I've done a similar setup at my house for probably similar reasons - solves the problem of friends bringing over their nappy computers with ancient video cards and me having to upgrade that and get it all running every lan party.
I used opensolaris (and now openindiana) for the back end server. It has lots of ram and some SSDs for the l2arc cache so most things end up being cached. I use zfs snapshots for the clone systems.
This works well, and performance falls somewhere between an SSD and a Hard drive for gaming load times... with one exception.. boot times for the diskless systems is horrible. Like 2 minutes each.
It looks like you are booting your windows off VHD and putting the differential on the local SSD? Any benefits of doing that over just running everything from the backend server?
I'm thinking about getting an SSD for mine and using that for the boot partition of the systems, and then having steam on an iscsi mount off the backend server. That should give great performance, and fast boot times. But I wouldn't be able to rollback the OS to the master state like I can do now.
People don't do enough research and buy the wrong stuff. If you need fast writes, you get the Intel X25-E. If you need fast reads, the Intel X25-M is fine. If you need the SSD to take a punishing amount of writes over the years, you get the X25-E again. If you aren't planning on punishing it with writes, the X25-M is again fine. If you need cheap, then you get an extra helping of crappy write I/O.:-)
If you want to have monstrously fast storage, you build a raid with zfs, and use one or two X25-Ms for the L2ARC cache and a mirrored (through zfs) X25-M pair for the zil cache.
I'd rather use SSD to supplement traditional storage rather than to run straight off it. But that said, I have been running my mythtv box off of an 8GB Transcend compact flash for the OS for over two years now with great results. It is a gentoo system, so I compile stuff on it all the time. Of course it has a mysql database that gets updated daily because of the schedules it pules down for mythtv. No problems there, and no regrets.
But more to your point, the problems that plagued crappy SSD controllers and designs are being worked out, and probably somewhat soonish won't be a relevant issue anymore for most people.
SSDs are tiny compared to the spinning rust variety.
But here is where ZFS kicks butt. You can attach SSDs to be the read and/or write cache for your large array of magnetic disk. You use an MLC SSDs for read cache, and a SLCs for write cache. If you do this you effectively turn your big slow array into crazy crazy fast storage.
I set up a opensolaris/zfs setup like this at home. I connect to it from my gaming pc, and have my apps installed on an iscsi target on opensolaris through regular gigabit. Do you know how crazy fast games launch? All games? It is like I am playing them all from a ram drive. I probably can't express how awesome it is.
Also, from a recent conference on ZFS it looks like they will have the ability to increase/decrease widths of strips, encryption, and single instance storage soon. Single instance storage will be great for me since my home systems get their disk from the server like I mentioned about.
It is my understanding (I read it on the internet somewhere) that every flash device has some form of wear leveling built in, except for the actual raw flash chips. So if you solder flash chips onto some device, you'll need to format those flash chips with jffs2 or similar because jffs2 will perform its own wear leveling.
As for compact flash as a hard drive, I have been using an 8GB Transcend 266x CF connected to an addonics CF->Sata adapter for use as the OS drive in my gentoo based mythtv system. Man, it really kicks butt.
I've only been running it for about 10 or 11 months now, but so far I've had no problems. The mysql database mythtv uses gets updated all the time. Since it is a gentoo system and I like to keep it up to date, the CF sees a lot of compiling action.
Speaking of which, having portage run off the flash has sped up my compiles way more than distc or ccache ever did for me. Or put another way, for compiling from, the flash drive is a godsend.
Performance is good as I get a consistent 40MB a second for sequential reads, and a consistent 34MB a second for sequential writes.
The article talks about large solid state drives, but because of the price premium, I've been experimenting with smaller SSDs. In particular, I've been using an 8GB 266x CF card coupled with a CF->SATA adapter as the OS drive for my mythtv system for 5 months now with great success.
Not only is the flash drive completely silent, it is reasonably fast. Reads always benchmark at 40MB a second and writes benchmark at 34MB a second.
I've been a bit worried about the flash wearing out after repeated writes, but so far so good. Since my mythtv mysql installation is stored on it, as well as the normal system log files, I'm sure it sees quite a lot of action.
But to my point...
One common problem with systems such as mythtv that are under heavy IO stress is that during these moments of stress (lots of recordings going on at once) the whole operating system grinds to a halt or at least becomes sluggish waiting on some needed IO.
It was very common on my old mythtv setup where I used the extra space on the OS hard drives as extra storage space for mythtv recordings. I'm not experiencing any of that sluggishness with the new setup.
This has got me thinking that for my future desktop system, maybe instead of getting a raptor for the OS drive, and a large, slower hard drive for the rest of my stuff in order to minimize IO bottlenecks, I should swap out the raptor for a 16GB SSD for the OS drive. I'd end up with something that has almost no latency, good speed, silent, and it may be possibly just as reliable in that role.
I can see where the author is coming from. I've replaced my Redhat 6.2 colo server with Gentoo 2003.0 4 years ago and have not looked back. This server has a couple hundred thousand uniques visit various hosted domains every month. It isn't mission critical, but it is important enough that if it goes down I get calls nearly right away from people wondering what happened to their stuff. (I've been very unlucky with my choice of hard drives)
I had two big problems with using RedHat.
One, it was amazingly incredibly out of date after a year. When too many newer packages wouldn't install on the system, I'd update to a newer redhat, but then I'd have to take huge amounts of time reapplying and checking all the various customizations I had made to the previous install. I hated doing that.
Two, RedHat did not provide RPMS for what seemed like most applications that I wanted on the server. This invariably meant that I was compiling what I wanted from sources, and so I had to deal with library hell anyway. There are other repositories for packages now like via Yum... but the last time I checked (and this has been a while ago) most of the obscure packages I like to install were not in the repository, but they are in portage.
Anyway, in those 4 years of running Gentoo on the server, I have had a couple of traumatic experiences.
One of my profiles was deprecated and I had to switch to a new updated one, and there was a good deal of configuration breakage.
Some time ago, the gentoo folks decided to change they way apache behaved and since I wasn't diligently keeping up with new changes, my apache server was down for a while while I figured out what happened. There was good documentation, I just didn't pay attention.
The gentoo folks switched from using Xfree86 to Xorg at some point, and I spent quite a long bit of time recompiling everything X and getting that sorted out.
I also got bitten by mysql upgrade issues when going from 4 to 5 that brought down my service for a while.
One annoying thing that happens fairly often is a lot of peoples websites break when I upgrade php. But I at least have it down pat enough to force upgrade several users photo galleries when I upgrade.
Besides for hardware issues, those have been my biggest issues while running gentoo on a server these past 4 years.
The benefits I get are of course being able to easily install pretty much any open source application I want which is just killer. Because I have a multiuser server, I also originally set up gentoo to be hardened. I had enough issues with lame users trying to get root with their shell account that I truly appreciate the benefits of a hardened install.
I do know that it is one thing to use gentoo on one server you pay a lot of attention to rather than deploying it on a lot of other servers that you probably don't want to pay so much attention to. There are tools and procedures in place to make multiple gentoo deployments reasonable. Building binary packages on a development system and deploying them to the production systems would be a must. I do that a lot on my home systems and it works well.
Oh yeah, I've been running on unstable this entire time. When I first set up the system I didn't realize you could run stable and then grab a few unstable packages. Oops.
It's not just the internet. Have you ever looked at North Korea using Google maps nighttime? North Korea is the black patch to the left of Japan. It is more amusing if you switch to "Dusk Map" as you can clearly see that the lack of lights match exactly with the boarders of South Korea and China.
Man, sucks to be them. My guess is the lack of electricity in the country is some sort of ploy to confuse all of our advanced weapons and smart bomb technology.;-)
It is also worth checking out Afghanistan and Mongolia at night. From looking at their night time maps, I admit that I am just AMAZED at how awesome their energy conservation programs are. California could learn a lot from Afghanistan for sure. And Mongolia better not give the US any lip.
And if you are looking at the map, check out how well lit Iran is. I don't know about you, but with the amount of bright lights all over that country, I'm guessing the US wouldn't hit that. I think we like our bitches more backwards and with a southern accent.:-)
I've been running DD-WRT V23 Final since right after Christmas, and the only time it has gone down on me was due to a power outage in my area a few months ago. Otherwise, it has been rock solid stable. I always had to reboot my WRT54g every week when using the linksys firmware... especially when I was downloading torrents and stuff. If the router didn't slow to a crawl, the wireless link would totally quit working until I rebooted the unit. I'm even using QOS, PPTP, and a few of the other enhancements that linksys didn't provide with their bum firmware.
All that is a thing of the past. In fact, here's what my router says now:
~ # date
Mon Jun 26 15:00:10 UTC 2006
~ # uname -a
Linux cerberus 2.4.32 #431 Sun Dec 25 16:58:55 UTC 2005 mips unknown
~ # uptime
14:52:33 up 100 days, 1:58, load average: 0.00, 0.00, 0.00
I realize that mediawiki is the current myspace of wikis, but it is not really ment to be deployed in a corporate environment. Most corporations would need to use a wiki with more access controls like TWiki or confluence.
I've heard that a while ago, some folks inside Intel set up a mediawiki site for internal documentation, and when the lawyers heard about it, they had the project shut down. There was too much liability I gather.
Anyway, I've set up a TWiki installation at my work three years ago now and am not totally sure if I would call the project a success story.
First off, in order for management to buy in, we needed to demostrate that we could lock down sections and properly secure the site. With TWiki it was easy to create 'Webs' with different access permissions.
Second, in order to get users to buy in, we needed a reason for them to go to the site. For example, they go to the site and find useful content, and so therefore get an idea that the wiki is useful for them. When starting out, there is no useful content in a wiki. Two or three of us have been dutifully putting in what we know into the site, but it has taken maybe two years of effort before there is enough content for other people to find the site semi worthwhile.
It would be better if everyone just 'got it' and contributed to the site right away.
Another issue is the employees that know a lot but like to horde information. Having a nice collaboration tool around will not get them to release nugets of information they certainly have.
Some people insist on publishing perfect documents to the wiki. Those kinds of documents are nearly impossible to come up with on the first pass, and so a lot of documentation that is partially captured and written up has never made it to the site. I'm sure this issue is partly my fault as I haven't impressed upon the users that it is ok ot put up half ass content into a wiki, and then work on it later to polish it up if it would be useful.
One more problem we've had is that most of the employees think that documents in the wiki are owned by the creator of the document and not to be touched by them. I know I have explained about colaboration and cooperative editing a lot, but still, most people will not fix minor errors in documents created by other people. I have had people come up to me and tell me that there is an error in a document I've written in the wiki, or some have sent me emails telling me so. I have demonstrated that they should have edited the page themselves and fixed the error, and now a few of my fellow employees are more willing to do that.
Anyway, even dispite all of these hardships and difficulties, I think the wiki installation is invaluable, and at least for the few of us that 'get it' with regards to wikis, the site has made us maybe 20-30% more productive then we were before we had it around.
The reason there is one torrent download instead of two has more to do with managability issues than anything else. Over the lifetime of this torrent (which could be years) it will be easier to manage and maintain as one file instead of two. It will be easier for seeders to seed one file instead of two. And I would think it would be easier for users to download just one file instead of two.
If you wanted to download just the mp3s you could use one of the many mirrors listed on the doom site.
If you wanted to download the flac version so that you could burn them to CD, what is the big deal with downloading an extra 134MB of mp3 data along the way? You could always move the mp3s onto your iPod or whatever anyway, or just trash them.
Of course this makes me look insensitive to the modem users out there... but if you have broadband, this is just not a big deal.
I've wondered about this kind of thing as well. The one thing that the RDP based implementations seem to have over standard X versions is that they can pipe the sound over along with display.
Is there any way to get full sound support when connecting remotely to your (flavor of) unix server? Preferably a free/open solution...
I'm rolling out an openvpn 2 setup now, and I have to say I'm quite impressed by the package. It seems very stable, gives good performance, has clients for everything we are going to use, and is open source.
The big reason why I chose openvpn over other solutions like IPSEC was basically because I couldn't find free IPSEC clients for windows.
If I had any advice for someone setting up openvpn, it would be to figure out what kind of a setup you want before you try to implement it. There are so many options and possibilities that you probably won't need. I had trouble keeping everything straight before I diagramed what I wanted the VPN setup to look like in the end (routed/nat/two factor authentication/etc)
I've actively been using the bayesian filter that Mozilla comes with for a bit over a year now. Although it seemed to take forever to get 'trained' to what I consider spam, I've found that it works exceptionally well, maybe mismarking a legitimate email to me probably less than a dozen times so far (after the initial round of training).
Maybe six months ago, I noticed I was receiving quite a lot of these hash busting spams and I was bummed that maybe the bayesian filter wasn't the be all end all of spam filters.
But I pressed on using it, and in time, almost all of the hash busting emails are again getting filtered as spam.
I'd guess there are only so many different ways people can write Vi@gra and still have it be readible...
Slashdot Mirror
I've done a similar setup at my house for probably similar reasons - solves the problem of friends bringing over their nappy computers with ancient video cards and me having to upgrade that and get it all running every lan party.
I used opensolaris (and now openindiana) for the back end server. It has lots of ram and some SSDs for the l2arc cache so most things end up being cached. I use zfs snapshots for the clone systems.
This works well, and performance falls somewhere between an SSD and a Hard drive for gaming load times... with one exception.. boot times for the diskless systems is horrible. Like 2 minutes each.
It looks like you are booting your windows off VHD and putting the differential on the local SSD? Any benefits of doing that over just running everything from the backend server?
I'm thinking about getting an SSD for mine and using that for the boot partition of the systems, and then having steam on an iscsi mount off the backend server. That should give great performance, and fast boot times. But I wouldn't be able to rollback the OS to the master state like I can do now.
Yeah, so what did you do software wise?
People don't do enough research and buy the wrong stuff. If you need fast writes, you get the Intel X25-E. If you need fast reads, the Intel X25-M is fine. If you need the SSD to take a punishing amount of writes over the years, you get the X25-E again. If you aren't planning on punishing it with writes, the X25-M is again fine. If you need cheap, then you get an extra helping of crappy write I/O. :-)
If you want to have monstrously fast storage, you build a raid with zfs, and use one or two X25-Ms for the L2ARC cache and a mirrored (through zfs) X25-M pair for the zil cache.
I'd rather use SSD to supplement traditional storage rather than to run straight off it. But that said, I have been running my mythtv box off of an 8GB Transcend compact flash for the OS for over two years now with great results. It is a gentoo system, so I compile stuff on it all the time. Of course it has a mysql database that gets updated daily because of the schedules it pules down for mythtv. No problems there, and no regrets.
But more to your point, the problems that plagued crappy SSD controllers and designs are being worked out, and probably somewhat soonish won't be a relevant issue anymore for most people.
SSDs are tiny compared to the spinning rust variety.
But here is where ZFS kicks butt. You can attach SSDs to be the read and/or write cache for your large array of magnetic disk. You use an MLC SSDs for read cache, and a SLCs for write cache. If you do this you effectively turn your big slow array into crazy crazy fast storage.
I set up a opensolaris/zfs setup like this at home. I connect to it from my gaming pc, and have my apps installed on an iscsi target on opensolaris through regular gigabit. Do you know how crazy fast games launch? All games? It is like I am playing them all from a ram drive. I probably can't express how awesome it is.
Also, from a recent conference on ZFS it looks like they will have the ability to increase/decrease widths of strips, encryption, and single instance storage soon. Single instance storage will be great for me since my home systems get their disk from the server like I mentioned about.
It is my understanding (I read it on the internet somewhere) that every flash device has some form of wear leveling built in, except for the actual raw flash chips. So if you solder flash chips onto some device, you'll need to format those flash chips with jffs2 or similar because jffs2 will perform its own wear leveling.
As for compact flash as a hard drive, I have been using an 8GB Transcend 266x CF connected to an addonics CF->Sata adapter for use as the OS drive in my gentoo based mythtv system. Man, it really kicks butt.
I've only been running it for about 10 or 11 months now, but so far I've had no problems. The mysql database mythtv uses gets updated all the time. Since it is a gentoo system and I like to keep it up to date, the CF sees a lot of compiling action.
Speaking of which, having portage run off the flash has sped up my compiles way more than distc or ccache ever did for me. Or put another way, for compiling from, the flash drive is a godsend.
Performance is good as I get a consistent 40MB a second for sequential reads, and a consistent 34MB a second for sequential writes.
The article talks about large solid state drives, but because of the price premium, I've been experimenting with smaller SSDs. In particular, I've been using an 8GB 266x CF card coupled with a CF->SATA adapter as the OS drive for my mythtv system for 5 months now with great success.
Not only is the flash drive completely silent, it is reasonably fast. Reads always benchmark at 40MB a second and writes benchmark at 34MB a second.
I've been a bit worried about the flash wearing out after repeated writes, but so far so good. Since my mythtv mysql installation is stored on it, as well as the normal system log files, I'm sure it sees quite a lot of action.
But to my point...
One common problem with systems such as mythtv that are under heavy IO stress is that during these moments of stress (lots of recordings going on at once) the whole operating system grinds to a halt or at least becomes sluggish waiting on some needed IO.
It was very common on my old mythtv setup where I used the extra space on the OS hard drives as extra storage space for mythtv recordings. I'm not experiencing any of that sluggishness with the new setup.
This has got me thinking that for my future desktop system, maybe instead of getting a raptor for the OS drive, and a large, slower hard drive for the rest of my stuff in order to minimize IO bottlenecks, I should swap out the raptor for a 16GB SSD for the OS drive. I'd end up with something that has almost no latency, good speed, silent, and it may be possibly just as reliable in that role.
What do you think?
I can see where the author is coming from. I've replaced my Redhat 6.2 colo server with Gentoo 2003.0 4 years ago and have not looked back. This server has a couple hundred thousand uniques visit various hosted domains every month. It isn't mission critical, but it is important enough that if it goes down I get calls nearly right away from people wondering what happened to their stuff. (I've been very unlucky with my choice of hard drives)
I had two big problems with using RedHat.
One, it was amazingly incredibly out of date after a year. When too many newer packages wouldn't install on the system, I'd update to a newer redhat, but then I'd have to take huge amounts of time reapplying and checking all the various customizations I had made to the previous install. I hated doing that.
Two, RedHat did not provide RPMS for what seemed like most applications that I wanted on the server. This invariably meant that I was compiling what I wanted from sources, and so I had to deal with library hell anyway. There are other repositories for packages now like via Yum... but the last time I checked (and this has been a while ago) most of the obscure packages I like to install were not in the repository, but they are in portage.
Anyway, in those 4 years of running Gentoo on the server, I have had a couple of traumatic experiences.
One of my profiles was deprecated and I had to switch to a new updated one, and there was a good deal of configuration breakage.
Some time ago, the gentoo folks decided to change they way apache behaved and since I wasn't diligently keeping up with new changes, my apache server was down for a while while I figured out what happened. There was good documentation, I just didn't pay attention.
The gentoo folks switched from using Xfree86 to Xorg at some point, and I spent quite a long bit of time recompiling everything X and getting that sorted out.
I also got bitten by mysql upgrade issues when going from 4 to 5 that brought down my service for a while.
One annoying thing that happens fairly often is a lot of peoples websites break when I upgrade php. But I at least have it down pat enough to force upgrade several users photo galleries when I upgrade.
Besides for hardware issues, those have been my biggest issues while running gentoo on a server these past 4 years.
The benefits I get are of course being able to easily install pretty much any open source application I want which is just killer. Because I have a multiuser server, I also originally set up gentoo to be hardened. I had enough issues with lame users trying to get root with their shell account that I truly appreciate the benefits of a hardened install.
I do know that it is one thing to use gentoo on one server you pay a lot of attention to rather than deploying it on a lot of other servers that you probably don't want to pay so much attention to. There are tools and procedures in place to make multiple gentoo deployments reasonable. Building binary packages on a development system and deploying them to the production systems would be a must. I do that a lot on my home systems and it works well.
Oh yeah, I've been running on unstable this entire time. When I first set up the system I didn't realize you could run stable and then grab a few unstable packages. Oops.
-Supertux
Man, sucks to be them. My guess is the lack of electricity in the country is some sort of ploy to confuse all of our advanced weapons and smart bomb technology. ;-)
It is also worth checking out Afghanistan and Mongolia at night. From looking at their night time maps, I admit that I am just AMAZED at how awesome their energy conservation programs are. California could learn a lot from Afghanistan for sure. And Mongolia better not give the US any lip.
And if you are looking at the map, check out how well lit Iran is. I don't know about you, but with the amount of bright lights all over that country, I'm guessing the US wouldn't hit that. I think we like our bitches more backwards and with a southern accent. :-)
All that is a thing of the past. In fact, here's what my router says now:
-SuperTux
I realize that mediawiki is the current myspace of wikis, but it is not really ment to be deployed in a corporate environment. Most corporations would need to use a wiki with more access controls like TWiki or confluence.
I've heard that a while ago, some folks inside Intel set up a mediawiki site for internal documentation, and when the lawyers heard about it, they had the project shut down. There was too much liability I gather.
Anyway, I've set up a TWiki installation at my work three years ago now and am not totally sure if I would call the project a success story.
First off, in order for management to buy in, we needed to demostrate that we could lock down sections and properly secure the site. With TWiki it was easy to create 'Webs' with different access permissions.
Second, in order to get users to buy in, we needed a reason for them to go to the site. For example, they go to the site and find useful content, and so therefore get an idea that the wiki is useful for them. When starting out, there is no useful content in a wiki. Two or three of us have been dutifully putting in what we know into the site, but it has taken maybe two years of effort before there is enough content for other people to find the site semi worthwhile.
It would be better if everyone just 'got it' and contributed to the site right away.
Another issue is the employees that know a lot but like to horde information. Having a nice collaboration tool around will not get them to release nugets of information they certainly have.
Some people insist on publishing perfect documents to the wiki. Those kinds of documents are nearly impossible to come up with on the first pass, and so a lot of documentation that is partially captured and written up has never made it to the site. I'm sure this issue is partly my fault as I haven't impressed upon the users that it is ok ot put up half ass content into a wiki, and then work on it later to polish it up if it would be useful.
One more problem we've had is that most of the employees think that documents in the wiki are owned by the creator of the document and not to be touched by them. I know I have explained about colaboration and cooperative editing a lot, but still, most people will not fix minor errors in documents created by other people. I have had people come up to me and tell me that there is an error in a document I've written in the wiki, or some have sent me emails telling me so. I have demonstrated that they should have edited the page themselves and fixed the error, and now a few of my fellow employees are more willing to do that.
Anyway, even dispite all of these hardships and difficulties, I think the wiki installation is invaluable, and at least for the few of us that 'get it' with regards to wikis, the site has made us maybe 20-30% more productive then we were before we had it around.
The reason there is one torrent download instead of two has more to do with managability issues than anything else. Over the lifetime of this torrent (which could be years) it will be easier to manage and maintain as one file instead of two. It will be easier for seeders to seed one file instead of two. And I would think it would be easier for users to download just one file instead of two.
If you wanted to download just the mp3s you could use one of the many mirrors listed on the doom site.
If you wanted to download the flac version so that you could burn them to CD, what is the big deal with downloading an extra 134MB of mp3 data along the way? You could always move the mp3s onto your iPod or whatever anyway, or just trash them.
Of course this makes me look insensitive to the modem users out there... but if you have broadband, this is just not a big deal.
-SuperTux
I've wondered about this kind of thing as well. The one thing that the RDP based implementations seem to have over standard X versions is that they can pipe the sound over along with display.
Is there any way to get full sound support when connecting remotely to your (flavor of) unix server? Preferably a free/open solution...
-supertux
I'll second this.
I'm rolling out an openvpn 2 setup now, and I have to say I'm quite impressed by the package. It seems very stable, gives good performance, has clients for everything we are going to use, and is open source.
The big reason why I chose openvpn over other solutions like IPSEC was basically because I couldn't find free IPSEC clients for windows.
If I had any advice for someone setting up openvpn, it would be to figure out what kind of a setup you want before you try to implement it. There are so many options and possibilities that you probably won't need. I had trouble keeping everything straight before I diagramed what I wanted the VPN setup to look like in the end (routed/nat/two factor authentication/etc)
-SuperTux
I've actively been using the bayesian filter that Mozilla comes with for a bit over a year now. Although it seemed to take forever to get 'trained' to what I consider spam, I've found that it works exceptionally well, maybe mismarking a legitimate email to me probably less than a dozen times so far (after the initial round of training).
Maybe six months ago, I noticed I was receiving quite a lot of these hash busting spams and I was bummed that maybe the bayesian filter wasn't the be all end all of spam filters.
But I pressed on using it, and in time, almost all of the hash busting emails are again getting filtered as spam.
I'd guess there are only so many different ways people can write Vi@gra and still have it be readible...
SuperTux
Dude!
I remember you from the old BBS days! And Michelle and Julian's wedding.
Did you ever convert off the video you took at the wedding to vcd?
SuperTux (aka Aubrey at the wedding, -=Thunder=- on WWIVnet)
I've got a 150 megabit (like 10 T1s) connection that isn't doing anything today, so here are the links from my server.
Unreal 2003 Windows demo
Unreal 2003 Linux demo via ftp
SuperTux