Agreed. The author is confused and has described a form of compatibility not scalability - very different animals. A simple check with whatis.com would have provided a reasonable definition of scalability.
In short, scalability refers to the ability of a system to 1) retain reasonable performance levels as the demand for a resource grows and 2) accept additional resources to meet a growing demand. ("Reasonable" addressed below.)
To scale a system "vertically" for example, one might add more processors, memory, or disk. In this case, a system's "scalability" refers to the ability to add those resources and get a reasonable increase in performance for having done so.
To scale a system "horizontally", one might add more nodes (servers) to a cluster of load-balanced servers. This method can sometimes compensate for the lack of "vertical scalability" in a system.
So what is "reasonable" performance? Obviously it depends on one's goals and the trade-offs one is willing to make. If one doubles the number of processors in a server, but gets only a 10% improvement in performance, one could easily say that the system does not scale well. On the other hand if one doubles the number of processors and gets nearly double the performance, one would say that the system has scaled extremely well.
Likewise, if one doubles the demand on a system, and the throughput only increases by 10%, the system is not scaling well (which may or may not be solvable by adding additional resource).
Also, AIX does not run on mainframes. In the second paragraph of the Scalability section, the author appears to believe otherwise. AIX runs only on pSeries (formerly known as RS/6000).
Linux however, runs on all IBM eServer hardware: pSeries (POWER), zSeries (mainframe), iSeries (AS/400), and of course xSeries (Intel).
There was such a thing a few years back. You could guess at the command line options and be right. It was incredibly consistent. The authors obviously knew unix and wanted to start all over with a good design. Graphics were fast too cause it didn't need X11. And good filesharing w/o NFS. Had a cool editor build into the windowing system. It was called Aegis/DomainOS. It was from Apollo computer in Chelmsford, MA before HP bought them and killed it. And it used to scream in 4MB. If you wanted to run the cool gui debugger (unheard of at that time) you needed about 8MB. They started adding Unix stuff to it around the time HP bought it and it got slow and ugly. But it WAS COOL. see comp.os.apollo.
Slashdot Mirror
Agreed. The author is confused and has described a form of compatibility not scalability - very different animals. A simple check with whatis.com would have provided a reasonable definition of scalability.
In short, scalability refers to the ability of a system to 1) retain reasonable performance levels as the demand for a resource grows and 2) accept additional resources to meet a growing demand. ("Reasonable" addressed below.)
To scale a system "vertically" for example, one might add more processors, memory, or disk. In this case, a system's "scalability" refers to the ability to add those resources and get a reasonable increase in performance for having done so.
To scale a system "horizontally", one might add more nodes (servers) to a cluster of load-balanced servers. This method can sometimes compensate for the lack of "vertical scalability" in a system.
So what is "reasonable" performance? Obviously it depends on one's goals and the trade-offs one is willing to make. If one doubles the number of processors in a server, but gets only a 10% improvement in performance, one could easily say that the system does not scale well. On the other hand if one doubles the number of processors and gets nearly double the performance, one would say that the system has scaled extremely well.
Likewise, if one doubles the demand on a system, and the throughput only increases by 10%, the system is not scaling well (which may or may not be solvable by adding additional resource).
Also, AIX does not run on mainframes. In the second paragraph of the Scalability section, the author appears to believe otherwise. AIX runs only on pSeries (formerly known as RS/6000).
Linux however, runs on all IBM eServer hardware: pSeries (POWER), zSeries (mainframe), iSeries (AS/400), and of course xSeries (Intel).
There was such a thing a few years back. You could guess at the command line options and be right. It was incredibly consistent. The authors obviously knew unix and wanted to start all over with a good design. Graphics were fast too cause it didn't need X11. And good filesharing w/o NFS. Had a cool editor build into the windowing system. It was called Aegis/DomainOS. It was from Apollo computer in Chelmsford, MA before HP bought them and killed it. And it used to scream in 4MB. If you wanted to run the cool gui debugger (unheard of at that time) you needed about 8MB. They started adding Unix stuff to it around the time HP bought it and it got slow and ugly. But it WAS COOL. see comp.os.apollo.