[email protected] writes: > At $WORK, I've got a lot of Dell equipment and some Supermicro (from Silicon > Mechanics, actually). Overall, the performance, build quality, and cost (with > the $LARGE.EDU discount) are very comparable. There are some noticeable > differences, which I think have to do with scale. The following are > anecdotes, meant to be representative, rather than any kind of formal data. > > One of the SM boxes (with an Nvidia GPU) has vents on the top & bottom of the > case, as well as at the rear. This led me to the conclusion that the box > needed free space in the rack above & below for cooling. I brought this up w. > SM tech support...and after several calls/email and days of delays, was told > that this was unnecessary. My experience with the Dell world suggests that > this kind of cooling requirement would be documented much more clearly, and > that the answer would have been delivered more quickly.
Hm. well, > => > => For someone who likes grubbing around inside computers, who has been > => doing so for the last fifteen years, the support offered by dell > => and hp has negative value. If it's something obvious, sure, dell > > Hmmm...I like grubbing around inside computers (and have been for, oh, about > 28 years). I'd much rather deal with the inside of a Dell server than most of > the smaller companies' products. For example, the Dell cases don't have > screws (or they are captive), and the rack slide mechanisms are much nicer. > Within the case, things that must be removed (fan shrouds or cables, for > example) usually are tool-less and have color-coded latches. Small points, > but when you've got to open 50+ servers in sequence to upgrade RAM, it > becomes important, and it's a real convenience when you've got several > generations of boxes to know that you just look for the blue plastic for an > indication of where different parts are secured. Eh, my preference is still for supermicro. You can take supermicro chassis for the last 10 years or so and exchange chassis parts. (I mean, you probably don't want to do this with power supplies, but with the other chassis parts? it's pretty nice that it's all compatable.) > => or hp can come out and fix it for you. But you've got to sit on hold, > => take the guy through troubleshooting, etc... and in that time, I > => could have just swapped the goddamn power supply myself. it's easy. > > Huh? I'm saying that usually I can swap out a power supply faster than I can convince a dell/hp tech that the power supply is bad and it needs to be swapped out. Not that either one is an impossible task, but one requires scheduling and granting someone access, etc... and the other is five minutes of work. > => Now, if I didn't have that experience, sure, it'd be different, talking > => to dell, I'm sure, would be easier than doing it myself and getting > => dust all over my nice suit. > => > => Now, when you get a /hard/ problem; something where you are not > => entirely sure if it's hardware or software, for me, the value of > => support swings massively to the negative. > => > => Do you know how hard it is to get dell to replace a maybe flaky > => motherboard? it takes a hell of a lot more than $250 worth of > => effort, even when it's only a $100 motherboard. > => > > That's almost completely the opposite of my experience, and may depend > largely on the level of service contract and the size of your organization. This is quite possible. It wouldn't suprise me if the organizations I worked for had low-end service contracts to save money. My understanding was that the good service contracts come close to doubling the cost of the server. > Our 'critical' equipment has enterprise class contracts (24x7, 4hr parts & > technician dispatch, etc), and $WORKPLACE is a small dept in a large > university. My calls with Dell often go something like: > > Me: I'm running Linux and getting IPMI errors on DIMM 3A. It's a 4GB > DIMM. I know that you > want me to swap that DIMM with another one in the same machine > to see if the problem > is with the DIMM or the memory riser, but I want to limit the > downtime on this production > machine. Please send me a new DIMM and when it arrives, I'll > swap them, run memory > diagnostics, and send back the failed DIMM instead of having > downtime now and > again when the replacement arrives. > > Dell: OK. We've got XYZ as your shipping address. The part will be > there by 9AM tomorrow. Please > follow-up with me when you swap the memory and I'll keep the > ticket open in case that > doesn't solve the issue. Yeah, sure, if it's clear what the problem is, it's trivial. But that's true even without a support contract. In my case, I swap the dimm with one in the spares bin, bring the thing back up, and then at my leisure test the bad dimm. (It's ecc, and it sounds like single bit-errors, so even if I'm wrong and it was the memory socket, the box isn't down) > => It's worse for most peripherals. When a drive fails a SMART test, > => I return it to western digital. They send me a refurb. everyone > => is happy, and minimal effort is expended. When a drive fails > => a SMART test but still appears to work in a dell? I have to > => explain to three different idiots what "SMART" is and why that > => means the drive should be replaced. > > See above for my experience with RAM... My experience has been that if something failed POST or gave me a failure in the BIOS logs (like single bit ecc errors on most ram does) getting a replacement wasn't so difficult. A failed SMART test on a drive? much more difficult, as that isn't in the BIOS logs (unless it's status FAILED but my policy is to return a drive when it gives a 'failed on read error' to /any/ SMART test.) > As a side note, I've got very mixed feelings about 'returning' bad drives to > the vendor/manufacturer. I don't return 'failing' drives unless: > > the vendor will send written certification of secure deletion (physical > destruction or an > acceptable method of overwriting data) > > the drive was part of a large striped array (personally, if the drive > fails and was > one disk in a large RAID-5 or -6 stripe, I think that data recovery > possibilities are very, > very low) > > I can securely erase the drive in advance (in which case it's probably > functioning just fine) how many unrecoverable errors do you tolerate before returning a drive? a single unrecoverable read error is essentially a remappable bad sector, but, on the other hand, you've lost the data you were trying to read. (bad sectors that you find when you write are much less bad, as the disk just writes the data elsewhere. But if it can't read the data, if it's not in drive ecc, it's gone.) A RAID will fail the drive after one unrecoverable read. For me, that's enough to say the drive is bad. > Depending on what was stored on the drive, sometimes we'll simply eat the > cost of a replacement even if > the warranty is in effect. One option is to purchase the vendor's "you keep > the drive" warranty (offered by Dell, I don't know about others), so that we > get the replacement and keep the failed disk. How do you destroy the drive? Obviously, just dumping the drives in the garbage is worse than sending 'em in. > => > => On top of that, dell shortens the warranty on both disks and ram, > => the parts most likely, in my experience, to go bad. > => > > I realize that this message makes me sound like a Dell fanboy...I'm not, but > overall, I'm surprised that I like their stuff as much as I do...there are > plenty of things about them not to like, and there are some product lines > that I won't buy from them at all. I think that a lot of my satisfaction > comes from their behavior based on my department's (and university's) > purchase volume. They treat me well because there's significant money at > stake. If I was purchasing 10 servers/year, I probably wouldn't have the same > experience. I think that the same could be said of almost any vendor. Eh, most of my dell experience comes at a place that bought on the order of 250 (rather large) servers a year. They paid top dollar, as far as I could tell, much more than my smaller consulting clients, in spite of talking up how hard they negotiate. _______________________________________________ Tech mailing list [email protected] https://lists.lopsa.org/cgi-bin/mailman/listinfo/tech This list provided by the League of Professional System Administrators http://lopsa.org/
