3 ms·
I agree. I went from dot-com Unix admin to investment bank Unix admin in 2000. Luckily it was via a consulting company. They said my interview process was un
by Ologn 12y ago
I agree. I went from dot-com Unix admin to investment bank Unix admin in 2000. Luckily it was via a consulting company. They said my interview process was unusual, because every internal interviewer passed on me, while every interviewer from a dot-com gave me a strong recommend. I was interviewed by them many, many times and finally accepted.
It was very different. At the dot-com's that I was at, web servers and application servers were Red Hat Linux, and only the databases were Sun boxes, with E4500 being a high end box. At a post-IPO dot-com there might be one E6500 and maybe a few Sun E4500's, and lots of Red Hat machines. At the investment bank they were all Solaris on the Unix side, and had dozens of E6500's in my city, and hundreds of E4500's. The investment bank had admins in the OpenBoot PROM all the time, whereas I had barely been in it. The investment bank was always replacing disks in the hundreds of external RAID arrays they had whereas only our database had a large external RAID array. They used NFS extensively and you had to know it thoroughly. My knowledge about Red Hat Linux, Apache web servers, Java application servers etc. didn't amount to much.
It was qualitatively (and quantitatively) very different, and in interviews it sounded like I didn't know what I was talking about.
On the developer side, the bureaucracy is very different. If a minor change in a software package is being pushed to New York, London and Tokyo, which often it is, you might have five meetings about it, and the code push might take weeks or even months. No matter how trivial the change. Three different divisions might have to sign off on the change. It's definitely not an agile code process, with code going from your fingers to the production web/app within the week.
- kokey 12y agoUnix admin here too. I also spent more time with Linux and FreeBSD during the dotcom era, but, strangely, for a lot of enterprise companies because I was working more on the security side of things. When I had to cast my net wider after the bubble burst, and after having moved to a new country, my lack of deep enterprise Solaris experience did bite me. Fortunately after a few years Red Hat was starting to do really well in the enterprise market and I ended up in banking, where I picked up a bit of Solaris again but I didn't really bother too much since the bank was actively getting rid of Solaris. I suspect many technologies used in the startup world now will get taken on in the enterprise after the bubble burst, but it will certainly take a few years and these companies will be weary of people who doesn't have enterprise experience.
- spacecowboy_lon 12y agoHaving worked at large telco's it makes me smile when i got marked down on for suegsting that you sould have identical hardware on test dev and live and I mean identical down to havng the systems from the same prodcution run as ideal woudl all the DASD the whole devlop on a mac book when the actual production systems are linux makes me wince.
- kephra 12y agoI was lucky to be an old sack. I moved from mainframes to so called open systems, around the time when IBM decided to ship OS without source code. And when the dot.com burst, I had my mainframe skills, still rusty but good enough that my Y2K PTFs had been the ones distributed by CBT Tapes. So I was in hot demand to supervise Y2K tests, and later to migrate mainframe scripts and applications to Unix. I did not stay in enterprise for long. I like mainframes, but I hate the culture of big companies. I next helped founding a university, and when tired by ivory tower, I left uni to fund a company to focus on machine learning for industrial applications. So, most important advise from an old sack: stay flexible, and dont be surprised that old skills in solid technology might be worth good money at the moment when the hypes end.
- DonHopkins 12y agoI must ask: What were investment bank admins doing in the OpenBoot PROM (aka Open Firmware), exactly? Hacking FORTH, using Emacs on the keys, and then saving it in NVRAM? Did you know that Open Firmware has its own theme song? https://web.archive.org/web/20060111132138/http://playground.sun.com/1275/misc/ofwsong.au https://web.archive.org/web/20060111132138/http://playground... http://everything2.com/title/The+Open+Firmware+Song http://everything2.com/title/The+Open+Firmware+Song Sung to the tune of "The Flintstones" Firmware Open Firmware It's the appropriate technology, Features FCode booting Hierarchical DevInfo tree. Hack Forth Using Emacs on the keys, Save in NVRAM if you please. With your Open Firmware You can fix the bugs in no time Bring the kernel up in no time We'll have an FCode time! I hacked FORTH with Mitch Bradley, the father of Open Firmware, as a summer intern at Sun in 1987. I love OpenBoot PROM, and his 32 bit 68k/SPARC Sun FORTH system that it's based on, which was derived from Langston and Perry's Forth-83! He also developed a version of it used in the OLPC XO-1. http://wiki.laptop.org/go/Open_Firmware http://wiki.laptop.org/go/Open_Firmware
- Ologn 12y agoCool. For production slated machines, we were usually running commands such as "copy-clock-tod-to-io-board", nothing too creative. While I was there, the Sparcstation 20's reached end-of-life and we were allowed to bring one home (I still have it). At home I had gotten ADSL through a reseller that had an odd routing setup - they wanted the MAC address of my desktop so that they could assign an IP to it. I guess I should have gotten a hub that could do NAT, but I gave them my Intel PC MAC, and then went into the SS20 OBP and changed the MAC address to that of my PC. It worked too, both machines got ADSL. Actually what I wound up doing is doing NAT with the SS20, after I swapped out the PC ethernet card with a spare one lying around.
- DonHopkins 12y agoIf you're wondering how it works, it's self documenting: besides the usual "help", it also has a decompiler/disassembler built in: "see", which you use as a prefix before other words, like "see see" to decompile the decompiler!