On Thu, 2003-06-05 at 15:43, William Page wrote: > Have you considered Network-Attached Storage? I've not got a whole lot of > knowledge in that specific area, but I think the idea is to have your > network storage independent of your servers - "It just works".
NAS is really cool. But in the final analysis it's really just some sort of NFS server, or it's basically a huge logical disk that looks like scsi to the server (fiber channel). The benefit is nice logical volume management build into the device. You still have to implement a file system on top of it (say ext3 or XFS) and when you add disks you have to stretch the partition across it. The NAS does take care of RAID and mirroring and rebuilding of individual disks and in many cases has a built-in backup system. The constraints are still bus speed and the speed of the disks. If everything gets obsolete in 3 years, then we'll be replacing our servers and NAS every three years anyway, at least in the sizes we are using (1 TB). In a data center where they use hundreds or thousands of TB, that's another story and the only way to go is with the NAS. Anyway, storage and backup is the problem of the next decade, in my opinion. Michael > > William > > > -----Original Message----- > > From: [EMAIL PROTECTED] [mailto:[EMAIL PROTECTED] > > On Behalf Of Evan McNabb > > Sent: Thursday, June 05, 2003 1:53 PM > > To: BYU Unix Users Group > > Subject: Re: to MP or not MP? that is the question (was [uug] NFS vs > > SAMBA) > > > > Soren and I are buds so I don't want him to think I'm attacking him and > > CADEM... :-) > > > > The CS dept network and the CADEM labs are not the same. We each have > > different > > classes to cater to and problems to worry about. Being a sys admin for two > > years > > I've heard plenty of complaints about BOTH the CS and CADEM setups. > > > > If the students think this problem with our file server has caused them > > problems, it > > has been literally a 500x worse than us. We can't figure out what the > > exact source > > of the problem is, nor the many people who have helped us look into the > > issue. Besides > > this problem (and it is a major one that will be solved within a few > > weeks), I haven't > > heard of any recent problems (ok, well Windows profiles suck too, but > > Byron fixed that). > > > > I respect the CADEM admins for their work and I believe they have a good > > setup. But > > since we have different needs, we do things differently. After a lot of > > experience > > we have found that one fileserver is what we need; we could do it with a > > cluster, but > > it has overall worked well with the single server (besides this problem > > we've had for > > the last 6 months with IO). > > > > I think think it's funny how people (users) complain about problems. Until > > you've been a > > sys admin over a large network (in our case 200 linux boxes, 150 windows, > > ~7 - 8 > > servers) you really don't know about how complicated things get. People > > come in > > to the office and complain about our samba setup and I see them post > > questions to > > the list a week later because they can't even get two machines to talk > > together. There > > is a HUGE difference between admining 5 and 50 computers. > > > > Also, I know that as a student you pay for computers to use and those > > should > > be maintained well. Just to let you know, we really do care that the boxes > > and networks > > are running smoothly. You see, sys admins do care. :-) > > > > I'm not trying to bash anyone, just reminding people to have constructive > > criticism > > and remember how hard it can be. I don't give OIT as hard of a time as I > > used to > > since I've personally had to deal with serious issues. > > > > Well, my rant is over. Kudos to the world's sys admins, and good work guys > > in the CADEM. > > > > -Evan > > > > P.S. Beware of the L.A.R.T :-) > > http://bofhcam.org/co-larters/lart-reference/ > > > > > So? The same thing happens when a monolithic server fills up. I > > > guarantee you that all the space that the new CS server has won't last > > more > > > than 4 years; they will eventually have to get a new one, or an > > > additional one. Building a modular system at the beginning forces you > > > into the plan-for-expansion mindset. If you're starting with 6 NFS > > > servers, it's easy to design the system to accomodate 8 or 20. However, > > > if you start with the mindset of having only one server, then try as you > > > might, there will be some design fallacies that make migrating to 2 or > > > more difficult, if not impossible. Nobody takes the time to set up a > > > load-balancing system for just one webserver, and always ends up kicking > > > themselves when they have to re-design their network when adding the > > > second server becomes necessary. > > > > > > On top of that, there's the financial aspect. A lot of small servers is > > > almost always cheaper than one big server, and believe it or not, > > > they're almost always easier to maintain. > > > > -- > > /********************************************************************\ > > Evan McNabb: <[EMAIL PROTECTED]> <[EMAIL PROTECTED]> > > http://evan.mcnabbs.org > > System Administrator, CS Department, BYU > > GnuPG Fingerprint: 53B5 EDCA 5543 A27A E0E1 2B2F 6776 8F9C 6A35 6EA5 > > \********************************************************************/ > > > ____________________ > BYU Unix Users Group > http://uug.byu.edu/ > ___________________________________________________________________ > List Info: http://uug.byu.edu/cgi-bin/mailman/listinfo/uug-list -- Michael L Torrie <[EMAIL PROTECTED]> ____________________ BYU Unix Users Group http://uug.byu.edu/ ___________________________________________________________________ List Info: http://uug.byu.edu/cgi-bin/mailman/listinfo/uug-list
