The future of news will be with the news aggregators like Topix.net. A
system that auto-categorizes the news, finding news from far-flung sources about anything you might be interested in. No one has time to read every news source. The number of sources is only going to grow as more and more citizen journalists and blogs evolve into a network of reliable local news.
Yahoo news crawls some 7000+ news sites, Google News crawls 4500+
English news sites, and Topix.net crawls 10,000+ news sites. Once you
add in the thousands of local blogs, you will need a system like
Topix.net to filter the relentless stream of news articles and posts
that are generated every day. You will need something that can sort
through the news, determine the trends, and ignore the old repeated
stories for you and present them to you in RSS for consumption with
your favorite RSS news reader.
Topix.net factors in site registration when it decide which articles to show. Given ten copies of the same/similar story it will bias the source selection to ones that do not require registration.
It's all Apache's fault. Load is at 0.2 and running with the CPU 90% idle. For some reason apache.conf thinks you should never have more than 150 simultaneous open connections. Grrr. Apache should ship with an alternate config called slashdot.porn.conf with all the right performance tweaks.:) Anyway, server is all better now. At least until all those horny hackers wake up tomorrow morning and read slashdot...
Don't be trash talking our ISP, Meer.net rocks the net. A quick shout out to JG and DG.
OK, I've built four such arrays ranging from 0.5TB to 0.8TB, one for home the others for work. The first 3 used 3ware IDE RAID cards. They suck. Support in Linux is OK, but the card's BIOS and user interface suck.
Linux software raid is easier to configure and it is easier to fix bad drives. Plus its cheaper.
The trick is to only ever put one IDE device per channel. Which means you need extra PCI ide controlers. SATA works great too.
If you're archiving your photos, music, and movies you don't need 130Mb/s 30Mb/s is just fine.
Also, any design that requires your disk to be fast, is a bad one. Do you think your Google queries are waiting for a disk seek? Well, they aren't.
All you need is linux, a $80 tower case, a bunch of 250GB drives at $0.50/GB, and Linux Software raid. So starting around $1k for a terabyte of storage. With a cheap CPU and motherboard.
Of course it helps to be within driving distance of a Fry's...
If its not your money, sure, buy it off the shelf. If it is your money, build your own.
Its just that simple. You need 3 (or more) drives. Last time I checked Fry's had 250GB drives for $139 each. So you'll have almost half a terabyte of storage for about $420.
Linux SW RAID is a breeze, just google for "linux sw raid howto"
If you go to disk, just once, you need about 9ms just to get the disk heads in position. If you're reading a file system of complex database, you now have multiple disk seeks and reads. That adds up. Seeking in RAM is orders of magnitude faster. That's why all the good search engines keep *everything* in RAM all the time.
I'm currently building a very similar device, but with a high end audio card with a > 100db signal to noise ratio and digital out. I just bought an ATI AIW card to use to drive the display through the TV.
The bad news is: there is no Linux support for TV-out. You need the Windoze software to configure the TV display.
I've seen some discussion about using the frame buffer interface for the ATI to set the right mode for TV out. But just talk, no bits.
Apparently nVidia has a card that has TV-out support under Linux. I'm heading to Fry's this afternoon to pick one up.
Blekko
There are several motherboards designed for servers that have nearly everything you describe onboard already: 1 or 2 ethernet ports, AGP video, sound chip, even SCSI. Tyan makes one and so does ASUS.
Here is just one example:
http://tyan.com/products/html/thunderhe_p.html
Slashdot Mirror
Yahoo news crawls some 7000+ news sites, Google News crawls 4500+ English news sites, and Topix.net crawls 10,000+ news sites. Once you add in the thousands of local blogs, you will need a system like Topix.net to filter the relentless stream of news articles and posts that are generated every day. You will need something that can sort through the news, determine the trends, and ignore the old repeated stories for you and present them to you in RSS for consumption with your favorite RSS news reader.
-AS
Topix.net factors in site registration when it decide which articles to show. Given ten copies of the same/similar story it will bias the source selection to ones that do not require registration.
-AS
Don't be trash talking our ISP, Meer.net rocks the net. A quick shout out to JG and DG.
-Bryn
Linux software raid is easier to configure and it is easier to fix bad drives. Plus its cheaper.
The trick is to only ever put one IDE device per channel. Which means you need extra PCI ide controlers. SATA works great too.
If you're archiving your photos, music, and movies you don't need 130Mb/s 30Mb/s is just fine.
Also, any design that requires your disk to be fast, is a bad one. Do you think your Google queries are waiting for a disk seek? Well, they aren't.
-SA
Of course it helps to be within driving distance of a Fry's...
If its not your money, sure, buy it off the shelf. If it is your money, build your own.
-AS
Topix.net also classifies all the news into 150,000 categories including 30,000 local pages with every city and town in the US.
Both Topix.net and Yahoo News have RSS/XML feeds as well.
Microsoft is using moreover.com to supply its news (As mentioned in the Mercury News Merc Article)
-AS
Linux SW RAID is a breeze, just google for "linux sw raid howto"
Bryn
If you go to disk, just once, you need about 9ms just to get the disk heads in position. If you're reading a file system of complex database, you now have multiple disk seeks and reads. That adds up. Seeking in RAM is orders of magnitude faster. That's why all the good search engines keep *everything* in RAM all the time.
That is why Google has multiple copies of the entire web in memory.
-AS
Otherwise, this is definitely good news.
I'm currently building a very similar device, but with a high end audio card with a > 100db signal to noise ratio and digital out. I just bought an ATI AIW card to use to drive the display through the TV. The bad news is: there is no Linux support for TV-out. You need the Windoze software to configure the TV display. I've seen some discussion about using the frame buffer interface for the ATI to set the right mode for TV out. But just talk, no bits. Apparently nVidia has a card that has TV-out support under Linux. I'm heading to Fry's this afternoon to pick one up. Blekko
There are several motherboards designed for servers that have nearly everything you describe onboard already: 1 or 2 ethernet ports, AGP video, sound chip, even SCSI. Tyan makes one and so does ASUS.
Here is just one example:
http://tyan.com/products/html/thunderhe_p.html
Blekko