This pretty well sums it up. As I said in a post elsewhere, you make a partnership with your users with the common goal of helping them reach their objective as efficiently as possible. With that as a starting point, a lot of the bullshit evaporates, as you working together, not against each other.
Like anything in life, sometimes you have to bend a little to get where you need to go.
It sounds like this guy has had some bad experiences and therefore infers that it is the case for all of IT. I must say I'm a bit a affronted by his sweeping generalisation that IT admin is out of control.
I currently look after a team about 100 developers, testers and technical writers. The situation I walked into was one where each software group had control over hardware budgets, admin, the whole shooting match. Needless to say it was an absolute mess. Permissions were all over the place, NFS and Oracle servers haphazardly brought up, no fault tolerance, poor performance and reliability, the list goes on. I took the approach that my job was to help these folks get software products out the door as efficiently as possible. With that in mind, I sought to make sense of the environment and let the developers write code and the testers test it. They still had some control over day to day stuff, but I manage their total infrastructure for them.
This involved retiring unneeded systems, consolidating storage and servers, upgrading some machines, implementing some proper change management, auditing, performance monitoring, backups, standard system images etc etc. We are not there yet, but needless to say that there has been a significant improvement in the last six months. The coders seem to be happier and the infrastructure is more stable. That sounds like a win-win situation to me.
I think the key for any admin is to make a partnership with your users, understand their needs and address their pain. I agree that some admins rule with an iron fist and have users cowering in their presence. This is counterproductive. But so is having no admin staff in a development centre. If the chance arose, I would actually like to work with the author of this article and perhaps help ease some of that pain:) I deal with folks like him in my work at the moment and I think it gets down to having them trust that you know what you're doing and getting out of their way.
A method I favour on our Unix/Linux network is to use automated installation tools such as Jumpstart (Solaris), Roboinst (IRIX) or KickStart (Linux) to rebuild a system unattended. If we have a root disk go bad on a machine, I tell the system to install over the network, with all of the little config changes in a central set of files and scripted to be put in place. For instance, we turn off Sendmail on most of our machines, so the script handles it. Other systems need particular kernel tweaks for running Oracle etc. The script also handles that. I can get a machine back online from a complete disk failure in about 30 minutes. Of course, if the system is *really* important, we mirror the O/S disks anyway, so there's little or no downtime.
On our desktop PC's we use a variation of this, but with a standard reload CD (using Ghost, I think), which has all the SP's etc etc needed to get a machine up and running quickly. The systems are partitioned into C: and D:, with all user data on D: (or network drives), so it will either be untouched or restorable from our backup system (depending on the severity of the failure). This has worked pretty well with our 50,000+ PC's so far.
There are an awful lot of things we don't get right as well. Our handling of refugee claims is one that I can think of (I don't necessarily disagree with the policy of screening everyone, more on how it was done). Anyway...
The OZ approach is more about making things practical and workable rather than necessarily adhering to some high minded ideal (I think high minded ideals are good things btw). This may have something to do with our history and the harsh conditions both indigenous and settlers had to deal with. We also tend to be rather "no nonsense" kind of people, which folks from other places either find refreshing or rather crass.
Yep, me too. This is how it works where I am and I'm quite happy with it. It gives you a sense of direction in your work and you get rewarded for it come review time....
I guess it has something to do with the fact that MIT is considered one of the U.S.'s best engineering schools and by implication is attended and staffed by some of the best technical minds in the country.
These HP calculators also last forever. Many times I have been sitting at an Engineer/Scientist's desk and seen an ancient HP still working. The $150 is money well spend IMHO. It also harkens back to a time when the name Hewlett Packard was synonymous with top quality. It is less so now...
A bit of both. There is no doubt that the Tagged Command Queing in the SCSI protocol helps enormously with high transaction rates...The drives are often better made as well.
Ah, but it doesn't even out necessarily. You are still dealing with the limitations of the drive/bus technology whether it's RAID'd or not. A stripe of SCSI disks will outperform a stripe of ATA disks. I'll take SCSI or FC over IDE/ATA in most situations for servers (databases, data processing, file serving etc). Workstations are another matter. It may not buy you much more to spend the extra cash.
Take a look at the benchmarks done at: http://www.storagereview.com/comparison.html http://www.storageperformance.org/results.html
or better still, try them out with your own applications. In most cases, I have found that SCSI/FC has spanked IDE and chews up less CPU in doing so.
1, We have several sites with 50-100TB or more of data. How do you propose I back that all up to disk? In fact, most of the data is already on tape in an HSM archive. It gets staged in from tape to disk for processing then released when not needed. There are also multiple processed versions of that data. Disk only solutions start to get expensive at that point.
2, True, but how much space are you going to need to store those multiple versions? In some cases, the changes are small enough that you can store only the increments, but that doesn't cut it for some data sets. Doing this with tape is way cheaper.
3, Fair enough, but drives are most likely to fail when they are started/stopped. You have reduced the reliability by spinning them up and down all the time.
4, But how practical is it *really* to take disks off site? They will need to be stored in a much more controlled environment than tape (in my experience) and are more susceptible to damage in transit.
5, I didn't mention anything about speed. No question, disk is faster, but I maintain tape is more robust. I could say the same re: fire for disks.
Your point about recovering data from disks sounds nice in theory, but wouldn't it be much easier to just grab a tape and restore it? I have been in several situations where tape was superior: retrieving some old field tapes on a VAX, the disk was shot, we replaced it and restore the O/S and data from an old 9 track tape, worked perfectly; the aforementioned 100TB site, where it is simply not practical cost wise to have all that data on spinning disk, the StorageTek Powderhorn(s) (each with 24 3590 drives and 6000 slots) can supply the processing cluster with enough data to keep them running and do it much cheaper than disk; we had a request to restore data from ten years ago to do with a lawsuit, we had the data on an Exabyte 8mm tape, restored perfectly and ours chances of recovering the disk would have been slim.
Oh, and about the obsolete bit...I have had little difficulty finding drives to read 3480, L490E, 3590 (B,E,H), 9 track, 8mm, 4mm, QIC blah blah tapes. One of the nice things is that most manufacturers build some sort of backwards read compatibility into their drives.
Don't get me wrong, disk backups are great and I have made significant use of them in the past (for HSM, DR and Oracle backups in particular) but I am reluctant to give up tape any time soon. For small/medium sized enterprises, disk is probably the way to go, but if you have strict archiving/version requirements I can't see how disk solves this problem adequately.
I can't see disk only backups replacing tape any time soon, for the following reasons:
1, How do take your backups offsite with disks? Sure, you can have an offsite DR centre, but that requires far more expense in terms of floorspace, power etc than a simple bank vault with a box of tapes in it. 2, How do you handle version control with disks? Say I have a customer who wants to back 12 months (or recently in our case *10 years*) for recover a backup. This is much easier with tape or optical, as you can implement a policy of write once, read many or cycling out tapes every month/quarter/year. The cost of doing this with disks would be a lot larger (never mind the fact that you might have difficulty getting the disk online after a long period of time). 3, With disk based backup, you have to keep these disks spinning all the time, thus consuming power, A/C, floor space etc etc. Tapes are way cheaper because you don't have to spin them all the time. You also save on floor space, because you can use a smaller library than you have actual tapes (through a rotation policy). 4, How to backup the backup? That is, what if you have an event that knocks out your backup array? You can potentially lose all the data that's in your "online" backup device. Tapes can avoid this issue because they are off site or not in use. 5, Tapes are lot more robust and have a longer shelf life than disks. I have been able to recover tapes that are 10 or 15 years old with a minimum of fuss. This is a lot more difficult with disks of that vintage.
I think a hybrid disk-tape approach fulfils the need for speedy backup/restore, but addresses the above issues. You can then stage the backups on the disks, and archive them to tape at your leisure. This is the approach that software like SAM-FS takes (small disk cache, big library at the back end). Having said all of that, I do rather like the snapshot and remote backup features that my NetAPP filer has:)
Big Sister is pretty powerful and quite extensible too. Be aware that it takes a non-trivial amount of effort to set up, as I found out. It works on all major O/S flavours though, which is a plus. It also interfaces with other packages, such as OpenView, should you ever need it to.
We are doing a similar evaluation where I work. I think we'll end up with OpenView if the costs work out OK. There are other good commercial solutions on the market, such as Foglight, Storage Profiler, Sun Management Console, Tivoli. It really depends on how much one wants to pay.
The best policy, I think, is to describe the kind of work you are doing, but also mention that NDA prevents you from saying much more. You can impart a fair bit about what you did without giving away the important details.
Often times, honesty will win you points with a potential employer (particularly one is as honest a scrupulous as you are).
...that such an important document in the history of human civiization is #10000. Lovers of freedom and more importantly those who would take it away (are you listening John Ashcroft?) should take the time to read this. For my money, it forms the basis of the common law system that we take for granted in the west today.
Your point if well made, but population density is a different (but related) issue. There is plenty of free space in North America for people to inhabit, so perhaps there should be some encouragment for a more even distribution. This is also the case with other continents (particularly in my home country of Australia). The problem, of course, is that not all of the land is inhabitable. This is why we have people congregating in cities.
I think there will be a move towards less dense areas as time goes on, now that technology and social structures make it easier (I'm thinking of telecommuting and "satellite" cities in particular). The market will have a part to play in this, as people go towards the most attractive areas.
Interestingly enough, there was a program on Discovery-Times called "Nowhere Fast", where all sorts of data and solutions were put forward about the traffic problem.
I personally think that a single approach won't solve the problem. Simply building more roads or buying more busses is not the answer. There has to be a coordinated approach between the public and private sector in terms of town planning, traffic control, etc etc. Where I live, public transport is laughable (there basically isn't any), the city is sprawled out into a mass of little satellite towns (that have grown into each other), the roads are in terrible shape (such that the constant construction slows traffic) and the road system is openly hostile (IMHO) to pedestrians.
I think the solution needs to involve: 1, Better public transport that provides an incentive for people to use it (i.e. cheaper than driving and can get you where you want to go). 2, More HOV lanes on freeways with single occupant cars able to pay a premium (this was one of the good suggestions from "Nowhere fast" 3, More roads linking suburban areas with the fairly typical "one entry - one exit" road systems 4, Better town planning, where you don't have the old style "hub and spoke" system. 5, Revitalising down town areas, to encourage more people to live in inner urban localities, thus reducing the amount of people driving 45-60 minutes on freeways every day. 6, Make cycle paths easier for people to use if they actually want to ride to work.
I used to bitch about the city I am from (Perth) being crap about traffic, but after living in Houston, I doubt I'll ever complain again:)
The two clusters are different enough that making accurate comparisons is difficult. The new G5's have a more recent PCI architecture, they use Infiniband as the interconnect and it's possible that they made use of the AltiVec (though I hear that this may not be the case because of 32 bit limitations). I believe none of these apply to the Xeon's. In high speed computing, the interconnect is vital, so that alone may push this cluster ahead for the time being. I don't doubt that the individual G5 processor are bloody quick (and as a Mac user and fan, I'm kinda glad) though.
See my comment above. I work for Schlumberger and we acquired InsideReality a while ago. The software does pretty much what MagicEarth does. Check it out at:
http://www.sis.slb.com/content/about/campaigns/per formance/insidereality.asp
I think it depends on the situation. I personally prefer fewer systems, but that's just me. A few things to think about:
1, What are the dependencies? That is, do you have an application server that depends on an RDBMS or the like? If so, it might make sense to put them on one machine, assuming you have enough capacity. In a 2-tier system setup, if one server goes down, the service is unavailable anyway. Or you could take those 2 boxes and cluster them for better availability. 2, How much capacity do you need? I have seen a lot of server environments where there were many systems, but most of them were drastically under-utilised. This is most common, IMHO, in Wintel environments, but I have also seen it in Unix shops. 3, Do you need to separate systems for security reasons (e.g. DMZ's for Internet facing apps)? 4, Do you need to charge different clients/departments for the use of these systems? Do these customers insist on buying their own systems and have you manage them? 5, Do you need to support different O/S flavours or different revs of the same O/S? 6, Are the applications likely to consume *all* available resources on a given system? They sometimes warrant their own system or a logical partition of a bigger box (e.g. domain on a Sun system or LPAR on IBMs)
I am currently doing a consolidation project for a large-ish software development and testing team. I'm taking the 19 server environment and bringing it down to about 11 systems, with more to follow when I can move some licenses that are tied to specific nodes. Most of these machines are under-utilised and have direct attached disk spread all over the place. My strategy was to get a NetAPP filer for NFS in the middle and retire all of the old file servers. The next phase will have me retiring four Oracle servers onto one system with SAN storage. So far, so good.
The crew I work for makes some pretty flash software for visualising oil and gas reservoirs in 3D in an immersive environment. It's not exactly the same as a CAVE, but we have variously used active/passive stereo and domes in the past to achieve pretty much that same effects (I'm sure it would work in a real CAVE too).
It never ceases to draw the "oohs and ahhs" from customers. The thing is, they actually find it useful, not just as some pretty eye candy. I guess you'd want to for the price of that SGI Onyx and the Barco projection system required to run it.
On the other hand, some of the most brilliant lecturers I had at university were a little eccentric. Their somewhat unorthodox methods didn't really bother me at the time. I have also worked in the university environment and if you think they're sometimes difficult to deal with as an undergrad, try working for them!!
Electrostatic headphones rule! I have some Stax SR-30's. Love that silky top end and mid range. If you want earth shaking bass, they are not the best choice, but I have speakers for that:)
In the same price range as the aforementioned Sennheisers (I have some of their mid range phones and love 'em), try the Stax Lambda series...
Slashdot Mirror
This pretty well sums it up. As I said in a post elsewhere, you make a partnership with your users with the common goal of helping them reach their objective as efficiently as possible. With that as a starting point, a lot of the bullshit evaporates, as you working together, not against each other.
Like anything in life, sometimes you have to bend a little to get where you need to go.
It sounds like this guy has had some bad experiences and therefore infers that it is the case for all of IT. I must say I'm a bit a affronted by his sweeping generalisation that IT admin is out of control.
:) I deal with folks like him in my work at the moment and I think it gets down to having them trust that you know what you're doing and getting out of their way.
I currently look after a team about 100 developers, testers and technical writers. The situation I walked into was one where each software group had control over hardware budgets, admin, the whole shooting match. Needless to say it was an absolute mess. Permissions were all over the place, NFS and Oracle servers haphazardly brought up, no fault tolerance, poor performance and reliability, the list goes on. I took the approach that my job was to help these folks get software products out the door as efficiently as possible. With that in mind, I sought to make sense of the environment and let the developers write code and the testers test it. They still had some control over day to day stuff, but I manage their total infrastructure for them.
This involved retiring unneeded systems, consolidating storage and servers, upgrading some machines, implementing some proper change management, auditing, performance monitoring, backups, standard system images etc etc. We are not there yet, but needless to say that there has been a significant improvement in the last six months. The coders seem to be happier and the infrastructure is more stable. That sounds like a win-win situation to me.
I think the key for any admin is to make a partnership with your users, understand their needs and address their pain. I agree that some admins rule with an iron fist and have users cowering in their presence. This is counterproductive. But so is having no admin staff in a development centre. If the chance arose, I would actually like to work with the author of this article and perhaps help ease some of that pain
What a brilliant idea. That would also be in keeping with Apple wanting to portray that hipness angle...
A method I favour on our Unix/Linux network is to use automated installation tools such as Jumpstart (Solaris), Roboinst (IRIX) or KickStart (Linux) to rebuild a system unattended. If we have a root disk go bad on a machine, I tell the system to install over the network, with all of the little config changes in a central set of files and scripted to be put in place. For instance, we turn off Sendmail on most of our machines, so the script handles it. Other systems need particular kernel tweaks for running Oracle etc. The script also handles that. I can get a machine back online from a complete disk failure in about 30 minutes. Of course, if the system is *really* important, we mirror the O/S disks anyway, so there's little or no downtime.
On our desktop PC's we use a variation of this, but with a standard reload CD (using Ghost, I think), which has all the SP's etc etc needed to get a machine up and running quickly. The systems are partitioned into C: and D:, with all user data on D: (or network drives), so it will either be untouched or restorable from our backup system (depending on the severity of the failure). This has worked pretty well with our 50,000+ PC's so far.
As an Aussie, thanks for the compliment :)
There are an awful lot of things we don't get right as well. Our handling of refugee claims is one that I can think of (I don't necessarily disagree with the policy of screening everyone, more on how it was done). Anyway...
The OZ approach is more about making things practical and workable rather than necessarily adhering to some high minded ideal (I think high minded ideals are good things btw). This may have something to do with our history and the harsh conditions both indigenous and settlers had to deal with. We also tend to be rather "no nonsense" kind of people, which folks from other places either find refreshing or rather crass.
Yep, me too. This is how it works where I am and I'm quite happy with it. It gives you a sense of direction in your work and you get rewarded for it come review time....
I guess it has something to do with the fact that MIT is considered one of the U.S.'s best engineering schools and by implication is attended and staffed by some of the best technical minds in the country.
These HP calculators also last forever. Many times I have been sitting at an Engineer/Scientist's desk and seen an ancient HP still working. The $150 is money well spend IMHO. It also harkens back to a time when the name Hewlett Packard was synonymous with top quality. It is less so now...
A bit of both. There is no doubt that the Tagged Command Queing in the SCSI protocol helps enormously with high transaction rates...The drives are often better made as well.
Ah, but it doesn't even out necessarily. You are still dealing with the limitations of the drive/bus technology whether it's RAID'd or not. A stripe of SCSI disks will outperform a stripe of ATA disks. I'll take SCSI or FC over IDE/ATA in most situations for servers (databases, data processing, file serving etc). Workstations are another matter. It may not buy you much more to spend the extra cash.
Take a look at the benchmarks done at:
http://www.storagereview.com/comparison.html
http://www.storageperformance.org/results.html
or better still, try them out with your own applications. In most cases, I have found that SCSI/FC has spanked IDE and chews up less CPU in doing so.
OK, so what about:
1, We have several sites with 50-100TB or more of data. How do you propose I back that all up to disk? In fact, most of the data is already on tape in an HSM archive. It gets staged in from tape to disk for processing then released when not needed. There are also multiple processed versions of that data. Disk only solutions start to get expensive at that point.
2, True, but how much space are you going to need to store those multiple versions? In some cases, the changes are small enough that you can store only the increments, but that doesn't cut it for some data sets. Doing this with tape is way cheaper.
3, Fair enough, but drives are most likely to fail when they are started/stopped. You have reduced the reliability by spinning them up and down all the time.
4, But how practical is it *really* to take disks off site? They will need to be stored in a much more controlled environment than tape (in my experience) and are more susceptible to damage in transit.
5, I didn't mention anything about speed. No question, disk is faster, but I maintain tape is more robust. I could say the same re: fire for disks.
Your point about recovering data from disks sounds nice in theory, but wouldn't it be much easier to just grab a tape and restore it? I have been in several situations where tape was superior: retrieving some old field tapes on a VAX, the disk was shot, we replaced it and restore the O/S and data from an old 9 track tape, worked perfectly; the aforementioned 100TB site, where it is simply not practical cost wise to have all that data on spinning disk, the StorageTek Powderhorn(s) (each with 24 3590 drives and 6000 slots) can supply the processing cluster with enough data to keep them running and do it much cheaper than disk; we had a request to restore data from ten years ago to do with a lawsuit, we had the data on an Exabyte 8mm tape, restored perfectly and ours chances of recovering the disk would have been slim.
Oh, and about the obsolete bit...I have had little difficulty finding drives to read 3480, L490E, 3590 (B,E,H), 9 track, 8mm, 4mm, QIC blah blah tapes. One of the nice things is that most manufacturers build some sort of backwards read compatibility into their drives.
Don't get me wrong, disk backups are great and I have made significant use of them in the past (for HSM, DR and Oracle backups in particular) but I am reluctant to give up tape any time soon. For small/medium sized enterprises, disk is probably the way to go, but if you have strict archiving/version requirements I can't see how disk solves this problem adequately.
I can't see disk only backups replacing tape any time soon, for the following reasons:
:)
1, How do take your backups offsite with disks? Sure, you can have an offsite DR centre, but that requires far more expense in terms of floorspace, power etc than a simple bank vault with a box of tapes in it.
2, How do you handle version control with disks? Say I have a customer who wants to back 12 months (or recently in our case *10 years*) for recover a backup. This is much easier with tape or optical, as you can implement a policy of write once, read many or cycling out tapes every month/quarter/year. The cost of doing this with disks would be a lot larger (never mind the fact that you might have difficulty getting the disk online after a long period of time).
3, With disk based backup, you have to keep these disks spinning all the time, thus consuming power, A/C, floor space etc etc. Tapes are way cheaper because you don't have to spin them all the time. You also save on floor space, because you can use a smaller library than you have actual tapes (through a rotation policy).
4, How to backup the backup? That is, what if you have an event that knocks out your backup array? You can potentially lose all the data that's in your "online" backup device. Tapes can avoid this issue because they are off site or not in use.
5, Tapes are lot more robust and have a longer shelf life than disks. I have been able to recover tapes that are 10 or 15 years old with a minimum of fuss. This is a lot more difficult with disks of that vintage.
I think a hybrid disk-tape approach fulfils the need for speedy backup/restore, but addresses the above issues. You can then stage the backups on the disks, and archive them to tape at your leisure. This is the approach that software like SAM-FS takes (small disk cache, big library at the back end). Having said all of that, I do rather like the snapshot and remote backup features that my NetAPP filer has
Solaris also has this, btw, as well as RBAC (Role Based Access Control) among other niceties...
Big Sister is pretty powerful and quite extensible too. Be aware that it takes a non-trivial amount of effort to set up, as I found out. It works on all major O/S flavours though, which is a plus. It also interfaces with other packages, such as OpenView, should you ever need it to.
We are doing a similar evaluation where I work. I think we'll end up with OpenView if the costs work out OK. There are other good commercial solutions on the market, such as Foglight, Storage Profiler, Sun Management Console, Tivoli. It really depends on how much one wants to pay.
The best policy, I think, is to describe the kind of work you are doing, but also mention that NDA prevents you from saying much more. You can impart a fair bit about what you did without giving away the important details.
Often times, honesty will win you points with a potential employer (particularly one is as honest a scrupulous as you are).
...that such an important document in the history of human civiization is #10000. Lovers of freedom and more importantly those who would take it away (are you listening John Ashcroft?) should take the time to read this. For my money, it forms the basis of the common law system that we take for granted in the west today.
Your point if well made, but population density is a different (but related) issue. There is plenty of free space in North America for people to inhabit, so perhaps there should be some encouragment for a more even distribution. This is also the case with other continents (particularly in my home country of Australia). The problem, of course, is that not all of the land is inhabitable. This is why we have people congregating in cities.
I think there will be a move towards less dense areas as time goes on, now that technology and social structures make it easier (I'm thinking of telecommuting and "satellite" cities in particular). The market will have a part to play in this, as people go towards the most attractive areas.
Interestingly enough, there was a program on Discovery-Times called "Nowhere Fast", where all sorts of data and solutions were put forward about the traffic problem.
:)
I personally think that a single approach won't solve the problem. Simply building more roads or buying more busses is not the answer. There has to be a coordinated approach between the public and private sector in terms of town planning, traffic control, etc etc. Where I live, public transport is laughable (there basically isn't any), the city is sprawled out into a mass of little satellite towns (that have grown into each other), the roads are in terrible shape (such that the constant construction slows traffic) and the road system is openly hostile (IMHO) to pedestrians.
I think the solution needs to involve:
1, Better public transport that provides an incentive for people to use it (i.e. cheaper than driving and can get you where you want to go).
2, More HOV lanes on freeways with single occupant cars able to pay a premium (this was one of the good suggestions from "Nowhere fast"
3, More roads linking suburban areas with the fairly typical "one entry - one exit" road systems
4, Better town planning, where you don't have the old style "hub and spoke" system.
5, Revitalising down town areas, to encourage more people to live in inner urban localities, thus reducing the amount of people driving 45-60 minutes on freeways every day.
6, Make cycle paths easier for people to use if they actually want to ride to work.
I used to bitch about the city I am from (Perth) being crap about traffic, but after living in Houston, I doubt I'll ever complain again
The two clusters are different enough that making accurate comparisons is difficult. The new G5's have a more recent PCI architecture, they use Infiniband as the interconnect and it's possible that they made use of the AltiVec (though I hear that this may not be the case because of 32 bit limitations). I believe none of these apply to the Xeon's. In high speed computing, the interconnect is vital, so that alone may push this cluster ahead for the time being. I don't doubt that the individual G5 processor are bloody quick (and as a Mac user and fan, I'm kinda glad) though.
See my comment above. I work for Schlumberger and we acquired InsideReality a while ago. The software does pretty much what MagicEarth does. Check it out at:r formance/insidereality.asp
http://www.sis.slb.com/content/about/campaigns/pe
I think it depends on the situation. I personally prefer fewer systems, but that's just me. A few things to think about:
1, What are the dependencies? That is, do you have an application server that depends on an RDBMS or the like? If so, it might make sense to put them on one machine, assuming you have enough capacity. In a 2-tier system setup, if one server goes down, the service is unavailable anyway. Or you could take those 2 boxes and cluster them for better availability.
2, How much capacity do you need? I have seen a lot of server environments where there were many systems, but most of them were drastically under-utilised. This is most common, IMHO, in Wintel environments, but I have also seen it in Unix shops.
3, Do you need to separate systems for security reasons (e.g. DMZ's for Internet facing apps)?
4, Do you need to charge different clients/departments for the use of these systems? Do these customers insist on buying their own systems and have you manage them?
5, Do you need to support different O/S flavours or different revs of the same O/S?
6, Are the applications likely to consume *all* available resources on a given system? They sometimes warrant their own system or a logical partition of a bigger box (e.g. domain on a Sun system or LPAR on IBMs)
I am currently doing a consolidation project for a large-ish software development and testing team. I'm taking the 19 server environment and bringing it down to about 11 systems, with more to follow when I can move some licenses that are tied to specific nodes. Most of these machines are under-utilised and have direct attached disk spread all over the place. My strategy was to get a NetAPP filer for NFS in the middle and retire all of the old file servers. The next phase will have me retiring four Oracle servers onto one system with SAN storage. So far, so good.
The crew I work for makes some pretty flash software for visualising oil and gas reservoirs in 3D in an immersive environment. It's not exactly the same as a CAVE, but we have variously used active/passive stereo and domes in the past to achieve pretty much that same effects (I'm sure it would work in a real CAVE too).
It never ceases to draw the "oohs and ahhs" from customers. The thing is, they actually find it useful, not just as some pretty eye candy. I guess you'd want to for the price of that SGI Onyx and the Barco projection system required to run it.
...and to get more research money :)
On the other hand, some of the most brilliant lecturers I had at university were a little eccentric. Their somewhat unorthodox methods didn't really bother me at the time. I have also worked in the university environment and if you think they're sometimes difficult to deal with as an undergrad, try working for them!!
Electrostatic headphones rule! I have some Stax SR-30's. Love that silky top end and mid range. If you want earth shaking bass, they are not the best choice, but I have speakers for that :)
In the same price range as the aforementioned Sennheisers (I have some of their mid range phones and love 'em), try the Stax Lambda series...
Definitely not. Analogue all the way for me. Preferably Swiss too. They look cooler for a start (never really was into digital timepieces...)