The dude that runs this project is a douche. Don't support it. I attempted to build his stuff from scratch to see if it's even possible . Build scripts were poorly documented, and I knew after I had downloaded like 8GB of source that something was fishy.
It may work for you - but this guy does very little to help openwrt.
Use xbmc - it's an excellent HTPC with a 10 foot user interface and a TON of content available. Plays pictures, movies and music, and has the best plug-in in the world (IMO)
NaviX - Internet content automagic / site scraper
http://www.xbmc.org
Get yourself a teeny atom/nvidia ion box (acer revo/zotac )
take a look at S3 or jungledisk and see if you can somehow have that make sense. Mozy might not be a bad idea either. The big tradeoff is
1) limited SLAs (privacy / latency / bandwidth) for getting to your data; once you host with a provider and accept their physical / logical storage footprint - you are constrained to living in their hosting model vs 2) providing quick access to your dataset b/c you have special sauce
The timing / requirements to get back to the data are the things that should drive your behavior. It might make sense to turn the data over to the customer - so when they need you to work on it- they provide it back.
I think you might be looking at this all wrong - why not redo the analysis and charge for the whole thing again? (According to the RIAA/MPAA isn't that what should happen when we scratch our movies or music?)
I did a write-up on this product in the beginning of this month - can run quagga routers in the UML image of your choice - wrote / ran a 12 router lab that ran on a p4 with 512MB / RAM. (http://www.vlcg.net/content/cloonix-clownix-rocks)
If this product was used - you would only be able to functionally test the protocols in a particular topology - wouldn't be cisco, and it wouldn't be the same as production (different protocols, different topologies).
I discovered this trying to figure out a way to run quagga in a gns3-like setup. GNS3 is great for testing a specific cisco thing that you need to learn about - but it didn't do well for me beyond 3 routers - (too much hand-holding getting the environment tweaked).
My ultimate vision for quagga would be to run it on the hypervisor and let it scale (in numbers of routing instances) wrt to the number of hypervisors - it's a pipe dream for now, but I think that routing that can scale with hypervisors is going to be a big challenge for cisco (esp if they try to do it in silicon) -
, and placed 3x rave 2u cases with some ultrsSparcAXI hardware inside, couple of 1u switches, etc.
http://www.rave.net/PRODUCTS/axi-2U5-specs.html
The shockmount racks were expensive, (about $500 on ebay), but worth it so when I move out of the current house, I just decable the systems, and put the panels on for the movers. (It's nice to have the systems in a compact case setup as well, as oppposed to all spread out over creation.)
My wife got sick of the 3 tables/desks worth of computer equipment hogging up the rec room space.
Not so great picture, here: http://stick.tookmymoney.com/gallery/tanabday2003/ IMG_1551
The lack of clutter is well worth the expense.
(O yeah, btw, don't overbuy the rack space, I have 25u of rack space, and I'm only using 10u.)
If you have legacy 10/100 devices that are plugged into that segment, jumbo gigE frames will NEVER work with the legacy devices. gigE frames appear to be L2 MAC errors as the preamble, source, destination, length addressing may line up in the front of the frame, but the crc at the rear will never line up. (Ethernet II frame illustrated below)
But anyways, I think there are demonstrations with some workloads saturating a gigE w/o using jumbo frames.
[snip] from http://sd.wareonearth.com/~phil/net/overhead/
Gigabit Ethernet with Jumbo Frames Gigabit ethernet is exactly 10 times faster than 100 Mbps ethernet, so for standard 1500 byte frames, the numbers above all apply, multiplied by 10. Many GigE devices however allow "jumbo frames" larger than 1500 bytes. The most common figure being 9000 bytes. For 9000 byte jumbo frames, potential GigE throughput becomes (from Bill Fink, the author of nuttcp):
No matter what, if you have an active cpu that fails on a sun host, the host panics. (It will blacklist the CPU once it reboots, but there's still an issue of that "reboot").
There isn't a "standard" SMP unix out there that wouldn't die a horrible death if you pull a cpu out that was running a critical section of the kernel when it was pulled. (I mean jese, how could it; let me pull yer brain out while yer thinking...)
Sun does let you stop/remove and add hardware on the fly gracefully, but they don't protect you against hard hardware failures thru any type of fault tolerance in the os/hardware.
Sun did offer a fault/tolrant unix platform that would deal with component failure/etc, (not sure how - checkpoints? voting? logging?), but it was a combined h/w s/w solution that was pretty expensive.
I'm not sure if Sun/Linux are really fighting the battle that you speak of (stability/performance).
The battle IMHO is horizontal vs vertical scalablity, sun has won the vertical scalability, but it's expensive.
Is it cheaper to have a 6-node 4-way SMP linux cluster running OPS or a single 6800 running oracle? Dunno.
My best guess is that with hardware advancing so quickly in the x86 world, the only boxes where sun has a huge advantage is the high end (> 8GB RAM).
The average techincal user should be anxious to RTFM for any understanding that it will bring to bear on the problem.
Installing debian is no better/worse than redhat, (granted it's curses, not X11) but it still requires that you understand the basic of partioning, package selection, and application configuration, (networking, X11, etc).
The autodetection of redhat/mandrake are great, but when they don't work, you end up RTFM anyways.
And how the heck do you propose I boot my 90 Mhz PI with no cdrom boot support? The whole cdrom boot process is based upon boot floppies, without the floppy images, your cdrom wouldn't boot.
Is there anyone attempting to use one of the A -> D converters to play xbox/ps2 on a monitor at a decent resolution, (1024 x 768)? I'd like to get something that would do S-VID -> SVGA/XGA to play xbox on a display projector. (But not get robbed blind in the process.) Does it look better than a TV? Is there flicker? Is it a bitch to setup?
--Adrian
I thought it said Verizon fatigue
on
Version Fatigue
·
· Score: 0, Offtopic
Like when you get sick of those assholes telling you yer line isn't good enough for DSL. (Well shit, if I owned that piece of copper, it'd be fibre!)
Just typical bs stuff that Microsoft(TM) intends to use, what is pretty neat is the recognition of the bottom up approach of most *nix implementations which makes it VERY difficult to fight from a M$ standpoint, *any vendor will have trouble replacing a well implemented free system*.
The idea that sales reps have open access to customer data centers (ie - OWNS the floor) is a silly notion and lets you know clueless people are allowing sales reps into the data centers.
All I can say is bring me yer M$ sales engineer and I'll feed his ass to the wolves in front of clients.
Bear's "Blood Magic" was a book so far beyond it's time it's frightening. It was beyond nanotech before it even started. It and Vernor Vinge's "A Fire upon the Deep" as books that are beyond examples of genre, and actually books that are ground breaking for their era, and therefore classics.
"Dune" was 2nd to only LOTR, (but I'm only modestly read.)
Windows 2000 still sucks! Did the quicklaunch bar make it that much better?
Architecture vs engineering
on
Coder or Architect?
·
· Score: 4, Informative
What you see when you move from a more in depth position (network/systems/software engineer) to a more abstract position is a requirement to understand and deploy/spearhead solutions to solve business "problems". If you are lucky, you are a revenue stream and life is easy. If you are unlucky, you are eventually required to support Micro$oft products. (grin)
The basic priciples descibed for software engineering always apply, but it's my opinion that the cycle is much more fluid at a higher level. (More people to appease, more requirements to understand.)
Raw talent is good to have, but the soft skills to interface and move projects that invariably cross business units forward become quite important. Don't worry about offering specific advice to folks; they are prolly more interested in following the steps of a high riser. Make sure to keep adept at the technologies that made you successful (DBA/software/networking), but also try to consider solutions with different types of technologies.
Do things that make peoples jobs easier, look for patterns in problems, look for the same in solutions. Try to learn from people that have been there, use newgroups, discussion forums, friends to your advantage. Be good at being a leader; don't be afraid to say you fucked something up when it happens.
As you said, the idea actually is to cluster the backplanes and replicate memory state across the boxes relative to an executing user process(Single system image). The seems like an N^2 instance, as you have N processes accessing N memory locations, across a shared backplane. (assuming a nice uniform process/kernel thread density)
I dunno enough about the hardware architecture of memory to understand the bandwidth/latency concerns of replication across a shared backplane, but the first thing to come down will be the single system image. (Which will require kernel caged stuff to be replicated, accessed in a sane manner)
Shouldn't be too bad of an extrapolation to get user space library/code sharing working after a kernel strategy is defined.
--Adrian
Portsentry is your friend
on
Code Redux
·
· Score: 1
been running full smtp, http, https, and ssh on @ home forever, with a portsentry monitor to blackhole scanners. With the apache logs, at least I can blackhole more hosts, and not have any risk of future exploits of those hosts impacting me.
I'm not too concerned about the TOS violations/enforcement, as services can move to any port no problem. (Just have to set up the a new mailer in sendmail for the desired port from yer buddies relay.)
There really can't be anyone more savvy working @home than a group of 10-15 unix/networking people "thinking out loud" on irc. (Except fer them with their irc buds...)
Randomly selecting portions of the binary for checksum are not going to be too difficult for aol to do, so any solution should probably point to an instance of the aim binary at some predetermined location in the aim clone tree. (What we really need is some nullsoft peeps to smack these aim developers around a bit.)
It may make sense to have clients install the aim binary seperately into a pretermined location of the source tree./usr/local/share/gaim/aim.exe? Then any transform that can be gleaned from aim should be easily applicable to gaim.
This is crappy of aol, but the PIA factor from the aim folks isn't nearly high enough to stop the competing products.
I'm not sure anyone else picked up on this, but isn't IBM the leading media researcher in the world right now? If they are represented on the committee/body/whatever, doesn't this make great sense to them?
ergo
1) lets develop a method to MAKE people use our technology
2) lets get a standard approved and hide copy prevention inside of it
3) in parallel, lets get a digital media copy prevention standard thru some dumbass body in need of protection like RIAA
4) they distribute their intellectual property to the public
5) then lets get that website ready so people can login with their credit card and PAY US to move a file.
Shew, it sounds like a stretch, but its so out there, it makes me nervous.
IBM if I find out you are the advocate of such an ill thought out technology, I will make a point to NEVER buy your hardware again. More importantly, I will persuade my clients to not buy/use any RS/6000 solution.
Please don't let IBM off the hook. S7A's are expensive relative to suns anyway.
I'm pretty sure the FreeBSD dummynet device can do this, I haven't configured it, but in the right kind of combination, (serial ethernet Token Ring ATM) interface/tunnel mixes, you could simulate PPP,ring and broadcast technologies for WAN or LAN....
http://www.iet.unipi.it/~luigi/ip_dummynet/
Got sick of getting jammed by inconsistent floppies, so I installed mod_dav and mod_ssl into apache and "mount" the sucker using web folders for windows. If yah tried this, you could buy a box or two or four (HW/RAID 5, 4 CPU's, FC/AL, 2+ GB RAM, TEST IT FIRST!), (charge each student like 2-3$/meg of remote storage), that should be fully capable of handling the load (it would be prudent to check the scaling prior to purchase, but it shouldn't be too bad to verify the scalablility using a bit of perl to emulate a population of client file operations.) You don't want students plugging their own drives into computers, or media into drives. They will break something. mod_dav isn't very quick for file transfers (256 kB/sec), (dunno why, the cable modem I'm on isn't rev limited upstream yet..., maybe some funky locking semantics in mod_dav, haven't looked at it yet.)I also have slow as crap ide drives on my box.
The nice thing about this solution is that it is usable thru any firewalls which permit tcp outbound 80/443.
Slashdot Mirror
The dude that runs this project is a douche. Don't support it. I attempted to build his stuff from scratch to see if it's even possible . Build scripts were poorly documented, and I knew after I had downloaded like 8GB of source that something was fishy.
It may work for you - but this guy does very little to help openwrt.
Please use openwrt - or x-wrt.
--Adrian
Use xbmc - it's an excellent HTPC with a 10 foot user interface and a TON of content available. Plays pictures, movies and music, and has the best plug-in in the world (IMO)
NaviX - Internet content automagic / site scraper
http://www.xbmc.org
Get yourself a teeny atom/nvidia ion box (acer revo/zotac )
take a look at S3 or jungledisk and see if you can somehow have that make sense. Mozy might not be a bad idea either. The big tradeoff is
1) limited SLAs (privacy / latency / bandwidth) for getting to your data; once you host with a provider and accept their physical / logical storage footprint - you are constrained to living in their hosting model
vs
2) providing quick access to your dataset b/c you have special sauce
The timing / requirements to get back to the data are the things that should drive your behavior. It might make sense to turn the data over to the customer - so when they need you to work on it- they provide it back.
I think you might be looking at this all wrong - why not redo the analysis and charge for the whole thing again? (According to the RIAA/MPAA isn't that what should happen when we scratch our movies or music?)
--Adrian
http://www.clownix.net
I did a write-up on this product in the beginning of this month - can run quagga routers in the UML image of your choice - wrote / ran a 12 router lab that ran on a p4 with 512MB / RAM. (http://www.vlcg.net/content/cloonix-clownix-rocks)
If this product was used - you would only be able to functionally test the protocols in a particular topology - wouldn't be cisco, and it wouldn't be the same as production (different protocols, different topologies).
I discovered this trying to figure out a way to run quagga in a gns3-like setup. GNS3 is great for testing a specific cisco thing that you need to learn about - but it didn't do well for me beyond 3 routers - (too much hand-holding getting the environment tweaked).
My ultimate vision for quagga would be to run it on the hypervisor and let it scale (in numbers of routing instances) wrt to the number of hypervisors - it's a pipe dream for now, but I think that routing that can scale with hypervisors is going to be a big challenge for cisco (esp if they try to do it in silicon) -
--Adrian
It's not a smart card, it's a kind of tap and go card which won't require anything but proximity to the reader in order to get an auth.
We're looking to make micropayments faster, much faster than the cycle which you wait for now when doing the typical swipe transaction.
I bought a 10u and a 15u 24" deep AJ case
n t/ index.htm
/ IMG_1551
http://ajcases.com/ajtext/products/di-shock-mou
, and placed 3x rave 2u cases with some ultrsSparcAXI hardware inside, couple of 1u switches, etc.
http://www.rave.net/PRODUCTS/axi-2U5-specs.html
The shockmount racks were expensive, (about $500 on ebay), but worth it so when I move out of the current house, I just decable the systems, and put the panels on for the movers. (It's nice to have the systems in a compact case setup as well, as oppposed to all spread out over creation.)
My wife got sick of the 3 tables/desks worth of computer equipment hogging up the rec room space.
Not so great picture, here: http://stick.tookmymoney.com/gallery/tanabday2003
The lack of clutter is well worth the expense.
(O yeah, btw, don't overbuy the rack space, I have 25u of rack space, and I'm only using 10u.)
--Adrian
Framesize is a function of hardware capability.
If you have legacy 10/100 devices that are plugged into that segment, jumbo gigE frames will NEVER work with the legacy devices. gigE frames appear to be L2 MAC errors as the preamble, source, destination, length addressing may line up in the front of the frame, but the crc at the rear will never line up. (Ethernet II frame illustrated below)
Preamble|Source MAC|Destination MAC|length|data|CRC
This is exactly like MTU's not lining up.
But anyways, I think there are demonstrations with some workloads saturating a gigE w/o using jumbo frames.
[snip] from http://sd.wareonearth.com/~phil/net/overhead/
Gigabit Ethernet with Jumbo Frames
Gigabit ethernet is exactly 10 times faster than 100 Mbps ethernet, so for standard 1500 byte frames, the numbers above all apply, multiplied by 10. Many GigE devices however allow "jumbo frames" larger than 1500 bytes. The most common figure being 9000 bytes. For 9000 byte jumbo frames, potential GigE throughput becomes (from Bill Fink, the author of nuttcp):
Rocket propelled grenade. Dumbasses
;)
No matter what, if you have an active cpu that fails on a sun host, the host panics. (It will blacklist the CPU once it reboots, but there's still an issue of that "reboot").
There isn't a "standard" SMP unix out there that wouldn't die a horrible death if you pull a cpu out that was running a critical section of the kernel when it was pulled. (I mean jese, how could it; let me pull yer brain out while yer thinking...)
Sun does let you stop/remove and add hardware on the fly gracefully, but they don't protect you against hard hardware failures thru any type of fault tolerance in the os/hardware.
Sun did offer a fault/tolrant unix platform that would deal with component failure/etc, (not sure how - checkpoints? voting? logging?), but it was a combined h/w s/w solution that was pretty expensive.
I'm not sure if Sun/Linux are really fighting the battle that you speak of (stability/performance).
The battle IMHO is horizontal vs vertical scalablity, sun has won the vertical scalability, but it's expensive.
Is it cheaper to have a 6-node 4-way SMP linux cluster running OPS or a single 6800 running oracle? Dunno.
My best guess is that with hardware advancing so quickly in the x86 world, the only boxes where sun has a huge advantage is the high end (> 8GB RAM).
The average techincal user should be anxious to RTFM for any understanding that it will bring to bear on the problem.
Installing debian is no better/worse than redhat, (granted it's curses, not X11) but it still requires that you understand the basic of partioning, package selection, and application configuration, (networking, X11, etc).
The autodetection of redhat/mandrake are great, but when they don't work, you end up RTFM anyways.
And how the heck do you propose I boot my 90 Mhz PI with no cdrom boot support? The whole cdrom boot process is based upon boot floppies, without the floppy images, your cdrom wouldn't boot.
--Adrian
Is there anyone attempting to use one of the A -> D converters to play xbox/ps2 on a monitor at a decent resolution, (1024 x 768)? I'd like to get something that would do S-VID -> SVGA/XGA to play xbox on a display projector. (But not get robbed blind in the process.) Does it look better than a TV? Is there flicker? Is it a bitch to setup?
--Adrian
Like when you get sick of those assholes telling you yer line isn't good enough for DSL. (Well shit, if I owned that piece of copper, it'd be fibre!)
The most interesting thing I see with this is why?
It can't be managed. It can't be monitored. It can't be logged.
This may be fine as a novelty, but running a network secured with such a hack is silly.
Let's talk about shutting down all userspace processes on a box except syslog and snort and I say you've got an interesting box.
OTW - it seems just like a game.
--Adrian
Just typical bs stuff that Microsoft(TM) intends to use, what is pretty neat is the recognition of the bottom up approach of most *nix implementations which makes it VERY difficult to fight from a M$ standpoint, *any vendor will have trouble replacing a well implemented free system*.
The idea that sales reps have open access to customer data centers (ie - OWNS the floor) is a silly notion and lets you know clueless people are allowing sales reps into the data centers.
All I can say is bring me yer M$ sales engineer and I'll feed his ass to the wolves in front of clients.
"IIS, need I say more?"
--Adrian
Bear's "Blood Magic" was a book so far beyond it's time it's frightening. It was beyond nanotech before it even started. It and Vernor Vinge's "A Fire upon the Deep" as books that are beyond examples of genre, and actually books that are ground breaking for their era, and therefore classics.
"Dune" was 2nd to only LOTR, (but I'm only modestly read.)
--Adrian
Windows 2000 still sucks! Did the quicklaunch bar make it that much better?
What you see when you move from a more in depth position (network/systems/software engineer) to a more abstract position is a requirement to understand and deploy/spearhead solutions to solve business "problems". If you are lucky, you are a revenue stream and life is easy. If you are unlucky, you are eventually required to support Micro$oft products. (grin)
The basic priciples descibed for software engineering always apply, but it's my opinion that the cycle is much more fluid at a higher level. (More people to appease, more requirements to understand.)
Raw talent is good to have, but the soft skills to interface and move projects that invariably cross business units forward become quite important. Don't worry about offering specific advice to folks; they are prolly more interested in following the steps of a high riser. Make sure to keep adept at the technologies that made you successful (DBA/software/networking), but also try to consider solutions with different types of technologies.
Do things that make peoples jobs easier, look for patterns in problems, look for the same in solutions. Try to learn from people that have been there, use newgroups, discussion forums, friends to your advantage. Be good at being a leader; don't be afraid to say you fucked something up when it happens.
Take responsibilty.
--Adrian
As you said, the idea actually is to cluster the backplanes and replicate memory state across the boxes relative to an executing user process(Single system image). The seems like an N^2 instance, as you have N processes accessing N memory locations, across a shared backplane. (assuming a nice uniform process/kernel thread density)
I dunno enough about the hardware architecture of memory to understand the bandwidth/latency concerns of replication across a shared backplane, but the first thing to come down will be the single system image. (Which will require kernel caged stuff to be replicated, accessed in a sane manner)
Shouldn't be too bad of an extrapolation to get user space library/code sharing working after a kernel strategy is defined.
--Adrian
been running full smtp, http, https, and ssh on @ home forever, with a portsentry monitor to blackhole scanners. With the apache logs, at least I can blackhole more hosts, and not have any risk of future exploits of those hosts impacting me.
I'm not too concerned about the TOS violations/enforcement, as services can move to any port no problem. (Just have to set up the a new mailer in sendmail for the desired port from yer buddies relay.)
There really can't be anyone more savvy working @home than a group of 10-15 unix/networking people "thinking out loud" on irc. (Except fer them with their irc buds...)
--Toilethead
DAMMIT - I fergot my broom!!!
Randomly selecting portions of the binary for checksum are not going to be too difficult for aol to do, so any solution should probably point to an instance of the aim binary at some predetermined location in the aim clone tree. (What we really need is some nullsoft peeps to smack these aim developers around a bit.)
/usr/local/share/gaim/aim.exe? Then any transform that can be gleaned from aim should be easily applicable to gaim.
It may make sense to have clients install the aim binary seperately into a pretermined location of the source tree.
This is crappy of aol, but the PIA factor from the aim folks isn't nearly high enough to stop the competing products.
--Adrian
I'm not sure anyone else picked up on this, but isn't IBM the leading media researcher in the world right now? If they are represented on the committee/body/whatever, doesn't this make great sense to them?
ergo
1) lets develop a method to MAKE people use our technology
2) lets get a standard approved and hide copy prevention inside of it
3) in parallel, lets get a digital media copy prevention standard thru some dumbass body in need of protection like RIAA
4) they distribute their intellectual property to the public
5) then lets get that website ready so people can login with their credit card and PAY US to move a file.
Shew, it sounds like a stretch, but its so out there, it makes me nervous.
IBM if I find out you are the advocate of such an ill thought out technology, I will make a point to NEVER buy your hardware again. More importantly, I will persuade my clients to not buy/use any RS/6000 solution.
Please don't let IBM off the hook. S7A's are expensive relative to suns anyway.
You can't build redundancy with a single cable. They shoulda ran two cables, one from like American Samoa. heh
I'm pretty sure the FreeBSD dummynet device can do this, I haven't configured it, but in the right kind of combination, (serial ethernet Token Ring ATM) interface/tunnel mixes, you could simulate PPP ,ring and broadcast technologies for WAN or LAN....
http://www.iet.unipi.it/~luigi/ip_dummynet/
Got sick of getting jammed by inconsistent floppies, so I installed mod_dav and mod_ssl into apache and "mount" the sucker using web folders for windows. If yah tried this, you could buy a box or two or four (HW/RAID 5, 4 CPU's, FC/AL, 2+ GB RAM, TEST IT FIRST!), (charge each student like 2-3$/meg of remote storage), that should be fully capable of handling the load (it would be prudent to check the scaling prior to purchase, but it shouldn't be too bad to verify the scalablility using a bit of perl to emulate a population of client file operations.)
You don't want students plugging their own drives into computers, or media into drives. They will break something.
mod_dav isn't very quick for file transfers (256 kB/sec), (dunno why, the cable modem I'm on isn't rev limited upstream yet..., maybe some funky locking semantics in mod_dav, haven't looked at it yet.)I also have slow as crap ide drives on my box.
The nice thing about this solution is that it is usable thru any firewalls which permit tcp outbound 80/443.