Date: Tue, 29 Mar 2005 12:23:46 +0200 From: Anthony Atkielski <atkielski.anthony@wanadoo.fr> To: freebsd-questions@freebsd.org Subject: Re: Anthony's drive issues.Re: ssh password delay Message-ID: <1648629793.20050329122346@wanadoo.fr> In-Reply-To: <LOBBIFDAGNMAMLGJJCKNEEPAFAAA.tedm@toybox.placo.com> References: <42480F8B.1060405@makeworld.com> <LOBBIFDAGNMAMLGJJCKNEEPAFAAA.tedm@toybox.placo.com>
next in thread | previous in thread | raw e-mail | index | archive | help
Ted Mittelstaedt writes: > The main point I've been trying to make is that just because FreeBSD's > drivers don't support whatever modification has been made in the Adaptec > code on the Vectra, does not mean that the FreeBSD driver is "broken" > or "has a bug" in it. When something doesn't work, it's broken. If you consider a change in firmware to be a hardware problem, then a lack of proper handling of the firmware in the OS must also be a problem. I don't see why the same standards wouldn't apply to both. The key point here, though, is that Windows apparently works correctly with the firmware, whatever changes that firmware may contain. FreeBSD does not. Therefore FreeBSD is broken. > But you are correct in that these are trapdoor systems - if you do not > install the Compaq/HP-written drivers at the right times during the > install, then Windows loads it's default drivers which may or may not > (usually not) work. And once loaded you cannot unload them and replace > them with the manufacturer-supplied ones because the operating system > won't let you do things like unloading the device driver that runs the > controller that the system disk is on, things like that. You have to > nuke and repave. I often wonder why people even buy servers from these vendors when they have so much vendor-specific junk on them. I suppose there isn't much competition. It seems to me that it should be possible to build equally good servers with entirely off-the-shelf, standard components, and no magic firmware or software. But vendors apparently cannot resist changing _something_. > I think Dell is the same way, though. I suspect all the name brand > systems are - that is why people buy name-brand server systems, to get > the extra little features like the preemptive disk failure monitoring, > the case-open/case-closed, temperature, fanspeed, power supply voltage > monitoring, and all the other proprietary little features. I suppose it's seductive initially, but after fighting with proprietary hardware and software for a while, it gets old. Forget the case-open switch and the three-dimensional beeping animated temperature monitoring application, and just buy commodity hardware and software. In exchange for sacrificing a few frills, you get something that behaves predictably and can be maintained cheaply without critical dependencies on one supplier. I'm pleased that I built my current server myself out of stuff bought right off the shelf. It may be 1-2% less performant than a name-brand, all-in-one server, but at least I know exactly what's in the machine, and virtually none of it is dependent on any single supplier or single model of hardware component. If I want to buy spare disks, I can get them for €80, and I can choose from a wide variety of brands; if this were a proprietary name-brand machine, I'd have to pay €300 per disk, and I'd be at the mercy of the vendor (if he stopped selling the specially tweaked disks required by his server, I'd be out of luck). That's the problem with my HP machine. It still runs great and may continue to do so for a long time, but if it breaks down, there's no way to fix it, as just plugging in commodity parts won't do. Even the memory had to be ordered special. > It's very much like buying the Lexus that comes with the key chip - > you get the extra feature of not being able to start the car without a > key with a chip in it, with the downside that only Lexus supplies the > chipped keys (and charges you up the ass for them of course) Yes. > :-) Actualy I didn't cover that. Manufacturers put these proprietary > things in their server products because they are features that are > very useful to organizations that run hundreds if not thousands of > servers all over the country or the world - with the caveat of course > that every server has to be the same model and come from that same > manufacturer to get the full benefit of the little fancy features. But > to most of us who don't run these large networks, these features do > nothing at best, and are an annoyance at worst. The trend in the IT industry has always been away from proprietary and towards commodity. The fancy little features eventually disappear over time. And they often are not missed. > The HP disk sector atomicity thing was a great feature if you had > disks on an external cabinet that didn't have a UPS on it. Sure, > laugh, but when you have a large HP minicomputer with a disk pack the > size of a refrigerator that has 50 scsi disks in it, that consumes > 15Kw, you don't just go down and grab a UPS from Office Depot. But > naturally for small PC's it was a completely stupid and useless > feature which is why no other disk manufacturer bothered to license > HP's patent on it. What does sector atomicity do? > While I can't of course say that the Adaptec microcode in Anthony's > server was modified to support this particular feature, clearly HP had > some fancy feature support in mind which is why they tampered with the > microcode to begin with. And the sector atomicity thing was not the > only fancy feature that HP put in it's disks back when they were > manufacturing them. This machine was positioned as a workstation, so I don't know that it would contain any of the junk intended for servers (thank goodness). I would have preferred that HP just build a reliable machine, instead of a machine that is custom-built. I guess something like a custom cabinet wouldn't be too bad (which it has), as long as any part that might wear out and fail could still be replaced by commodity parts. -- Anthony
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?1648629793.20050329122346>