The idea of people owning multiple cars, a city car for communing and a traditional car for other times, is very interesting. I own a mini-van because sometimes I need to put a bunch of stuff in it to move, but most of the time I just commute with it. The main problem with this is auto insurance. The insurance companies want to charge me for two cars with two drivers not as one car with an additional part time car. I think that the idea of owning an alternate commuter car will not work until the insurance industry changes their pricing policies.
Out here on the left coast the extension programs at the various University Of California campuses have some Linux Kernel classes. These tend to be developed and taught by engineers in the industry with a real working knowledge of the subject. UC Santa Cruz Extension, http://www.ucsc-extension.edu/ has an "Linux Kernel Architecture and Programming" which looks like an intro course. You can take it online or as two Saturdays. There is also a Linux device drivers class which a once a week class and an Advanced Device Drivers class which is 4 Saturdays. I checked the other campuses but they all seem to be summer schedule with a limited set of classes. Red Hat also has a one week Kernel internals class which is a "hands on" which to me means a trade off of less information for some finger programing of the brain. All of these courses seem to have an introductory flavor to them. But I suspect that you will learn a lot about all of the various areas of the kernel and how the different parts hang together. My experience as kernel hacker is that I have learned a lot about the parts I am interested in, but that there are many big areas of the kernel that I only have a superficial understanding of.
First off, the Doctor/lawyer issue is a red herring. Private mail to your doctor/lawyer is one to one personal mail. Goodmail is all about commercial bulk mail whether one to one personalized mail or one to many mailing list mail. It is not about personal one to one mail.
The real focus of goodmail is on the legitimate business bulk mail senders like Citybank, Paypal or HP. Goodmail sells their service to these companies to assist in sending mail. If you are Citybank then paying 1/4 cent per message or $2,400 per million messages in order to have a third party deal with coordinating with the large ISPs to bypass their spam filters seems like a good deal. Do the math. Assuming a million customers and a bulk message twice a month comes to a yearly cost of $57,600. Less than half the cost of a staff person who is likely to be unable to make sure all of your mail goes through. S/he might miss an ISP blocking your mail for 2 hours which might mean that 10,000 customers did not get their monthly statement. From a business perspective it is really important that all the customers get their monthly statements and well worth the cost of $57,600 per year.
Now the 1/8 cent per message that gets passed to the ISP is a motivator for the ISP to allocate resources to make sure that the mail that is paid for gets delivered reliably. This is more important for the small or medium sized ISPs who don't have a lot of resources or motivation to throw at managing whitelists. Again do the math. If an ISP has 1000 customers and goodmail can say that on average we will pass 10 verified messages a month to each of your customers that means we will pay you $150 per year. Humm... maybe goodmail needs a sliding scale so that mail to small ISPs cost more than mail to large ISPs. Bulk mail rate of 23 cents a letter means: 1000 x 10 x 12 x ($0.23 / 2) = $13,800 per year. And the business senders still save money because they do not have to print, collate, and label the mail. This higher rate would only apply to a subset of the total recipients. Listen-up small to midsized ISPs, you should band together and start your own certified mail service. Maybe even negotiate as a group with goodmail for this higher rate.
The next logical step for goodmail is to work with the ISPs to brutally filter out non-goodmail messages that appears to come from goodmail's sending customer. So for 1/2 cent per message in addition to getting your messages past the whitelists, the participating ISPs will also filter out 95% of the phishing mail sent that appears to come from your domain. I would bet that a bank would see the additional $57,600 per year as a cheap way to get rid of 95% of the phishing attempts at participating ISPs. Does it solve the phishing problem for the bank, no it does not. You still have the 5% of messages that get through and mail to non-participating ISPs. But if 25% of your customers use ISPs who accept goodmail and reject non-goodmail, then you have cut the phishing problem by 1/4. If phishing fraud costs the bank $1,000,000 per year, that is almost a $200,000 savings. I realize this number is very optimistic. An additional argument would be that your customers will feel more trust in reading mail from the bank because most of the bogus phishing mail is never seen by the customer. I know personally I junk all mail from paypal.com without even thinking about it because I know most of it is phishing. This means that PayPal can not send me bulk mail that I might be interested in because I junk it before I even look at it.
To summarize: Goodmail is not about personal email. It is really not even about the recipient of the message or the ISPs. Goodmail is focused on the legitimate business senders who don't want to deal with whitelists and spam blocking. For a modest cost, goodmail will make sure your messages get through to every customer at ISPs that are willing split the revenue with goodmail. Having advised clients trying to send bulk mail and dealing with the spam filt
If record companies stop releasing albums for new bands, what is going to happen to the live performance business? Would you shell out $25-$50 to see a new band based on only hearing one or two singles. Oh, I forgot, the ring tone is going to get you hopping in the isles. I know personally that I am much more likely to go see a band if I have heard their album. I know what to expect and I can look forward hearing more than one song I like.
A big issue with loosing the live performances and the band going out on tour is that they do not get a chance to work their skills. To be a good musician you have to play a lot. It is called "getting your chops". Practicing by yourself is find for the physical aspects of playing an instrument. Ya, s/he can play the notes. Rehearsing with a band makes you better a playing with other musicians. But performing in front of a live audience puts you under the gun to perform well. Not just play the instrument well, but to reach past that and put some magic into the music. Playing in front of a live audience 5 days a week, allows get comfortable with playing so that it becomes natural. Hard to do that with a click track and a MIDI band.
Finally what about the money. A lot of bands make more money from live performances than from music sales. For a good read on why you are likely to make more money per hour flipping burgers than from making a gold record try reading Moses Avalon's "Secrets of Negotiating a Record Contract" or try his Moses Avalon Royalty Calculator at http://www.mosesavalon.com/cgi-bin/calculatecopy.p l. On the other hand when they perform live the either get a fixed fee or a percentage of ticket sales. So the lack of an album may mean that you have no fans willing to shell out the cash to see you live, so there is little point in touring to support the single. And because you are not performing regularly you do not get the practice and experience to be an excellent musician. So you single may be your only single. Oh ya, I forgot, that ring tone is going to make every one want to download your next single.
Just hoping to hear the album so that I might want to see the band. RLH
Man, realmolo must be young. CDs had nothing to do with the album concept or the rise of Album Oriented Rock. The term "album" came from the old vinyl records or "LP"s. The marketing of a complete album of music came from rock 'n' roll in the mid to late 60's. I remember buying albums like Adam Heart Mother, Electric Ladyland, Tommy, Sounds of Silence, etc. This was 15 years before the first CD was issued.
Having mixed live sound I know that a board mix is fine for the band, but a real disappointment for a concert goer: not enough reverb, strange EQ, improper balance between the musicians. When you are mixing live sound, you are taking into account the musical wash (mush) coming off stage, the room sound, and the sound of the speakers. A good concert recording requires a separate mix and with the better recordings a separate mixer with EQ and effects. Sometimes the FOH (front of house) engineer can do both mixes, but it is a stretch. The FOH engineer's primary responsibility is the FOH sound. The flip side of the board mix is that if you buy it and take it home and it sucks then you will feel ripped off and might wonder if the concert was really as good as you remembered it.
I am surprised that no one has mentioned the Slashdot article on the
study at LBL on use of cell processors in High Performance Computing:
The Potential of Science With the Cell Processor
http://science.slashdot.org/article.pl?sid=06/05/2 8/047223/
This discusses research at Lawrence Berkeley National Laboratory on using
the STI Cell processor for scientific computing. From the article quoting
the LBL paper:
"Overall results demonstrate the tremendous potential of the Cell
architecture for scientific computations in terms of both raw performance
and power efficiency,"
and
"We also conclude that Cell's heterogeneous multi-core implementation
is inherently better suited to the HPC environment than homogeneous
commodity multi-core processors."
The paper went on to say that while the Cell processor was designed
for single-precision 32-bit floating performance but with some simple
changes to the design it could be optimized for double precision 64-bit
floating performance.
This makes a lot of sense if this is the same Cell processor that IBM is using in their blade servers.
Really cheap, really fast 9 core processors!
I know this is off topic, US market not China, AMD/Intel not MIPS, but...
Why are no vendors competing with the Mac mini?
I would have expected it by now. Same basic specs, same basic features, but $100-$200 cheaper.
I like small computers, but I am totally frustrated with the PC vendors. If you want a smaller box, you have to pay more. Hugh??? The box is smaller, less metal and plastic. The mother board is smaller, less PC board. The system has less stuff in it (or potential stuff) so the power supply is smaller. Everything is smaller but you pay a 20%-30% premium for it? It just does not make sense.
Design: Take a medium powerful CPU, AMD Semtron or Intel Celeron, and mother board chip-set that includes video, 5.1 sound, USB, firewire, and either a PATA or SATA disk interface. Build a 5 1/4" by 6-8" mother board around it with 1 or 2 memory slots and no floppy or PCI slots. Put it in a small case with a 3 1/2" disk and a CD/DVD -R/RW drive (maybe external USB or Firewire). If there is space us an internal power supply, otherwise use an external supply like the Mac mini. Design the case so it can be mounted on the back of a LCD display. And wallah, you have an almost zero footprint computer.
In a normal size, you can by a similar system, CPU, 256Mb, 80G, CD-DVD, for about $300. It should not be hard to make a similar small foot print system for about the same amount.
I see two big markets for this type of system: Consumers who need a web browser, email, a word processor,
a Quick Books like financial application and simple games Businesses who need to deploy 100's of systems to specific job
functions: customer service, call centers, order entry,
warehouse, and other non-computer specific jobs.
In the first case my parents come to mind. She does not do a lot with their computer. Mostly web browsing, email, and keeping track of their fiances. The only thing the have upgraded in their computer is memory. Other upgrades like card readers and zip drives have used USB. The last few times they have thought about upgrading more they have decided to buy a new computer. Sound familiar? I think that there are many people who feel less intimidated if the computer was the size of a long hardbound book. My mom would like it in violet.
The second case is business. A smaller computer would mean less desk space or floor space waisted. It would also take less power and would use less air conditioning. With energy costs the way they are, this is a big concern. A computer with less bells and whistles might mean fewer distractions for the employee. And again, for a group, 20-100, of computers what was the last upgrade you added inside the case? Memory? Maybe a new disk drive? What I see is that when the company decides to upgrade their computers, they just buy new computers. At $300 a unit, it not much more than a nice office chair or desk.
What I am surprised at is that none of the major computer vendors have made a product like this.
Forget the gamers and desk top users. For those users just put the RAM in the system.
Think mail servers, database servers, web servers and e-commerce servers. Any application that does heavy I/O. Real I/O, not just swap or paging. In particular any application that is write heavy. For true write it to the disk activities, kernel buffers if they get used, and they should not, are a loose because the data is not committed to non-volatile memory. Wrote the data to disk. Ya, right. Oops, the system crashed. You want it back? Well we were only kidding about the write, sorry (;-(.
Sendmail for example is very write intensive. Write each message to disk before you acknowledge receiving the message. Sendmail's mqueue directory normally is the performance bottle neck of a sendmail mail server. Since most of the files that sendmail writes are small, less than 1Kb, seek time is the important spec, not transfer speed. When you compare access time between memory and disk, memory blows disk out of the water. It is odd that Gigabyte does not mention access time in their data sheet. Not even a range of access times based on RAM speed. I would think it would be an important selling point.
Another use of this type of device is any journalling application such as a database or perhaps a journalling file system. If you can split the journal from the underlying data, then you could put the journal on the solid state disk and then let the application write the data to slow magnetic disk in the background.
Now what gigabyte or some other enterprising company should do with this type of product is:
1) Make sure it works with Linux (well maybe that other OS, Windoze, too) and make sure it is rock-solid reliable.
2) Support ECC RAM with error recovery. ECC is not an option on a server.
3) Repackage it in a normal 3.5 inch SATA enclosure with the connectors in the correct place. Who wants to put a PCI card into a 1U rack mount server? Maybe even have the option of a bigger battery although battery time is not really an issue with a server.
4) Support bigger DIMMs, design the controller to allow 4G, 8G, or larger DIMMs, but only "officially" support 1G or 2G DIMMs. I don't think that for many server applications size is that big of an issue. The really heavily accessed data quite often can be highly localized. The other thing is that you can always RAID several of these disks together.
5) Work on the pricing and package. You are going to put the card in a metal box, so charge more. Maybe in the $250 to $400 range with out memory. Also offer it as a complete package with memory, and charge more. These days since a 1G DIMM costs less than $100, you should charge at least $150 per Gig. Humm... 4G * $150 + $250 = $850. Less than $1K for a 4G screaming fast hard disk. We are talking micro-second access time, not milli-second access time. That is a product I could sink my teeth into.
A bit pricey for the enthusiast or the early adopter though. Maybe you could offer bare bone evaluation/hobbyist units for $150. Limit it to two units with web based support only. Get the bleeding edge experts to bang at it for you to find the bugs and reliability issues. And if it is a good product, they will start a buzz for the product. Just look at what is happening here on slashdot with the definitely marginal Gigabyte product.
Slashdot Mirror
The idea of people owning multiple cars, a city car for communing and a traditional car for other times, is very interesting. I own a mini-van because sometimes I need to put a bunch of stuff in it to move, but most of the time I just commute with it. The main problem with this is auto insurance. The insurance companies want to charge me for two cars with two drivers not as one car with an additional part time car. I think that the idea of owning an alternate commuter car will not work until the insurance industry changes their pricing policies.
RLH
Out here on the left coast the extension programs at the various University Of California campuses have some Linux Kernel classes. These tend to be developed and taught by engineers in the industry with a real working knowledge of the subject.
UC Santa Cruz Extension, http://www.ucsc-extension.edu/ has an "Linux Kernel Architecture and Programming" which looks like an intro course. You can take it online or as two Saturdays. There is also a Linux device drivers class which a once a week class and an Advanced Device Drivers class which is 4 Saturdays.
I checked the other campuses but they all seem to be summer schedule with a limited set of classes.
Red Hat also has a one week Kernel internals class which is a "hands on" which to me means a trade off of less information for some finger programing of the brain.
All of these courses seem to have an introductory flavor to them. But I suspect that you will learn a lot about all of the various areas of the kernel and how the different parts hang together. My experience as kernel hacker is that I have learned a lot about the parts I am interested in, but that there are many big areas of the kernel that I only have a superficial understanding of.
Hope this helps
RLH
First off, the Doctor/lawyer issue is a red herring. Private mail to your doctor/lawyer is one to one personal mail. Goodmail is all about commercial bulk mail whether one to one personalized mail or one to many mailing list mail. It is not about personal one to one mail.
The real focus of goodmail is on the legitimate business bulk mail senders like Citybank, Paypal or HP. Goodmail sells their service to these companies to assist in sending mail. If you are Citybank then paying 1/4 cent per message or $2,400 per million messages in order to have a third party deal with coordinating with the large ISPs to bypass their spam filters seems like a good deal. Do the math. Assuming a million customers and a bulk message twice a month comes to a yearly cost of $57,600. Less than half the cost of a staff person who is likely to be unable to make sure all of your mail goes through. S/he might miss an ISP blocking your mail for 2 hours which might mean that 10,000 customers did not get their monthly statement. From a business perspective it is really important that all the customers get their monthly statements and well worth the cost of $57,600 per year.
Now the 1/8 cent per message that gets passed to the ISP is a motivator for the ISP to allocate resources to make sure that the mail that is paid for gets delivered reliably. This is more important for the small or medium sized ISPs who don't have a lot of resources or motivation to throw at managing whitelists. Again do the math. If an ISP has 1000 customers and goodmail can say that on average we will pass 10 verified messages a month to each of your customers that means we will pay you $150 per year. Humm... maybe goodmail needs a sliding scale so that mail to small ISPs cost more than mail to large ISPs. Bulk mail rate of 23 cents a letter means: 1000 x 10 x 12 x ($0.23 / 2) = $13,800 per year. And the business senders still save money because they do not have to print, collate, and label the mail. This higher rate would only apply to a subset of the total recipients. Listen-up small to midsized ISPs, you should band together and start your own certified mail service. Maybe even negotiate as a group with goodmail for this higher rate.
The next logical step for goodmail is to work with the ISPs to brutally filter out non-goodmail messages that appears to come from goodmail's sending customer. So for 1/2 cent per message in addition to getting your messages past the whitelists, the participating ISPs will also filter out 95% of the phishing mail sent that appears to come from your domain. I would bet that a bank would see the additional $57,600 per year as a cheap way to get rid of 95% of the phishing attempts at participating ISPs. Does it solve the phishing problem for the bank, no it does not. You still have the 5% of messages that get through and mail to non-participating ISPs. But if 25% of your customers use ISPs who accept goodmail and reject non-goodmail, then you have cut the phishing problem by 1/4. If phishing fraud costs the bank $1,000,000 per year, that is almost a $200,000 savings. I realize this number is very optimistic. An additional argument would be that your customers will feel more trust in reading mail from the bank because most of the bogus phishing mail is never seen by the customer. I know personally I junk all mail from paypal.com without even thinking about it because I know most of it is phishing. This means that PayPal can not send me bulk mail that I might be interested in because I junk it before I even look at it.
To summarize: Goodmail is not about personal email. It is really not even about the recipient of the message or the ISPs. Goodmail is focused on the legitimate business senders who don't want to deal with whitelists and spam blocking. For a modest cost, goodmail will make sure your messages get through to every customer at ISPs that are willing split the revenue with goodmail. Having advised clients trying to send bulk mail and dealing with the spam filt
If record companies stop releasing albums for new bands, what is going to happen to the live performance business? Would you shell out $25-$50 to see a new band based on only hearing one or two singles. Oh, I forgot, the ring tone is going to get you hopping in the isles. I know personally that I am much more likely to go see a band if I have heard their album. I know what to expect and I can look forward hearing more than one song I like.
p l. On the other hand when they perform live the either get a fixed fee or a percentage of ticket sales. So the lack of an album may mean that you have no fans willing to shell out the cash to see you live, so there is little point in touring to support the single. And because you are not performing regularly you do not get the practice and experience to be an excellent musician. So you single may be your only single. Oh ya, I forgot, that ring tone is going to make every one want to download your next single.
A big issue with loosing the live performances and the band going out on tour is that they do not get a chance to work their skills. To be a good musician you have to play a lot. It is called "getting your chops". Practicing by yourself is find for the physical aspects of playing an instrument. Ya, s/he can play the notes. Rehearsing with a band makes you better a playing with other musicians. But performing in front of a live audience puts you under the gun to perform well. Not just play the instrument well, but to reach past that and put some magic into the music. Playing in front of a live audience 5 days a week, allows get comfortable with playing so that it becomes natural. Hard to do that with a click track and a MIDI band.
Finally what about the money. A lot of bands make more money from live performances than from music sales. For a good read on why you are likely to make more money per hour flipping burgers than from making a gold record try reading Moses Avalon's "Secrets of Negotiating a Record Contract" or try his Moses Avalon Royalty Calculator at http://www.mosesavalon.com/cgi-bin/calculatecopy.
Just hoping to hear the album so that I might want to see the band.
RLH
Man, realmolo must be young. CDs had nothing to do with the album concept or the rise of Album Oriented Rock. The term "album" came from the old vinyl records or "LP"s. The marketing of a complete album of music came from rock 'n' roll in the mid to late 60's. I remember buying albums like Adam Heart Mother, Electric Ladyland, Tommy, Sounds of Silence, etc. This was 15 years before the first CD was issued.
Just my 2 cents
RLH
Having mixed live sound I know that a board mix is fine for the band, but a real disappointment for a concert goer: not enough reverb, strange EQ, improper balance between the musicians. When you are mixing live sound, you are taking into account the musical wash (mush) coming off stage, the room sound, and the sound of the speakers. A good concert recording requires a separate mix and with the better recordings a separate mixer with EQ and effects. Sometimes the FOH (front of house) engineer can do both mixes, but it is a stretch. The FOH engineer's primary responsibility is the FOH sound. The flip side of the board mix is that if you buy it and take it home and it sucks then you will feel
ripped off and might wonder if the concert was really as good as you remembered it.
Been there, done that.
RLH
The Potential of Science With the Cell Processor
http://science.slashdot.org/article.pl?sid=06/05/
It reference a second article:
Researchers Analyze HPC Potential of Cell Processor
http://www.hpcwire.com/hpc/671376.html/
This discusses research at Lawrence Berkeley National Laboratory on using the STI Cell processor for scientific computing. From the article quoting the LBL paper:
"Overall results demonstrate the tremendous potential of the Cell architecture for scientific computations in terms of both raw performance and power efficiency,"
and
"We also conclude that Cell's heterogeneous multi-core implementation is inherently better suited to the HPC environment than homogeneous commodity multi-core processors."
The paper went on to say that while the Cell processor was designed for single-precision 32-bit floating performance but with some simple changes to the design it could be optimized for double precision 64-bit floating performance.
This makes a lot of sense if this is the same Cell processor that IBM is using in their blade servers.
Really cheap, really fast 9 core processors!
An interesting read.
RLH
I know this is off topic, US market not China, AMD/Intel not MIPS, but...
Why are no vendors competing with the Mac mini?
I would have expected it by now. Same basic specs, same basic features,
but $100-$200 cheaper.
I like small computers, but I am totally frustrated with the PC vendors.
If you want a smaller box, you have to pay more. Hugh??? The box
is smaller, less metal and plastic. The mother board is smaller, less
PC board. The system has less stuff in it (or potential stuff) so the
power supply is smaller. Everything is smaller but you pay a 20%-30%
premium for it? It just does not make sense.
Design:
Take a medium powerful CPU, AMD Semtron or Intel Celeron, and mother board
chip-set that includes video, 5.1 sound, USB, firewire, and either a PATA
or SATA disk interface. Build a 5 1/4" by 6-8" mother board around it
with 1 or 2 memory slots and no floppy or PCI slots. Put it in a small
case with a 3 1/2" disk and a CD/DVD -R/RW drive (maybe external USB
or Firewire). If there is space us an internal power supply, otherwise
use an external supply like the Mac mini. Design the case so it can be
mounted on the back of a LCD display. And wallah, you have an almost zero
footprint computer.
In a normal size, you can by a similar system, CPU, 256Mb, 80G, CD-DVD,
for about $300. It should not be hard to make a similar small foot
print system for about the same amount.
I see two big markets for this type of system:
Consumers who need a web browser, email, a word processor,
a Quick Books like financial application and simple games
Businesses who need to deploy 100's of systems to specific job
functions: customer service, call centers, order entry,
warehouse, and other non-computer specific jobs.
In the first case my parents come to mind. She does not do a lot with
their computer. Mostly web browsing, email, and keeping track of their
fiances. The only thing the have upgraded in their computer is memory.
Other upgrades like card readers and zip drives have used USB. The last
few times they have thought about upgrading more they have decided
to buy a new computer. Sound familiar? I think that there are many
people who feel less intimidated if the computer was the size of a long
hardbound book. My mom would like it in violet.
The second case is business. A smaller computer would mean less desk
space or floor space waisted. It would also take less power and would
use less air conditioning. With energy costs the way they are, this
is a big concern. A computer with less bells and whistles might mean
fewer distractions for the employee. And again, for a group, 20-100, of
computers what was the last upgrade you added inside the case? Memory?
Maybe a new disk drive? What I see is that when the company decides to
upgrade their computers, they just buy new computers. At $300 a unit,
it not much more than a nice office chair or desk.
What I am surprised at is that none of the major computer vendors
have made a product like this.
Just my thoughts.
RLH
Forget the gamers and desk top users. For those users just put the RAM in the system.
Think mail servers, database servers, web servers and e-commerce servers. Any application that does heavy I/O. Real I/O, not just swap or paging. In particular any application that is write heavy. For true write it to the disk activities, kernel buffers if they get used, and they should not, are a loose because the data is not committed to non-volatile memory. Wrote the data to disk. Ya, right. Oops, the system crashed. You want it back? Well we were only kidding about the write, sorry (;-(.
Sendmail for example is very write intensive. Write each message to disk before you acknowledge receiving the message. Sendmail's mqueue directory normally is the performance bottle neck of a sendmail mail server. Since most of the files that sendmail writes are small, less than 1Kb, seek time is the important spec, not transfer speed. When you compare access time between memory and disk, memory blows disk out of the water. It is odd that Gigabyte does not mention access time in their data sheet. Not even a range of access times based on RAM speed. I would think it would be an important selling point.
Another use of this type of device is any journalling application such as a database or perhaps a journalling file system. If you can split the journal from the underlying data, then you could put the journal on the solid state disk and then let the application write the data to slow magnetic disk in the background.
Now what gigabyte or some other enterprising company should do with this type of product is:
1) Make sure it works with Linux (well maybe that other OS, Windoze, too) and make sure it is rock-solid reliable.
2) Support ECC RAM with error recovery. ECC is not an option on a server.
3) Repackage it in a normal 3.5 inch SATA enclosure with the connectors in the correct place. Who wants to put a PCI card into a 1U rack mount server? Maybe even have the option of a bigger battery although battery time is not really an issue with a server.
4) Support bigger DIMMs, design the controller to allow 4G, 8G, or larger DIMMs, but only "officially" support 1G or 2G DIMMs. I don't think that for many server applications size is that big of an issue. The really heavily accessed data quite often can be highly localized. The other thing is that you can always RAID several of these disks together.
5) Work on the pricing and package. You are going to put the card in a metal box, so charge more. Maybe in the $250 to $400 range with out memory. Also offer it as a complete package with memory, and charge more. These days since a 1G DIMM costs less than $100, you should charge at least $150 per Gig. Humm... 4G * $150 + $250 = $850. Less than $1K for a 4G screaming fast hard disk. We are talking micro-second access time, not milli-second access time. That is a product I could sink my teeth into.
A bit pricey for the enthusiast or the early adopter though. Maybe you could offer bare bone evaluation/hobbyist units for $150. Limit it to two units with web based support only. Get the bleeding edge experts to bang at it for you to find the bugs and reliability issues. And if it is a good product, they will start a buzz for the product. Just look at what is happening here on slashdot with the definitely marginal Gigabyte product.
Just my 4.2 CentOSs
RLH
Robert Harker
harker-slashdot@harker.com