A Beowulf cluster can't do everything - that's why we have SMP machines. If you want a Beowulf cluster of Suns, you can have that too. Sun produce a wide range of kit, not just big SMP machines.
On the Linux front, I find the comments about Sun and Linux on Slashdot totally bizarre - one moment Sun are informed by Slashdot readers that Linux and Intel are the future, the next moment Sun are criticised for selling a server that addresses that part of the market.
There's a lawsuit going on with potentially large implications for Linux, but it's not clear at this stage - Sun say they're looking at the implications. How exactly does this make Sun 'fair weather friends'.
Have they dropped their Linux strategy, Linux blades, stoppped supporting the various Open Source projects, dropped their 100% Unix background and started selling NT boxes like Unix' other 'fair weather friends'? Thought not...
What do you mean by long term? There's a ten year road map for Solaris, there are hundreds of thousands of customers and Solaris excels over Linux in various areas.
The two OSes can sit together well in Sun's strategy.
I appreciate your point about Veritas Volume Manager, but I think the point of the post is that this functionality should be available, free, within Linux. Solaris has DiskSuite (or Solaris Volume Manager as it's been renamed in Solaris 9) for example. Considering everyone's whingeing about the 'cost' of Unix and advocating Linux as it's 'cheap' the fact you have to buy an expensive piece of volume manager software doesn't fit into the general scheme of things.
I was under the impression that Linux didn't scale very well over 4 cpus? Certainly you don't get linear scalability out of an 8 way Intel box running Linux, based on what I've read, but if you can point me to some evidence suggesting otherwise, cool.
But surely you should have heeded what people on Slashdot tell us is the thing to do - namely replace that 'expensive' Sun box with lots of 'cheap' Intel boxes in a cluster?
UIII is not 'slow', that's a childishly simplistic approach.
What's your application? On the 480 the internal bandwidth is 9.6Gbs, Level 2 cache on the CPUs is 8Mb, the cpus scale linearly. You have two on board copper Gb ethernet ports. You get Solaris for free, warranty is three years (probably the same as the Dell).
Also, you're not just buying a box, you're buying a box that's going to be part of the rest of your strategy - how well does it connect/support with other systems, can the vendor sell decent storage solutions around it, etc, etc.
I'd disagree that the low end Dell is going to be 'faster' than the low end Sun. That depends on the apps. Also, the 480 is priced similarly to the equivalent 4 cpu Dell, except the cpus scale linearly so I actually get to use 4 cpus for the money I've spent, the internal bandwidth is higher, build is higher quality, support is integrated for hardware and OS, I can run the same proven OS on that small box up to the F15k, I can get a large number of supported apps for Solaris, storage solutions are better, an excellent Professional Services organisation that can help me implement stuff and I have a partnership with a company that has vision and should be a decent partner with me as IT Manager as I roll stuff out throughout the business. If I really require Linux I can get LX50s and get my support from the same vendor as I do with the Sparc/Solaris boxes.
There's more to Sun vs Dell than 'my Dell is cheaper so Sun is dead'.
Shame Sun Rays aren't mentioned this time around. They'd still have a big place to play in the scenario the author's talking about - ability to hot desk around campus, zero maintenance, etc, etc.
What companies do you know that run TPC-C benhcmarks as their business application? Why should we care how well IBM optimise their compilers to do well in benchmarks that don't tell us that the app we want to run will run fast?
Why would you choose such a large server based on a TPC-C benchmark? What about the rest of the server's components, storage, services, etc, etc?
I recently moved from the UK to the Middle East. The situation in the Gulf States such as Qatar, Oman, the UAE, etc, is that the local citizens of each state make up, on average, about 20% of the population of the country. 'Locals', meaning the indigenous Arab population, generally work for the government in managerial positions and the majority of hands on IT work is performed by people from every country in the world, from Kyrgystan to South Africa, to Great Britain, to Somalia, to India and everywhere in between.
Certifications in this region are extremely important. As a manager looking to employ someone, whether in a small or large company, you have to have some benchmark to use before you even interivew someone for a hands on IT role as you have no way of knowing what the general standard is in Iran, or Somalia or England, so there is no point interviewing everyone who claims they have a CS degree from the University of Tehran or Dundee or Hyderabad followed by sys admin experience at the Al Eadffg Coat Factory - it simply means nothing to you.
So, you need some kind of benchmark from which to work up from.
If you need an MS admin, you start off with people who are MS certified. If you need Sun skills you interview only those with Sun certification. Clearly a good manager will try and delve a little more deeply into what each individual can actually do and make a decision based on the results.
As an Englishman, I'd be happy accepting a CV from someone from the UK without certification as I could look at a CV and make some judgements based on my own experience as to whether they are worth interviewing. We'd speak the same language and would have had similar experiences which would let me make that judgement. For other nationalities, I'd expect to start with at least some kind of 'official' level of skill and take things from there.
Equally, an Indian manager wouldn't trust a UK CV - and quite rightly so - as they don't have the experience to judge what a UK guy's CV *really* means.
So, even though the certification doesn't guarantee that the guy can do as much as someone without a certification, it gives one a good basis on which to work from.
Linux is taking off here as everyone is obsessed with price and since Linux is 'free' it must be a good thing to use.
So Red Hat are right on the nail in producing a 'benchmark' which the guys with the budgets and the influence can use when looking for potential employees.
I think you're missing the point rather. I'm clearly not suggesting that ID cards replace normal policing methods.
In the UK, you're not required to carry ID, there is no 'one card' and not everyone has a driving licence. Even if you do, you aren't forced to have it when driving a car, for example.
With a proper ID card, through which one can access unified sources of information, all that info you mention about outstanding warrants can quickly be accessed by the police. This can help clear you quickly, or help the police arrest the right person more quickly. At the moment in the UK, if the police stop anyone they feel suspicious about, there is no official way of identifying that person quickly. This is silly and this is what needs to be changed.
ID cards would clearly be irrelevant here, as we both value people by the content of their character and don't need any of the bother of those awful formal means of proving who we are - as the poster said, he knows who he is, so that's clearly enough for anyone else.
Fair point. Can this be the first example of someone actually reacting to a posting on Slashdot, asking us all to 'imagine a Beowulf cluster' of something? Did Larry read it and tell his engineers to make it happen?
>I value people by the content of their character, >not by government issued papers.
So do I. You're clearly a nice guy. In fact, if you give me your bank account details, can I ask you to help transfer some money out of an account in Nigeria, to help some other nice people who've had a misunderstanding with the international finance community? They know who they are too.
Slashdot Mirror
A Beowulf cluster can't do everything - that's why we have SMP machines. If you want a Beowulf cluster of Suns, you can have that too. Sun produce a wide range of kit, not just big SMP machines.
Got any real reference sites of those Dells running Oracle RAC, or are you just reading the marketing brochures?
What exactly do you think OpenOffice is?
On the Linux front, I find the comments about Sun and Linux on Slashdot totally bizarre - one moment Sun are informed by Slashdot readers that Linux and Intel are the future, the next moment Sun are criticised for selling a server that addresses that part of the market.
Scaling well on multi-cpu systems? Excellent clustering?
What exactly is wrong with adopting Gnome?
There's a lawsuit going on with potentially large implications for Linux, but it's not clear at this stage - Sun say they're looking at the implications. How exactly does this make Sun 'fair weather friends'.
Have they dropped their Linux strategy, Linux blades, stoppped supporting the various Open Source projects, dropped their 100% Unix background and started selling NT boxes like Unix' other 'fair weather friends'? Thought not...
What do you mean by long term? There's a ten year road map for Solaris, there are hundreds of thousands of customers and Solaris excels over Linux in various areas.
The two OSes can sit together well in Sun's strategy.
I appreciate your point about Veritas Volume Manager, but I think the point of the post is that this functionality should be available, free, within Linux. Solaris has DiskSuite (or Solaris Volume Manager as it's been renamed in Solaris 9) for example. Considering everyone's whingeing about the 'cost' of Unix and advocating Linux as it's 'cheap' the fact you have to buy an expensive piece of volume manager software doesn't fit into the general scheme of things.
I was under the impression that Linux didn't scale very well over 4 cpus? Certainly you don't get linear scalability out of an 8 way Intel box running Linux, based on what I've read, but if you can point me to some evidence suggesting otherwise, cool.
But surely you should have heeded what people on Slashdot tell us is the thing to do - namely replace that 'expensive' Sun box with lots of 'cheap' Intel boxes in a cluster?
UIII is not 'slow', that's a childishly simplistic approach.
What's your application? On the 480 the internal bandwidth is 9.6Gbs, Level 2 cache on the CPUs is 8Mb, the cpus scale linearly. You have two on board copper Gb ethernet ports. You get Solaris for free, warranty is three years (probably the same as the Dell).
Also, you're not just buying a box, you're buying a box that's going to be part of the rest of your strategy - how well does it connect/support with other systems, can the vendor sell decent storage solutions around it, etc, etc.
Also, the V480 price is a list price.
I'd disagree that the low end Dell is going to be 'faster' than the low end Sun. That depends on the apps. Also, the 480 is priced similarly to the equivalent 4 cpu Dell, except the cpus scale linearly so I actually get to use 4 cpus for the money I've spent, the internal bandwidth is higher, build is higher quality, support is integrated for hardware and OS, I can run the same proven OS on that small box up to the F15k, I can get a large number of supported apps for Solaris, storage solutions are better, an excellent Professional Services organisation that can help me implement stuff and I have a partnership with a company that has vision and should be a decent partner with me as IT Manager as I roll stuff out throughout the business. If I really require Linux I can get LX50s and get my support from the same vendor as I do with the Sparc/Solaris boxes.
There's more to Sun vs Dell than 'my Dell is cheaper so Sun is dead'.
Shame Sun Rays aren't mentioned this time around. They'd still have a big place to play in the scenario the author's talking about - ability to hot desk around campus, zero maintenance, etc, etc.
Except, er, Sun isn't 'starting to extinguish'.
What companies do you know that run TPC-C benhcmarks as their business application? Why should we care how well IBM optimise their compilers to do well in benchmarks that don't tell us that the app we want to run will run fast?
Why would you choose such a large server based on a TPC-C benchmark? What about the rest of the server's components, storage, services, etc, etc?
Why don't you find out who your nearest Sun reseller is and call them?
The gui is the last thing on the list when buying a compute cluster! And besides, Gnome (and KDE and others) work fine on Solaris.
But I read on Slashdot that you can do anything on clusters of small Intel machines and Linux!
Pardon? Sun spend the equivalent of about 15% of revenues on R and D (compared to Dell's 1.5%).
Sun Rays are not a 1000 squids! More like 400 or so.
Maybe you need to spend $20 on a dictionary.
Even as a poor student, $20 was always available for cools stuff!
Buy some new Sparc machines then!!!
...have differing levels of importance.
I recently moved from the UK to the Middle East. The situation in the Gulf States such as Qatar, Oman, the UAE, etc, is that the local citizens of each state make up, on average, about 20% of the population of the country. 'Locals', meaning the indigenous Arab population, generally work for the government in managerial positions and the majority of hands on IT work is performed by people from every country in the world, from Kyrgystan to South Africa, to Great Britain, to Somalia, to India and everywhere in between.
Certifications in this region are extremely important. As a manager looking to employ someone, whether in a small or large company, you have to have some benchmark to use before you even interivew someone for a hands on IT role as you have no way of knowing what the general standard is in Iran, or Somalia or England, so there is no point interviewing everyone who claims they have a CS degree from the University of Tehran or Dundee or Hyderabad followed by sys admin experience at the Al Eadffg Coat Factory - it simply means nothing to you.
So, you need some kind of benchmark from which to work up from.
If you need an MS admin, you start off with people who are MS certified. If you need Sun skills you interview only those with Sun certification. Clearly a good manager will try and delve a little more deeply into what each individual can actually do and make a decision based on the results.
As an Englishman, I'd be happy accepting a CV from someone from the UK without certification as I could look at a CV and make some judgements based on my own experience as to whether they are worth interviewing. We'd speak the same language and would have had similar experiences which would let me make that judgement. For other nationalities, I'd expect to start with at least some kind of 'official' level of skill and take things from there.
Equally, an Indian manager wouldn't trust a UK CV - and quite rightly so - as they don't have the experience to judge what a UK guy's CV *really* means.
So, even though the certification doesn't guarantee that the guy can do as much as someone without a certification, it gives one a good basis on which to work from.
Linux is taking off here as everyone is obsessed with price and since Linux is 'free' it must be a good thing to use.
So Red Hat are right on the nail in producing a 'benchmark' which the guys with the budgets and the influence can use when looking for potential employees.
I think you're missing the point rather. I'm clearly not suggesting that ID cards replace normal policing methods.
In the UK, you're not required to carry ID, there is no 'one card' and not everyone has a driving licence. Even if you do, you aren't forced to have it when driving a car, for example.
With a proper ID card, through which one can access unified sources of information, all that info you mention about outstanding warrants can quickly be accessed by the police. This can help clear you quickly, or help the police arrest the right person more quickly. At the moment in the UK, if the police stop anyone they feel suspicious about, there is no official way of identifying that person quickly. This is silly and this is what needs to be changed.
ID cards would clearly be irrelevant here, as we both value people by the content of their character and don't need any of the bother of those awful formal means of proving who we are - as the poster said, he knows who he is, so that's clearly enough for anyone else.
Fair point. Can this be the first example of someone actually reacting to a posting on Slashdot, asking us all to 'imagine a Beowulf cluster' of something? Did Larry read it and tell his engineers to make it happen?
>I value people by the content of their character, >not by government issued papers.
So do I. You're clearly a nice guy. In fact, if you give me your bank account details, can I ask you to help transfer some money out of an account in Nigeria, to help some other nice people who've had a misunderstanding with the international finance community? They know who they are too.