(I'm a woodworking geek as well as a computer geek. Interestingly, I noticed that in the past year or so my brain now defaults to the power tool as opposed to the network traffic management device when it hears "router". Heh.)
It's even worse here in the UK. We pronounce router as in "rooter" (because it "routes") for the network device, but as in "raowter" (because it "routs") for the power tool. I've worked for US IT firms, so my brain automatically converts US->UK pronunciation for the network device.
I've just spent two weeks making sawdust with the powertool, which I've been refering to as; the "rooter-no-raowter", the "roo-aowter" and "that thing there". No doubt this week I'll be dealing with "raow-ooters" and "that box there, with the flashing lights".:)
Actually I did have this happen once in all my years with a Debian stable production box
I once saw a vendor approved patch bring down a financial market running on some very big iron. The amount of money lost in compensation to the traders was many, many times the cost of the hardware involved. Automatic patching was the root cause - except rather than a shell script being to blame, the idiot was there in person, downloading patches and whacking them straight on.
They failed to do any testing whatsoever before applying them to production machines. This is an often overlooked part of availability engineering best practices.
A database is the last place you would need a supercomputer. This is one of the easily parellelised jobs...
True for read-only databases, or those with simple writes and easily partitionable data. For decent performance and availability on mixed work load databases, clustered large SMP boxes still rule the roost.
Oracle are pushing RAC primarily because of the outrageous license fees they gain when you grow the cluster. The RAC licenses are $20K per CPU over and above the outrageous standard Enterprise Edition prices. For large databases, the Oracle licences make the big iron look cheap.
I agree. This was a pointless exercise, particularly since he used GCC but mainly because we've known this ever since Solaris 7 shipped for UltraSPARC I.
Not only a fraction of the price, but a fraction of the megahertz as well. I used to disregard the MHz Myth as Apple told it, but a gap like this, with apples-to-apples benchmarks showing them neck-and-neck
There is no such thing as an apples-to-apples benchmark unless you happen to be weighing apples. Every chip architecture is different and will perform in different ways. A 64-bit apple PowerPC chip isn't the same as those that go into an IBM pSeries. A 64-bit Sun workstation UltraSPARCIII chip isn't the same chip that goes in an F15K. A 3GHz Xeon will spank a lot of chips on integer performance, but it's FP performance sucks.
The clock speed is irrelevant. Benchmarks are mostly irrelevant because they can't possibly model your application's workload.
You aren't getting future proofing from 64-bit CPU's (especially since some of these low-end 64-bit boxes can't even take 4GB of memory), you're only getting major performance improvements on applications which are either: a) specifically developed for this architecture or b) exhibit memory address space bottlenecks, i.e. for most people today, relational databases. Memory capacity, memory bandwidth and IO performance then become the dominant factors in scalability.
A fast chip which can't be fed with useful data will just spend it's time running nops. A CPU architecure which is not designed to scale beyond 8 CPUs might also become a constraint.
Not the only fast moving hardware developments going on sadly.
Actually video cards are OK these days (not out of the box - you need the Linux XFree86 compatability kit). The real problem is things like WLAN cards, scanners, ADSL modems, USB printers, PC cards etc.
I guess there are Sun-only places where this might be a big deal. I'm also guessing that they're in a minority, so what does Sun see in it all ? It must be a reasonably large cost to maintain another OS for a company, so there has to be an upside... Answers on a postcard, please:-)
Practically nobody uses Solaris x86 commercially (yet) - this was the reason they were planning to drop it a couple of years ago. Times have changed, Sun have a couple of newish Xeon based boxes out (really intended to run Linux not Solaris, but they'll do that too) plus some blades. The boxes to watch are the Opteron based systems coming out next year. I have reason to believe they will be priced *very* competitively.
Solaris is 95+% platform independent, porting to a new architechture is not that big a deal - keeping up with the fast moving ecosystem of x86 hardware is a real pain, which is why they've not really been interested in x86 to date. Sun makes their money on selling tin not software.
It used to be the standard method of dealing with Microsoft Service Packs that you never deployed the latest one on your boxes. You always stayed one step behind. This practice was proved right with the Service Pack 6/6a debacle.
+4 Insightful, not interesting (this stuff is dull:)
I'm sorry, I just fail to see why this needs saying at all. The standard practice for any platform in an enterprise environment is to test the damn patch before you apply it. You test the patch on your test and QA systems - if you go straight from development to production what can you expect?
128 cpus? 2.6 kernel no recompile? awww, shucks, I'm running a 128 cpu box and I don't know how to recompile!
If you're running a 128 CPU box, lack of knowledge will not be your problem, SLA's will be. If Linux is in there, you *will* only get to use an 'enterprise' flavour of Linux or you're on your own. Redhat or SuSE. You can't recompile your kernel even if you wanted to (not that you would) or you'd lose support.
Is Sun selling Solaris separate from 128 cpu boxes? Or are they installing Solaris on those boxes when setting them up for customers? Is IBM setting up linux on their 128 processor boxes? Or are they selling 128 processor boxes and handing the operating system to customers in boxes, requiring customers to recompile?
Hot swap? Who gives a rat's ass? Haven't you seen the latest sales? Big iron is out, clustering is in. You don't need hotswap anything when clustering, that includes drives. Just ask Oracle.
Let me tell you as someone who has just spent the last 3 weeks evaluating Oracle RAC for a major outsourcing company. My recommendation will be: stick to plain Oracle on mid-range Sun hardware with FOM software, this stuff is waay too immature and it sucks badly for even moderate OLTP workloads. Extended distance clustering? Forget it.
You pick the absolute smallest part of the market, 128 cpu boxes, which in some quarters absolutely no company sells, and use that to slam linux over the entire server market? Get a life.
The smallest part of the market has the most money to spend and are often extremely loyal. No one in their right mind deploys mission critical applications on a Solaris instance with that many CPU's because CPU's have about the worst MTBF after disks and PSUs - stick 128 CPUs in there and you'll be rebooting every few months! You deploy these boxes underspec'ed, partition them and dynamically add and remove boards between them as the business requires.
Let us know when Solaris fits in less than 1 MB of space, when Solaris is running on cell phones, when Solaris is used as device drivers, when Solaris is used in routers, when Solaris is used in mesh networks, when Solaris is used in embedded devices, when Solaris is used in consumer electronics, when Solaris...
Solaris isn't designed for those applications. Neither is windows (just look at the train wreck that is PocketPC), neither are the BSDs, neither is Linux. Is kernel 2.6 going to fit in 1MB? I'd be surprised, it was hard enough getting a 2.4 kernel with PCMCIA and soundcard support + libm and mpg123 onto a 1.4MB floppy disk 3 years ago.
You're confusing open source with open systems. The interfaces *must* be open, the source is nice to have open. You'd be mad to deploy an enterprise UNIX on consumer devices and even madder to do the reverse.
I find this parallel interesting. Homosexuality is arguably natural occurring but atypical (I choose those words carefully) - the same could be said for paedophiles. I would be surprised if they are a historically recent phenomena and they certainly make up a very small percentage of the population, yet their actions and desires are abhorent to the vast majority of us.
The same can be said of homosexuality 100 years ago.
I personally find paedophilia abhorant and believe the world would be a better place without them (first to the wall after wedding photographers and management consultants:) but of course, that's what Hitler thought about several groups of people.
I heard an argument (BBC Radio4, "Moral Maze", ages ago) which claimed that paedophiles are 9 times more likely than a random selection to be homosexual, but I can't find anything else to back that up.
Given that no one 'chooses' to be a paedophile, the argument, I suppose boils down to; are you born a paedophile, do you 'become' one through your environment (hormones in your mother's womb/child abuse/bullying/whatever), or is it an artifact of society's condemnation of homosexuality.
Because x86 outperforms sparc, unless I buy a really expensive piece of hardware with a lot of processors, but if I'm in that market why should I buy from you when I can buy from IBM and get the better thing?
If it's a good fit for your IT environment and you're happy with their roadmap then go ahead, buy IBM. Availability is about people and process, not just product, not everyone will find IBM's offering (or Sun's or HP's) an easy fit into their organisation.
Because I'm the customer and that's what I want. I'll worry about the apps, you give me what I ask for. BTW, I want apps on Linux so you better start leaning on your vendors to port them if I buy your sparc solution. I have the money you want, do what I ask and I'll give it to you, simple rules, eh?
You may be under the misapprehension that I work for Sun. I don't (I used to, and I'm still good friends with a lot of guys there). Sun's existing customers are not asking for Linux on SPARC, they're asking for linux on x86 based boxes and only on low end rack and stack kit. There may be a customer base which is interested in running Linux on SPARC in the future, but they are certainly keeping quiet (at least in the vertical markets which I and my friends are exposed to). If the ISV's thought there was a market for SPARC linux software, I'm sure there would be at least one example without being leaned on. There isn't.
Because I'm the customer and that's what I want. Trusted Solaris has had multiple security problems and I want to virtualize for security reasons, plus it's good enough for java so why not something else?
Yes, I believe you mentioned that you're a 'potential' customer. All operating systems have 'multiple security problems' many of which haven't been found yet, as I'm sure you know. You trust your virtualization software implicitly. I'm sure. I know I would.
I can do it on a z server and it outperforms Sun's enterprise hardware in every measure.
Except price.
...Sun should spend more time listening to their customers and less time listening to their prima donas. If a customer wants VMs on Solaris with linux in it (clue, they are giving solaris a life line you dumb asses! they are hedging on sun and you don't want to take it?!?) then they should make it. If a customer wants x86 solaris or Linux on sparc they should bloody well start providing those.
Sadly, some of Sun's customers are prima donnas (they work for investment banks mostly). Just out of curiosity, have you ever used Solaris x86? It's crap! The hardware support is pathetic, the install procedure is arcane if you have anything but very mainstream hardware and the performance sucks (again, mainly because of poor hardware support). Sun is not in the business of supporting esoteric PC hardware, they built Solaris x86 to support particular models of popular PC hardware (compaq, HP, fujitsu etc.) - and nobody bought them.
Linux on SPARC exists but was developed outside of Sun and only really runs on EOL kit. From my experience, Sun have absolutely no interest in SPARC linux - none what so ever - they're primary focus is on Solaris on SPARC - I happen to think that's a good thing.
And of course, everyone runs Linux on their RS/6000 boxes...
advertise! not to try and convert new customers, but to your existing core market. "think different" was all about consolidation.
Different target markets. You will find Sun marketing material in professional IT journals, at trade shows etc. They don't, in general advertise to the general public, because only a tiny proportion of the general public are in the position to sign PO's. Brand awareness is being built through corporate sponsorship.
fire some high-level people. just enough to get into time magazine.
One of Sun's problems is that they hired a lot of ex-IBM management during the 90s. They are very good at making themselves look invaluable to the business (plus they ruined the culture, but that's another story).
come up with something new and interesting - even if it's just packaging. hint: thin clients aren't interesting.
I'd say the blades and N1 provisioning server, the new UltraSPARCIIIi boxes, JES, UltraSPARCIV, Solaris 10 are all pretty cool - but maybe that's just me. Are you suggesting Sun come up with some sort of music player or PDA? (actually they did have plans for a PDA once...:)
foster a sense of elitism and cool amongst yr customer base. good lord, high school kid's have computers with the dell log on the front. this should be easy.
A sense of elitism among your customer base is often associated with arrogance on the part of the vendor. Sun have been accused of arrogance before and it's an image they are keen to avoid. High school kids don't have much purchasing power and Sun really concentrates on the high margin, low volume end of the market.
What always amuses me about slashdot is that both Apple and Sun develop and sell proprietry software, both develop and sell proprietry hardware with decent margins, both dislike Microsoft. Yet Sun is evil and Apple is holy.
However, you can "simulate" a single high available server with a bunch of commodity hardware and heartbeat monitoring quite well - think google.
No you can't. That only works on stateless protocols with simple client software. Try 'simulating' a payroll, SCM, risk analysis, ERP system - or indeed any other application which does real work on a medium to large data set. It's not possible to achieve high availability on a non-trivial database without using big iron, industrial strength storage arrays, mature DBMS products and a commercial FMS (VCS, SunCluster).
Oh, and what OSS disaster recovery solution did you have in mind?
Sun is a systems company, not just hardware or software. Solaris rocks and will continue to rock for many years to come (I'm on the Solaris 10 beta programme and there's some really cool new stuff in there). In an enterprise environment there is nothing that can touch it - JumpStart, processor sets, domaining, DR, HA Clustering etc. brilliant technology.
Performance is not a hugely important factor in enterprise computing - availability, manageability and support are much, much more important. Sun's products are designed to address these concerns and priced accordingly. SPARC development is a concern, but that's because the timescales are slipping, not because the ideas aren't there - something that can be rectified by adding more engineering resource, Solaris (or Java, or systems) development need not be affected.
When the global economy picks up, I predict that Sun and IBM will continue taking market share from HP, because they've really lost the plot in the UNIX server market.
Slashdot Mirror
(I'm a woodworking geek as well as a computer geek. Interestingly, I noticed that in the past year or so my brain now defaults to the power tool as opposed to the network traffic management device when it hears "router". Heh.)
:)
It's even worse here in the UK. We pronounce router as in "rooter" (because it "routes") for the network device, but as in "raowter" (because it "routs") for the power tool. I've worked for US IT firms, so my brain automatically converts US->UK pronunciation for the network device.
I've just spent two weeks making sawdust with the powertool, which I've been refering to as; the "rooter-no-raowter", the "roo-aowter" and "that thing there". No doubt this week I'll be dealing with "raow-ooters" and "that box there, with the flashing lights".
Actually I did have this happen once in all my years with a Debian stable production box
I once saw a vendor approved patch bring down a financial market running on some very big iron. The amount of money lost in compensation to the traders was many, many times the cost of the hardware involved. Automatic patching was the root cause - except rather than a shell script being to blame, the idiot was there in person, downloading patches and whacking them straight on.
They failed to do any testing whatsoever before applying them to production machines. This is an often overlooked part of availability engineering best practices.
To quote futurama:
:)
You're technically correct - the best kind of correct.
A database is the last place you would need a supercomputer. This is one of the easily parellelised jobs...
True for read-only databases, or those with simple writes and easily partitionable data. For decent performance and availability on mixed work load databases, clustered large SMP boxes still rule the roost.
Oracle are pushing RAC primarily because of the outrageous license fees they gain when you grow the cluster. The RAC licenses are $20K per CPU over and above the outrageous standard Enterprise Edition prices. For large databases, the Oracle licences make the big iron look cheap.
...No I don't mean clusters. Clusters are a cheap hack to pretend to a non-existant level of reliability...
:).
In my experience, the reason you cluster mainframes is primarily to minimize licensing costs
...a new interface is never as intuitive as the old one you are used to...
Not intuitive, familiar. As a wise man once said: the nipple is the only intuitive interface.
As a breast man I agree.
Sun will only support Solaris x86 in 32bit mode on the x86-64 systems.
Not true. I'm told we can expect Solaris 9 x86 64 bit support in the second half of 2004, in one of the quarterly Solaris update releases.
They'd be retarded to undercut the flagship UltraSPARC line of servers...
The SunFire systems have a different target market with different scalability, availability and managability requirements.
I agree. This was a pointless exercise, particularly since he used GCC but mainly because we've known this ever since Solaris 7 shipped for UltraSPARC I.
This guy gets the 'No Shit, Sherlock' award.
Not only a fraction of the price, but a fraction of the megahertz as well. I used to disregard the MHz Myth as Apple told it, but a gap like this, with apples-to-apples benchmarks showing them neck-and-neck
There is no such thing as an apples-to-apples benchmark unless you happen to be weighing apples. Every chip architecture is different and will perform in different ways. A 64-bit apple PowerPC chip isn't the same as those that go into an IBM pSeries. A 64-bit Sun workstation UltraSPARCIII chip isn't the same chip that goes in an F15K. A 3GHz Xeon will spank a lot of chips on integer performance, but it's FP performance sucks.
The clock speed is irrelevant. Benchmarks are mostly irrelevant because they can't possibly model your application's workload.
You aren't getting future proofing from 64-bit CPU's (especially since some of these low-end 64-bit boxes can't even take 4GB of memory), you're only getting major performance improvements on applications which are either: a) specifically developed for this architecture or b) exhibit memory address space bottlenecks, i.e. for most people today, relational databases. Memory capacity, memory bandwidth and IO performance then become the dominant factors in scalability.
A fast chip which can't be fed with useful data will just spend it's time running nops. A CPU architecure which is not designed to scale beyond 8 CPUs might also become a constraint.
Not the only fast moving hardware developments going on sadly.
Actually video cards are OK these days (not out of the box - you need the Linux XFree86 compatability kit). The real problem is things like WLAN cards, scanners, ADSL modems, USB printers, PC cards etc.
You with the zink cardigan, are you English?
I guess there are Sun-only places where this might be a big deal. I'm also guessing that they're in a minority, so what does Sun see in it all ? It must be a reasonably large cost to maintain another OS for a company, so there has to be an upside... Answers on a postcard, please :-)
Practically nobody uses Solaris x86 commercially (yet) - this was the reason they were planning to drop it a couple of years ago. Times have changed, Sun have a couple of newish Xeon based boxes out (really intended to run Linux not Solaris, but they'll do that too) plus some blades. The boxes to watch are the Opteron based systems coming out next year. I have reason to believe they will be priced *very* competitively.
Solaris is 95+% platform independent, porting to a new architechture is not that big a deal - keeping up with the fast moving ecosystem of x86 hardware is a real pain, which is why they've not really been interested in x86 to date. Sun makes their money on selling tin not software.
The only two relevant platforms on IA32 are Windows and Linux.
You are absolutely right. What you overlook is that IA32 is a platform which is already starting to become less and less relevant in the enterprise.
> Needle Nardle Noo
Would you like a gorilla?
It used to be the standard method of dealing with Microsoft Service Packs that you never deployed the latest one on your boxes. You always stayed one step behind. This practice was proved right with the Service Pack 6/6a debacle.
:)
+4 Insightful, not interesting (this stuff is dull
I'm sorry, I just fail to see why this needs saying at all. The standard practice for any platform in an enterprise environment is to test the damn patch before you apply it. You test the patch on your test and QA systems - if you go straight from development to production what can you expect?
128 cpus? 2.6 kernel
no recompile? awww, shucks, I'm running a 128 cpu box and I don't know how to recompile!
If you're running a 128 CPU box, lack of knowledge will not be your problem, SLA's will be. If Linux is in there, you *will* only get to use an 'enterprise' flavour of Linux or you're on your own. Redhat or SuSE. You can't recompile your kernel even if you wanted to (not that you would) or you'd lose support.
Is Sun selling Solaris separate from 128 cpu boxes? Or are they installing Solaris on those boxes when setting them up for customers? Is IBM setting up linux on their 128 processor boxes? Or are they selling 128 processor boxes and handing the operating system to customers in boxes, requiring customers to recompile?
Hot swap? Who gives a rat's ass? Haven't you seen the latest sales? Big iron is out, clustering is in. You don't need hotswap anything when clustering, that includes drives. Just ask Oracle.
Let me tell you as someone who has just spent the last 3 weeks evaluating Oracle RAC for a major outsourcing company. My recommendation will be: stick to plain Oracle on mid-range Sun hardware with FOM software, this stuff is waay too immature and it sucks badly for even moderate OLTP workloads. Extended distance clustering? Forget it.
You pick the absolute smallest part of the market, 128 cpu boxes, which in some quarters absolutely no company sells, and use that to slam linux over the entire server market? Get a life.
The smallest part of the market has the most money to spend and are often extremely loyal. No one in their right mind deploys mission critical applications on a Solaris instance with that many CPU's because CPU's have about the worst MTBF after disks and PSUs - stick 128 CPUs in there and you'll be rebooting every few months! You deploy these boxes underspec'ed, partition them and dynamically add and remove boards between them as the business requires.
Let us know when Solaris fits in less than 1 MB of space, when Solaris is running on cell phones, when Solaris is used as device drivers, when Solaris is used in routers, when Solaris is used in mesh networks, when Solaris is used in embedded devices, when Solaris is used in consumer electronics, when Solaris...
Solaris isn't designed for those applications. Neither is windows (just look at the train wreck that is PocketPC), neither are the BSDs, neither is Linux. Is kernel 2.6 going to fit in 1MB? I'd be surprised, it was hard enough getting a 2.4 kernel with PCMCIA and soundcard support + libm and mpg123 onto a 1.4MB floppy disk 3 years ago.
You're confusing open source with open systems. The interfaces *must* be open, the source is nice to have open. You'd be mad to deploy an enterprise UNIX on consumer devices and even madder to do the reverse.
If you look over the credible research, theres no real link between homosexuality and pedophilia.
Credible research would be peer reviewed, and published in a respected journal. There doesn't seem to be much research at all.
I find this parallel interesting. Homosexuality is arguably natural occurring but atypical (I choose those words carefully) - the same could be said for paedophiles. I would be surprised if they are a historically recent phenomena and they certainly make up a very small percentage of the population, yet their actions and desires are abhorent to the vast majority of us.
:) but of course, that's what Hitler thought about several groups of people.
The same can be said of homosexuality 100 years ago.
I personally find paedophilia abhorant and believe the world would be a better place without them (first to the wall after wedding photographers and management consultants
I heard an argument (BBC Radio4, "Moral Maze", ages ago) which claimed that paedophiles are 9 times more likely than a random selection to be homosexual, but I can't find anything else to back that up.
Given that no one 'chooses' to be a paedophile, the argument, I suppose boils down to; are you born a paedophile, do you 'become' one through your environment (hormones in your mother's womb/child abuse/bullying/whatever), or is it an artifact of society's condemnation of homosexuality.
Fact: There were 1 million downloads of iTunes for Windows.
/tmp/iTunesSetup.exe
/tmp/iTunesSetup.exe
At least one of those was me:
$ winex3
(error message)
$ rm -f
Oh well.
Because x86 outperforms sparc, unless I buy a really expensive piece of hardware with a lot of processors, but if I'm in that market why should I buy from you when I can buy from IBM and get the better thing?
...Sun should spend more time listening to their customers and less time listening to their prima donas. If a customer wants VMs on Solaris with linux in it (clue, they are giving solaris a life line you dumb asses! they are hedging on sun and you don't want to take it?!?) then they should make it. If a customer wants x86 solaris or Linux on sparc they should bloody well start providing those.
If it's a good fit for your IT environment and you're happy with their roadmap then go ahead, buy IBM. Availability is about people and process, not just product, not everyone will find IBM's offering (or Sun's or HP's) an easy fit into their organisation.
Because I'm the customer and that's what I want. I'll worry about the apps, you give me what I ask for. BTW, I want apps on Linux so you better start leaning on your vendors to port them if I buy your sparc solution. I have the money you want, do what I ask and I'll give it to you, simple rules, eh?
You may be under the misapprehension that I work for Sun. I don't (I used to, and I'm still good friends with a lot of guys there).
Sun's existing customers are not asking for Linux on SPARC, they're asking for linux on x86 based boxes and only on low end rack and stack kit. There may be a customer base which is interested in running Linux on SPARC in the future, but they are certainly keeping quiet (at least in the vertical markets which I and my friends are exposed to). If the ISV's thought there was a market for SPARC linux software, I'm sure there would be at least one example without being leaned on. There isn't.
Because I'm the customer and that's what I want. Trusted Solaris has had multiple security problems and I want to virtualize for security reasons, plus it's good enough for java so why not something else?
Yes, I believe you mentioned that you're a 'potential' customer. All operating systems have 'multiple security problems' many of which haven't been found yet, as I'm sure you know. You trust your virtualization software implicitly. I'm sure. I know I would.
I can do it on a z server and it outperforms Sun's enterprise hardware in every measure.
Except price.
Sadly, some of Sun's customers are prima donnas (they work for investment banks mostly). Just out of curiosity, have you ever used Solaris x86? It's crap! The hardware support is pathetic, the install procedure is arcane if you have anything but very mainstream hardware and the performance sucks (again, mainly because of poor hardware support). Sun is not in the business of supporting esoteric PC hardware, they built Solaris x86 to support particular models of popular PC hardware (compaq, HP, fujitsu etc.) - and nobody bought them.
Linux on SPARC exists but was developed outside of Sun and only really runs on EOL kit. From my experience, Sun have absolutely no interest in SPARC linux - none what so ever - they're primary focus is on Solaris on SPARC - I happen to think that's a good thing.
And of course, everyone runs Linux on their RS/6000 boxes...
Because for some operations, such as a CVS server, Solaris on x86 is rather faster than Solaris on SPARC. Linux is faster still.
Use linux on x86 then. It's clearly the right tool for the right job.
if ms offers you money, take it!
:)
Err, can't quite see that happening.
advertise! not to try and convert new customers, but to your existing core market. "think different" was all about consolidation.
Different target markets. You will find Sun marketing material in professional IT journals, at trade shows etc. They don't, in general advertise to the general public, because only a tiny proportion of the general public are in the position to sign PO's. Brand awareness is being built through corporate sponsorship.
fire some high-level people. just enough to get into time magazine.
One of Sun's problems is that they hired a lot of ex-IBM management during the 90s. They are very good at making themselves look invaluable to the business (plus they ruined the culture, but that's another story).
come up with something new and interesting - even if it's just packaging. hint: thin clients aren't interesting.
I'd say the blades and N1 provisioning server, the new UltraSPARCIIIi boxes, JES, UltraSPARCIV, Solaris 10 are all pretty cool - but maybe that's just me. Are you suggesting Sun come up with some sort of music player or PDA? (actually they did have plans for a PDA once...
foster a sense of elitism and cool amongst yr customer base. good lord, high school kid's have computers with the dell log on the front. this should be easy.
A sense of elitism among your customer base is often associated with arrogance on the part of the vendor. Sun have been accused of arrogance before and it's an image they are keen to avoid. High school kids don't have much purchasing power and Sun really concentrates on the high margin, low volume end of the market.
What always amuses me about slashdot is that both Apple and Sun develop and sell proprietry software, both develop and sell proprietry hardware with decent margins, both dislike Microsoft. Yet Sun is evil and Apple is holy.
However, you can "simulate" a single high available server with a bunch of commodity hardware and heartbeat monitoring quite well - think google.
No you can't. That only works on stateless protocols with simple client software. Try 'simulating' a payroll, SCM, risk analysis, ERP system - or indeed any other application which does real work on a medium to large data set. It's not possible to achieve high availability on a non-trivial database without using big iron, industrial strength storage arrays, mature DBMS products and a commercial FMS (VCS, SunCluster).
Oh, and what OSS disaster recovery solution did you have in mind?
1) Does Sun support x86 for Solaris?
Yes, but why would you want to, when you can have the real thing on SPARC.
2) Does Sun support Linux on Sparc?
No, why would you want to. Where are the applications? (I mean Enterprise applications, not desktop stuff).
3) Is Linux good, or bad?
Good for the low-end, good for annoying Microsoft.
4) Why can't you run multple Linux VMs on a single Solaris O/S?
Why would you want to. What benefit does that give you over say, linux on blades.
Top tip. Ignore the marketing - listen to your Sun SE, he'll know what's really going on.
Sun is a systems company, not just hardware or software. Solaris rocks and will continue to rock for many years to come (I'm on the Solaris 10 beta programme and there's some really cool new stuff in there). In an enterprise environment there is nothing that can touch it - JumpStart, processor sets, domaining, DR, HA Clustering etc. brilliant technology.
Performance is not a hugely important factor in enterprise computing - availability, manageability and support are much, much more important. Sun's products are designed to address these concerns and priced accordingly. SPARC development is a concern, but that's because the timescales are slipping, not because the ideas aren't there - something that can be rectified by adding more engineering resource, Solaris (or Java, or systems) development need not be affected.
When the global economy picks up, I predict that Sun and IBM will continue taking market share from HP, because they've really lost the plot in the UNIX server market.