Just cos they're taking advantage of what people want now (Linux, Opteron, Open Source) doesn't mean they're not also working on stuff that's cool that we don't know that we want yet, or even stuff that's not cool but is still worthy.
This is where Sun, IBM, SGI, even HP, do more for us than Dell and Microsoft. Though at least, and I hate myself for saying this, Microsoft are trying.
Cleary being first or having the best idea ever are no guarantees of esteem or profit - often the opposite, so kudos to Sun for slugging it out and continuing to bet on innovation. Ditto to IBM and AMD.
I think they're particularly looking at things like the C10K problem (http://www.kegel.com/c10k.html).
The new Solaris 10 networking code reputedly pays a lot of attention to exploiting, and serving threads well, particularly hardware multithreading if it's available.
If they could squeeze one of these and maybe 8GB+ of RAM into a 1U box or into their blade centre, then I think it'd do quite nicely for serving web.
It's not cos my fat arse doesn't fit mind but cos arms dig in to your thighs (c'mon, men, you know how we sit) and they get stuck on the edge of the desk.
Now, foot-rests, what the hell are they about? I mean, you end up rolling your chair back all the time, and that ain't useful.
The Microprocesor Report summarised the CPI vs MHz approaches to improved CPU performance not so long ago.
Their studied and authoritative conclusion was that process improvements over the last decade have allowed clockspeeds to drive performance up faster than any CPI improvements like superscalar designs or speculative execution.
There is no denying marketecture is the major driver for consumer computers, peecees and macintrashes, but there is a a real technical argument for it too.
An exmaple: IBM POWER designs have recently made a strong shift from a strong CPI focus to a very strong clockspeed focus (POWER 3 around 300 Mhz, POWER 4 around 2GHz), and they generally sell to a technically savvy market that's not so subject to the lure of marketecture.
NAS is more mature than SAN, also SAN products aren't and won't become available for some of the platforms mentioned, of that I'm confident.
My suggestion would be to use a low-end NetApp if it's in the price-range or a Solaris x86 box coupled with a RAID system like the Arena boxes you can get from Zero-D.
NetApp gives you the gamut of NAS facilities, including snapshots, RAID, Integrated Unix and NT permissions, and your choice of access protocols.
The x86 option lets you build your own, exactly as you would with Linux or Freebeard, but also nets you a working lockd, and a comprehendable, consistent, and on-line tunable system, including the network stack.
Obviously someone is going to flame me for this, but I'm wearing asbestos undies. 8)
D-Link do/did a 5-port portable hub that can take its power from a PS2 or DIN keyboard/mouse connector that was purpose designed for such arrangements.
I've run my old Matrox Millennium under most of the OS's named. I do 1280x1024x16bpp because I couldn't afford an 8MB card at the time. Works a treat though, especially using Xfree under solaris (such a bonus compared to the BrokenWindows X server).
Any Matrox card 8MB or better in the Millennium lineage (Millennium, Millennium II, G100, G200, G400) should do the trick. Finding the price shouldn't be too hard either, and getting a more recent card has obvious attractions.
I suppose I didn't make clear that I consider that there's a difference between tapping the wire (the physical medium) and doing 'wiretapping' in active devices like routers and switches.
Think about it for a while. Done right only quantum crypto defeats tapping the wire, while wiretapping higher-level protocols relies on a whole mess of technical know-how at the tap-point, (rather than back at the spook-cave) and collusion with multiple bodies (both human and corporate).
I don't think that anyone who's sufficently knowledgeable or paranoid has ever believed that spooks couldn't already tap IP communications. Just not in the way most IP-familiar techies would choose to do it.
The Echelon *email* concerns have always struck me as an unfeasible approach, given tapping the wire itself is (or at least has been) so much more achievable than getting ISPs to help the spooks in an organised fashion.
I wish I could recall the URL for the public guardians_of_the_law-ISP dialogue that went on in the UK a few months back, made this whole set of points about ISPs incurring costs for spook-work and jurisdictional difficulties and lack of guardians_of_the_law technical know-how.
And I also recall thinking how it was all a blind, given the spooks can almost certainly do all this stuff when they want to anyway.
To be honest it must be like herding cats getting the ISPs to pitch in when the spooks want, but the major carriers and infrastructure companies...they can be arm-twisted much more effectively.
Certainly that's the situation that sems to pertain here in the UK with BT, GCHQ, the NSA and the old-boys network.
The IETF, as a body of erudite folk, knows that it can specify, and pontificate and stay well on the side of right, (well, spooks are sinister aren't they?) and get away with it because the spooks have other ways to get what they want. Heck even though the IETF tries to be de jure, the Interenet itself tends to be de facto so whetever will be, will be.
Guess we'll need IPsec, and ssh and whatever else we can get even more than ever now the router giants are kow-towing along with the wire-owners.
Doesn't everyone and his dog recall the news stories over the last few months about the HP researchers building huge reconfigurable arrays of FPGA's and getting stonking performance out of them even when there were high numbers of defunt chips in the mix?
There's also at least one company producing circuit simulation platforms hundreds/thousands of times faster than pure software simulation platforms, for the IC design industry.
What marks these guys out is that they can write hype with the very best Microsofties.
Heck what do I know, they might be the same guys after some marketing courses.
I've long held that a quantum-noise RNG should be on-chip. It should be possible to read a *real* RN from a register every clock.
I'd also love there to be a UTC register so gettimeofday() would be trivial and no longer the performance monster in Motif etc. that it currently is.
The unique ID has no detractors, the press is being groundlessly alarmist. If it's used by OS's and software we write, we trust and we want to use it's fine, even potentially useful. If not, then it won't be used at all.
I'd love to know if it'll push the price of the chips up a few pence though - after all uniqueness is anathema to mass production.
I'd also like to know if there's any hardware included in the RNG to test for the characteristics of randomness - just in case Intel make mistakes, or there are manufacturing flaws, we need to know we can depend on the RNG, or it too will become a costly waste of space and time.
Altogether though, this stuff makes me happy that Intel have done something genuinely useful and relatively innovative.
Slashdot Mirror
..back on the statutes in the US eh.
Brocade have been the major player in Fibre Channel since that kicked off years ago.
Micron are long-time players in the memory field.
Cos that's what x2vnc likes best.
Machines called 'toasters' (by the faithful) already based upon NetBSD (long long ago, see the aknowledgements in a DataOnTap manual near you).
..of something getting 1 day older?
Must be a slow news day.
Isn't there some sort of pseudo-democracy thing we could be scrutinizing instead?
How do you cope with the blank-page problem and times when the story seems to dry up in your mind? Does it ever happen to you?
Why do you write the way you do - long-hand, typewritten, word-processed, or dictation?
They do everything they used to do.
Just cos they're taking advantage of what people want now (Linux, Opteron, Open Source) doesn't mean they're not also working on stuff that's cool that we don't know that we want yet, or even stuff that's not cool but is still worthy.
This is where Sun, IBM, SGI, even HP, do more for us than Dell and Microsoft. Though at least, and I hate myself for saying this, Microsoft are trying.
Cleary being first or having the best idea ever are no guarantees of esteem or profit - often the opposite, so kudos to Sun for slugging it out and continuing to bet on innovation. Ditto to IBM and AMD.
I think they're particularly looking at things like the C10K problem (http://www.kegel.com/c10k.html).
The new Solaris 10 networking code reputedly pays a lot of attention to exploiting, and serving threads well, particularly hardware multithreading if it's available.
If they could squeeze one of these and maybe 8GB+ of RAM into a 1U box or into their blade centre, then I think it'd do quite nicely for serving web.
This fat folk takes the arms off the chairs.
It's not cos my fat arse doesn't fit mind but cos arms dig in to your thighs (c'mon, men, you know how we sit) and they get stuck on the edge of the desk.
Now, foot-rests, what the hell are they about? I mean, you end up rolling your chair back all the time, and that ain't useful.
Is anyone fixing the fundamental resolution limit in the protocol?
I saw the take off and the landing live on BBC News24 and it looked very smooth.
Apparently there may have been some slight damage to the nose, but Mike Melvill declared it a 'mind-blowing experience'.
Burt Rutan seems quite moved too.
Get an Aeron specifically for your height and weight (three main sizes istr). Lovely.
The Microprocesor Report summarised the CPI vs MHz approaches to improved CPU performance not so long ago.
Their studied and authoritative conclusion was that process improvements over the last decade have allowed clockspeeds to drive performance up faster than any CPI improvements like superscalar designs or speculative execution.
There is no denying marketecture is the major driver for consumer computers, peecees and macintrashes, but there is a a real technical argument for it too.
An exmaple: IBM POWER designs have recently made a strong shift from a strong CPI focus to a very strong clockspeed focus (POWER 3 around 300 Mhz, POWER 4 around 2GHz), and they generally sell to a technically savvy market that's not so subject to the lure of marketecture.
*cough* let me try again.
NAS is more mature than SAN, also SAN products aren't and won't become available for some of the platforms mentioned, of that I'm confident.
My suggestion would be to use a low-end NetApp if it's in the price-range or a Solaris x86 box coupled with a RAID system like the Arena boxes you can get from Zero-D.
NetApp gives you the gamut of NAS facilities, including snapshots, RAID, Integrated Unix and NT permissions, and your choice of access protocols.
The x86 option lets you build your own, exactly as you would with Linux or Freebeard, but also nets you a working lockd, and a comprehendable, consistent, and on-line tunable system, including the network stack.
Obviously someone is going to flame me for this, but I'm wearing asbestos undies. 8)
D-Link do/did a 5-port portable hub that can take its power from a PS2 or DIN keyboard/mouse connector that was purpose designed for such arrangements.
I've run my old Matrox Millennium under most of the OS's named. I do 1280x1024x16bpp because I couldn't afford an 8MB card at the time. Works a treat though, especially using Xfree under solaris (such a bonus compared to the BrokenWindows X server).
Any Matrox card 8MB or better in the Millennium lineage (Millennium, Millennium II, G100, G200, G400) should do the trick. Finding the price shouldn't be too hard either, and getting a more recent card has obvious attractions.
I suppose I didn't make clear that I consider that there's a difference between tapping the wire (the physical medium) and doing 'wiretapping' in active devices like routers and switches.
Think about it for a while. Done right only quantum crypto defeats tapping the wire, while wiretapping higher-level protocols relies on a whole mess of technical know-how at the tap-point, (rather than back at the spook-cave) and collusion with multiple bodies (both human and corporate).
What's a spook gonna do?
The Echelon *email* concerns have always struck me as an unfeasible approach, given tapping the wire itself is (or at least has been) so much more achievable than getting ISPs to help the spooks in an organised fashion.
I wish I could recall the URL for the public guardians_of_the_law-ISP dialogue that went on in the UK a few months back, made this whole set of points about ISPs incurring costs for spook-work and jurisdictional difficulties and lack of guardians_of_the_law technical know-how.
And I also recall thinking how it was all a blind, given the spooks can almost certainly do all this stuff when they want to anyway.
To be honest it must be like herding cats getting the ISPs to pitch in when the spooks want, but the major carriers and infrastructure companies...they can be arm-twisted much more effectively.
Certainly that's the situation that sems to pertain here in the UK with BT, GCHQ, the NSA and the old-boys network.
The IETF, as a body of erudite folk, knows that it can specify, and pontificate and stay well on the side of right, (well, spooks are sinister aren't they?) and get away with it because the spooks have other ways to get what they want. Heck even though the IETF tries to be de jure, the Interenet itself tends to be de facto so whetever will be, will be.
Guess we'll need IPsec, and ssh and whatever else we can get even more than ever now the router giants are kow-towing along with the wire-owners.
Score one for the spooks.
I recommend Cricket:
o l/
http://www.munitions.com/~jra/cricket/
which is built around
http://ee-staff.ethz.ch/~oetiker/webtools/rrdto
plus whatever other tools and bits and bobs you find/create/get pointed to by everyone else.
Doesn't everyone and his dog recall the news stories over the last few months about the HP researchers building huge reconfigurable arrays of FPGA's and getting stonking performance out of them even when there were high numbers of defunt chips in the mix?
There's also at least one company producing circuit simulation platforms hundreds/thousands of times faster than pure software simulation platforms, for the IC design industry.
What marks these guys out is that they can write hype with the very best Microsofties.
Heck what do I know, they might be the same guys after some marketing courses.
I've long held that a quantum-noise RNG should be on-chip. It should be possible to read a *real* RN from a register every clock.
I'd also love there to be a UTC register so gettimeofday() would be trivial and no longer the performance monster in Motif etc. that it currently is.
The unique ID has no detractors, the press is being groundlessly alarmist. If it's used by OS's and software we write, we trust and we want to use it's fine, even potentially useful. If not, then it won't be used at all.
I'd love to know if it'll push the price of the chips up a few pence though - after all uniqueness is anathema to mass production.
I'd also like to know if there's any hardware included in the RNG to test for the characteristics of randomness - just in case Intel make mistakes, or there are manufacturing flaws, we need to know we can depend on the RNG, or it too will become a costly waste of space and time.
Altogether though, this stuff makes me happy that Intel have done something genuinely useful and relatively innovative.