While I agree with you that good test automation is challenging, I would say that it is no worse than on-premise product development. Effective testing requires engineering commitment to make test automation development co-equal with product development. It is an immense challenge. BUT, having worked CI/CD for both an on-premise product and a cloud-based (AWS) product, cloud seemed easier. If for no other reason than it is much easier to manage the equipment.
Taught myself various flavors of assembly in high school (8080, 6502, 6809) as well as Apple integer basic. In college (1979) CS100 was FORTRAN taught by a "visionary" that was convinced that learning how to punch cards was a useful job skill. Best language learned that year was Pascal in an elective class.
No matter what the first language is, the second one should be assembler. Modern languages are great, but a lot of people have lost touch with how their fancy functional language actually executes on the hardware and what the memory consequences are of their code.
In 1978, I spent my entire life savings (about $2500) on an Apple ][. Maxed out the memory at 48K (it was $150 per 16K). I can't remember how much it cost, but I footed for the massive 130K floppy drive because one of my friends had a TRS-80 model 1, and I refused to try to load programs by cassette. No monitor - couldn't afford one and they were all monochrome anyway. It ran Integer Basic in much less than 16K. The add-on Applesoft board, which would have made it essentially an Apple ][+ was just a dream.
It was a great machine. I learned assembler and simple device drivers on that machine. I have fond memories of the big red book with the entire ROM assembly in print form. I saved up for about 6 months to afford the Applesoft add-on board that had a toggle switch sticking out of the back so you could switch between the onboard Integer Basic ROM and the floating-point Applesoft Basic on the board. Never, ever flip that switch while the machine is running. I also remember driving 3 hours to a place that had the new ROM chip for the floppy interface card that would let it store 160K on a drive instead of the base 130K. Except for the 6502 CPU, there was nothing denser on that motherboard than MSI chips. And I don't mean the computer/motherboard maker.
It lasted a good 5 years, but the last year it started having some kind of thermal problem that required keeping the lid off of it or it would lock up after about 15-20 minutes. The only way to fix it was to power off, flex the motherboard, and power on again. Sometimes I had to pull the CPU and clean the legs of the chip off with an eraser.
For all my promises to my parents that it would be good for my engineering degree, I mostly used it to play games. Choplifter, Ultima 2 and 3, Castle Wolfenstein, Miner 2049er, and Wizardry were favorites. I also had the world's most primitive wireframe flight simulator with single-digit frames per second.
I really wanted one of those all-in-one 6809 systems, or , an Altair box with CP/M and the massive S-100 chassis. Couldn't afford them, and really it was just as well - this way, I was able to watch one of my friends almost ruin his college career by skipping most of his classes for 2 weeks to beat Ultima 3 on it.
Programming on it was so much fun. No memory protection. No O/S. You powered it on, and you were in Basic. Nothing like a BIOS, though you could hook the system reset and interrupt vectors. Video directly mapped to fixed memory ranges. Hacking the floppy disk driver was a piece of cake. There was no modem built in. There may have been an acoustic modem coupler that connected to the built-in audio out/mic jacks, but I might be mis-remembering that. If you put a radio too close to it, you could hear the apple work because of its EMR. The 1 MHz CPU
I'm at re:Invent, and as they laid it out, the difference is in storage, read replication, and provisioning. Storage is in log-based files, and they've set it up so that crash recovery runs in seconds since the redo log playback runs differently. Storage is also all in SSDs. They've optimized read replication and between that and the log-based files, read replicas do very little write processing and are thus able to serve more read requests. Finally, it auto-provisions in 10GB chunks up to a max of 64TB, and the storage cost depends only on what you use. You do have an hourly charge for the master node instance.
TL;DR: It has fast I/O and read replication because AWS can leverage the knowledge of their infrastructure, and it provisions itself.
Slashdot Mirror
While I agree with you that good test automation is challenging, I would say that it is no worse than on-premise product development. Effective testing requires engineering commitment to make test automation development co-equal with product development. It is an immense challenge. BUT, having worked CI/CD for both an on-premise product and a cloud-based (AWS) product, cloud seemed easier. If for no other reason than it is much easier to manage the equipment.
Taught myself various flavors of assembly in high school (8080, 6502, 6809) as well as Apple integer basic. In college (1979) CS100 was FORTRAN taught by a "visionary" that was convinced that learning how to punch cards was a useful job skill. Best language learned that year was Pascal in an elective class. No matter what the first language is, the second one should be assembler. Modern languages are great, but a lot of people have lost touch with how their fancy functional language actually executes on the hardware and what the memory consequences are of their code.
In 1978, I spent my entire life savings (about $2500) on an Apple ][. Maxed out the memory at 48K (it was $150 per 16K). I can't remember how much it cost, but I footed for the massive 130K floppy drive because one of my friends had a TRS-80 model 1, and I refused to try to load programs by cassette. No monitor - couldn't afford one and they were all monochrome anyway. It ran Integer Basic in much less than 16K. The add-on Applesoft board, which would have made it essentially an Apple ][+ was just a dream.
It was a great machine. I learned assembler and simple device drivers on that machine. I have fond memories of the big red book with the entire ROM assembly in print form. I saved up for about 6 months to afford the Applesoft add-on board that had a toggle switch sticking out of the back so you could switch between the onboard Integer Basic ROM and the floating-point Applesoft Basic on the board. Never, ever flip that switch while the machine is running. I also remember driving 3 hours to a place that had the new ROM chip for the floppy interface card that would let it store 160K on a drive instead of the base 130K. Except for the 6502 CPU, there was nothing denser on that motherboard than MSI chips. And I don't mean the computer/motherboard maker.
It lasted a good 5 years, but the last year it started having some kind of thermal problem that required keeping the lid off of it or it would lock up after about 15-20 minutes. The only way to fix it was to power off, flex the motherboard, and power on again. Sometimes I had to pull the CPU and clean the legs of the chip off with an eraser.
For all my promises to my parents that it would be good for my engineering degree, I mostly used it to play games. Choplifter, Ultima 2 and 3, Castle Wolfenstein, Miner 2049er, and Wizardry were favorites. I also had the world's most primitive wireframe flight simulator with single-digit frames per second.
I really wanted one of those all-in-one 6809 systems, or , an Altair box with CP/M and the massive S-100 chassis. Couldn't afford them, and really it was just as well - this way, I was able to watch one of my friends almost ruin his college career by skipping most of his classes for 2 weeks to beat Ultima 3 on it.
Programming on it was so much fun. No memory protection. No O/S. You powered it on, and you were in Basic. Nothing like a BIOS, though you could hook the system reset and interrupt vectors. Video directly mapped to fixed memory ranges. Hacking the floppy disk driver was a piece of cake. There was no modem built in. There may have been an acoustic modem coupler that connected to the built-in audio out/mic jacks, but I might be mis-remembering that. If you put a radio too close to it, you could hear the apple work because of its EMR. The 1 MHz CPU
I'm at re:Invent, and as they laid it out, the difference is in storage, read replication, and provisioning. Storage is in log-based files, and they've set it up so that crash recovery runs in seconds since the redo log playback runs differently. Storage is also all in SSDs. They've optimized read replication and between that and the log-based files, read replicas do very little write processing and are thus able to serve more read requests. Finally, it auto-provisions in 10GB chunks up to a max of 64TB, and the storage cost depends only on what you use. You do have an hourly charge for the master node instance. TL;DR: It has fast I/O and read replication because AWS can leverage the knowledge of their infrastructure, and it provisions itself.