics. His specialty was storage and retrieval of radiological images for screen devices, either with X11 or Motif or Win95. After trying Linux for few days (15) he adopted it, sending to the garbage 2 AIX, 1 HP800 and 1 DEC with Ultrix. Today he is a happy man with a "portable" and his Linux.
This company already had a few installations in Switzerland when a great opportunity presented itself in Canada. Almost without warning two salesmen from a large American company showed up. They were not coming to sell something but to check "something" that "someone" mentioned to then at the RSNA in Chicago. The RSNA is the world event in the subject of radiology. Two months later the director of DAPSYS went to Montreal with his "portable" to demonstrate the product to several doctors; three months later (504 years after Colon), I'm off to the Americas to perform the first installation in that continent of Iris. Iris is an application for the archival and retrieval of radiological images.
This first installation clearly showed the robustness of the product, of which Linux is great part, and that we could handle the problem correctly.
St. JohnAll that to get here, to what interests us: the INSTALLATION. But first let me explain the specifics of the problem.
Radiology is a specialty that is usually very expensive both in equipment and in film stock. This last item can take 15 to 20% of the annual budget of any radiological department. For this and other reasons it was desirable to eliminate film usage and substitute it for a computer screen.
Furthermore, Iris allows one to centralize ALL the radiological events by patient. Each image has recorded in it the first and family name and patient code. Iris permits retrieval of this information through OCR techniques, extracting this data and archiving it together with the existing images of the patient (in case there are any).
Iris not only saves in film expense, but it also saves storage space (Hospitals must archive the images for all its patients for years), images can be shared simultaneously among several physicians kilometers away (even with an ocean in between), and they can be searched very rapidly.
In the anglophone part of Canada there is a tendency to centralize healthcare in phases; first regionally and then at the provincial level. Each province has a certain independence and the province that hired us for this first installation has seven regions.
In this region we are going to connect six hospitals though a WAN ATM of 100 and 10 Megabit, Linux has nothing to do with this network, it only works on it, but it allows us to stress the importance of its tasks.
The headquarters for the whole system is located at St. John hospital. It has two archives, the first one has the following specs:
In a first phase the network will consist of six hospitals, with two more being added during the second phase of the project. In the first phase we are going to connect the following:
What is Linux doing in a Hospital ?Right, why Linux?.
In the healthcare environment Linux has an incomparable strength: its price and the type of hardware it uses.
For the same functionality the final cost usually is about 50% or less compared to Sun, HP, IBM or DEC stations and Linux functionality has nothing to envy in the others.
The saving on Linux allows hospitals to install a larger number of stations and more hardware throughout the hospital and it allows them to increment its services, which in Canada are not public but semi-public. The hospitals are regional and depend on the regional governments, but their budgets are only partly supported by the regional tax-payers. The other part comes from the services provided to local establishments. Linux, therefore means better and more services for less costs and more income. It also means that hospitals can pass down to smaller clinics and patients the savings and can stay competitive.
With regard to the robustness of Linux , all I have to say is that I have a server with a record up-time of 26 months.
Nobody and that means NOBODY, has given us any objections to using Linux for the installation. What really counts is the solution and that the solution WORKS. This concept is very far from the old idea of setting up an operating system (with a known name) as a solution and then a program that sharpens that solution.
The truth is that we have not had any problems putting Linux into the various hospitals, or suffered any funny comments. On the contrary I have detected a certain respect with regards to the word Linux, but for some reason, this system seems to be most supported in the USA rather than Canada (although this may appear strange, because Canada is, in a sense, like the USA).
What are those two servers doing?A radiological image usually has a size from 256x256x1 to 8192x4096x2, which means it takes up a lot of storage space and so will only hold up to 6 months worth of information. After this time, the study of a patient passes to a different support (non mobile), and for the moment, on to DAT tapes. In the future we will most probably change to magneto-optical robots of 600Gb.
The topology of the network is very simple, for each hospital:
In a serverFor a given patient several X-ray explorations can be presented in various modalities. An exploration is a radiological exam, each exam may have one or more X-ray images or shots. A modality is a device capable of producing radiological images by means of X-rays, ultrasound, NMR, etc..
When the images are shot, they arrive directly to the server which has to be capable of "reading" the image and extracting the patient data in order to forward it to the visualization stations and archive them with the other patient data.
Before archiving an image, it goes through a process of "clean up" and non-destructive compression. We now achieve compression ratios of 60%. We could achieve higher ratios, but it would be at the expense of performance in the archival process and later restoration of the images.
According to the tests performed, an image of 512x512x2 takes 0.3 seconds, on average, to appear on the computer screen from the beginning of the transaction. I speak of averages because a study is not always a single image, this average can be improved both by means of hardware and by improving the method of retrieval.
St.John has 12 halls producing 1,000 images per day, this gives us an
idea of the volume the servers must handle. During the test period we were
reaching the CPU usage of nearly 2%, with 6 halls sending images.
In a clientThis is the easy part, there are three types of clients:
The workstations IRIS/Motif are, mainly, for visualization work with diagnostics. Since they can pilot up to 4 screens, but for the moment they only have two screens of 20 inches each. The searches of the database can be summarized by:
ConclusionTo finish, I can only say that this system has as much room for expansion as any other OS. An Operating System is only the support, and this support can be good or bad. The success of the solution depends upon the system and the application being able to work together.
Lately someone in a magazine has managed to spread the idea that the
solution for a problem stops by his favorite OS, but if we are good professionals
we must be able to convince people that WE know better than any journalist
what works or not.
번역 : 이주호
© 1998 Manuel Soriano
본 웹사이트는 Miguel Angel Sepulveda씨에 의해 관리됩니다.
© Emre Demiralp 1998