It is amazing how hard it is to find true benchmarks that show actual real-world performance. When we were looking for our new database server we knew that we couldn't trust any SPEC benchmarks or anything similar.
We scoured the net for information on database performance. Now, keep in mind we are running Linux with MySQL so all those Windows performance tests didn't mean anything to us...and they were abundant.
What we ended up with was we decided to go the route of Xeon processors. For us it was a simple cost vs performance comparison and the availability of parts. I hate to say it but we don't want to be stuck with one machine vendor (Dell) so if we have problems in the future we can easily switch. That's why Apple will never have our core business...they MAY get our web cluster business but I doubt it.
We found that our true bottle necks were the memory throughput of our database machine (133 FSB) and the lack of physical memory. Our old database server was a Dell 2450 Dual PIII 933 with 1G RAM running on an 18G RAID 1 setup running RH Linux 7.2. Also, each CPU would sit at around 10 - 50% useage constantly. We also new that many of our heavy queries would nail each CPU for a split second. The MySQL server also AVERAGED over 32 queries per second so it was fairly busy.
With this information we knew that we were going to need an 8 way Xeon machine in about 12 months. What we didn't know is that we are growing so quickly right now that we needed a machine NOW. After doing some qick investigation we realized that we *should* go the route of Intels new HyperThreading Xeon processors. We know from real world tests that MySQL and Linux would recognize each CPU as 2 CPU's.
We quickly zeored in on the Dell 2650 with Dual Xeon 2.4's, 3G RAM (400 FSB), ~100G RAID5 with hot spare, redundant everything and dual GB NICs. Now, here is the BIG suprize this machine in raw inserts is over 4x faster than our old machine. The new machine could do over 2000 insert queries/second. Also, Linux is taking full use of the HyperThreading...as I'm sure OSX would ;-)
Now, how does this compare to Apple's machines? Namely the xServe? Good question. We are going to order an xServe just to play with it but we don't have any time right now so we haven't. I would be interested if someone has an xServe setup with MySQL to see what kind of performance they are getting.
Apple does have a long way to go in their hardware but there is no doubt their OS is on the right track. One of my guys actually wants an OSX workstation now and, although he doesn't know it, he'll get it later this year.
Now, I wonder what Apple has coming in a couple weeks...
<strong>I hate to say it but we don't want to be stuck with one machine vendor (Dell) so if we have problems in the future we can easily switch. That's why Apple will never have our core business...they MAY get our web cluster business but I doubt it.</strong><hr></blockquote>
And this is a big reason why Apple needs IBM to be a licensed OSX vendor.
Maybe there are too many people in here who know a little too much about electronics. So why don't I clarify the concerns of customers based on the understanding of someone who knows nothing about electronics.
"How fast?" is for us (users) measured in time to completed task. Whether such times are a result of software efficiency or hardware power, we don't care, we don't see that, we see only the speed with which our commands are carried out by the system.
"How expensive is it?" Can I afford to buy it? Sure, it might be twice as fast, but if it costs 10X as much, I'll take 2 or 3 or 4 of the cheaper system and work out a distributed solution.
Put those two questions together and you have the primary metric by which all processor designs are judged. The ONLY exceptions would be for highly specialized "cost-is-no-object" applications. A consumer, a pro, or a very high-end pro is each forced to evaluate a system in that way.
The next set of questions surround efficiency (but they are really just subsets of cost)
How much energy does it use? How much space does it take up? If a processor isn't the fastest out there but it's cheap enough and cool enough that I can cram a lot of them into 1 box and a lot of boxes into one space (rack, closet, room) then it is a superior design for some applications and not for others.
It really comes down to how fast and how much? People who know and love design might prefer a certain design over another because they find it intellectually gratifying/stimulating to apply certain philosophies, but I venture that the consumer doesn't care.
All the other issues come down to those 2 questions: "How fast?" and, "How much?" Support costs, ease of use, stability, available applications, trained personelle -- It's a balance that each buyer weighs a little differently, but on the whole shows a very homogenous presentation.
Does either Itanium or Power matter to the mac user? Nope, we'll not see either in a mac for quite a while. But, if either (providing leading performance) could be had cheap, then either would be good.
There is no better design as far as we're concerned. Just faster and cheaper. Don't care how you do it, just do it.
Well, powerful as it would be for Apple to cede it's hi end server offerings to IBM; what price would Apple pay for effecitvely giving away control of this market?
Well, powerful as it would be for Apple to cede it's hi end server offerings to IBM; what price would Apple pay for effecitvely giving away control of this market?
Would it really be worth it?</strong><hr></blockquote>
That wasn't posted by me <img src="confused.gif" border="0">
<strong>Maybe there are too many people in here who know a little too much about electronics. So why don't I clarify the concerns of customers based on the understanding of someone who knows nothing about electronics.
"How fast?" is for us (users) measured in time to completed task. Whether such times are a result of software efficiency or hardware power, we don't care, we don't see that, we see only the speed with which our commands are carried out by the system.
"How expensive is it?" Can I afford to buy it? Sure, it might be twice as fast, but if it costs 10X as much, I'll take 2 or 3 or 4 of the cheaper system and work out a distributed solution.
Put those two questions together and you have the primary metric by which all processor designs are judged. The ONLY exceptions would be for highly specialized "cost-is-no-object" applications. A consumer, a pro, or a very high-end pro is each forced to evaluate a system in that way.
The next set of questions surround efficiency (but they are really just subsets of cost)
How much energy does it use? How much space does it take up? If a processor isn't the fastest out there but it's cheap enough and cool enough that I can cram a lot of them into 1 box and a lot of boxes into one space (rack, closet, room) then it is a superior design for some applications and not for others.
It really comes down to how fast and how much? People who know and love design might prefer a certain design over another because they find it intellectually gratifying/stimulating to apply certain philosophies, but I venture that the consumer doesn't care.
All the other issues come down to those 2 questions: "How fast?" and, "How much?" Support costs, ease of use, stability, available applications, trained personelle -- It's a balance that each buyer weighs a little differently, but on the whole shows a very homogenous presentation.
Does either Itanium or Power matter to the mac user? Nope, we'll not see either in a mac for quite a while. But, if either (providing leading performance) could be had cheap, then either would be good.
There is no better design as far as we're concerned. Just faster and cheaper. Don't care how you do it, just do it.</strong><hr></blockquote>
I **believe** that Apple, in the next year, will make a big change. I believe that Apple will make the iMac line their consumer line...even more so than it is now.
I also belive that Apple will make the PowerMac line more of a professional lineup. I don't care what people say on these boards, 99% of the people out there would be just fine with a new iMac.
Well, powerful as it would be for Apple to cede it's hi end server offerings to IBM; what price would Apple pay for effecitvely giving away control of this market?
Would it really be worth it?</strong><hr></blockquote>
What are you talking about? Apple is barely a player in the low-end server market, much less the high end sector. What control would they be giving away, exactly?
What are you talking about? Apple is barely a player in the low-end server market, much less the high end sector. What control would they be giving away, exactly?</strong><hr></blockquote>
... that's certainly one way to see it, 100% control of nothing is still nothing; but the problem with that argument is it assumes there's a potential middle ground ... as if Apple can keep part of any future server market to itself while pulling off some sort of win-win deal with a corporate behemoth like IBM in that realm.
Well, maybe there is, it's certainly worth a thought, but before Apple goes merily liscencing away with the best of future market share intentions, it'd do well to strongly consider what kind of game it's really getting into.
Fools rush in where wise men fear to tread; and Apple's already learned humility the hard way after one OS liscencing fiasco.
I'm not saying it's a bad idea, I'm just saying they don't dare go into it without really knowing what they're doing.
Comments
We scoured the net for information on database performance. Now, keep in mind we are running Linux with MySQL so all those Windows performance tests didn't mean anything to us...and they were abundant.
What we ended up with was we decided to go the route of Xeon processors. For us it was a simple cost vs performance comparison and the availability of parts. I hate to say it but we don't want to be stuck with one machine vendor (Dell) so if we have problems in the future we can easily switch. That's why Apple will never have our core business...they MAY get our web cluster business but I doubt it.
We found that our true bottle necks were the memory throughput of our database machine (133 FSB) and the lack of physical memory. Our old database server was a Dell 2450 Dual PIII 933 with 1G RAM running on an 18G RAID 1 setup running RH Linux 7.2. Also, each CPU would sit at around 10 - 50% useage constantly. We also new that many of our heavy queries would nail each CPU for a split second. The MySQL server also AVERAGED over 32 queries per second so it was fairly busy.
With this information we knew that we were going to need an 8 way Xeon machine in about 12 months. What we didn't know is that we are growing so quickly right now that we needed a machine NOW. After doing some qick investigation we realized that we *should* go the route of Intels new HyperThreading Xeon processors. We know from real world tests that MySQL and Linux would recognize each CPU as 2 CPU's.
We quickly zeored in on the Dell 2650 with Dual Xeon 2.4's, 3G RAM (400 FSB), ~100G RAID5 with hot spare, redundant everything and dual GB NICs. Now, here is the BIG suprize this machine in raw inserts is over 4x faster than our old machine. The new machine could do over 2000 insert queries/second. Also, Linux is taking full use of the HyperThreading...as I'm sure OSX would ;-)
================================================== =======================
8:51am up 3 days, 23:54, 1 user, load average: 0.04, 0.07, 0.08
57 processes: 56 sleeping, 1 running, 0 zombie, 0 stopped
CPU0 states: 4.0% user, 7.0% system, 0.0% nice, 88.0% idle
CPU1 states: 0.0% user, 1.0% system, 0.0% nice, 98.0% idle
CPU2 states: 40.0% user, 11.0% system, 0.0% nice, 48.0% idle
CPU3 states: 0.0% user, 0.0% system, 0.0% nice, 100.0% idle
Mem: 3099304K av, 3082144K used, 17160K free, 0K shrd, 140196K buff
Swap: 3156648K av, 0K used, 3156648K free 2619892K cached
================================================== =======================
Now, how does this compare to Apple's machines? Namely the xServe? Good question. We are going to order an xServe just to play with it but we don't have any time right now so we haven't. I would be interested if someone has an xServe setup with MySQL to see what kind of performance they are getting.
Apple does have a long way to go in their hardware but there is no doubt their OS is on the right track. One of my guys actually wants an OSX workstation now and, although he doesn't know it, he'll get it later this year.
Now, I wonder what Apple has coming in a couple weeks...
<strong>I hate to say it but we don't want to be stuck with one machine vendor (Dell) so if we have problems in the future we can easily switch. That's why Apple will never have our core business...they MAY get our web cluster business but I doubt it.</strong><hr></blockquote>
And this is a big reason why Apple needs IBM to be a licensed OSX vendor.
<strong>
And this is a big reason why Apple needs IBM to be a licensed OSX vendor.</strong><hr></blockquote>
I couldn't agree more! No company in their right mind today will lock themselves into one vendor regardless of how good their books look on paper.
"How fast?" is for us (users) measured in time to completed task. Whether such times are a result of software efficiency or hardware power, we don't care, we don't see that, we see only the speed with which our commands are carried out by the system.
"How expensive is it?" Can I afford to buy it? Sure, it might be twice as fast, but if it costs 10X as much, I'll take 2 or 3 or 4 of the cheaper system and work out a distributed solution.
Put those two questions together and you have the primary metric by which all processor designs are judged. The ONLY exceptions would be for highly specialized "cost-is-no-object" applications. A consumer, a pro, or a very high-end pro is each forced to evaluate a system in that way.
The next set of questions surround efficiency (but they are really just subsets of cost)
How much energy does it use? How much space does it take up? If a processor isn't the fastest out there but it's cheap enough and cool enough that I can cram a lot of them into 1 box and a lot of boxes into one space (rack, closet, room) then it is a superior design for some applications and not for others.
It really comes down to how fast and how much? People who know and love design might prefer a certain design over another because they find it intellectually gratifying/stimulating to apply certain philosophies, but I venture that the consumer doesn't care.
All the other issues come down to those 2 questions: "How fast?" and, "How much?" Support costs, ease of use, stability, available applications, trained personelle -- It's a balance that each buyer weighs a little differently, but on the whole shows a very homogenous presentation.
Does either Itanium or Power matter to the mac user? Nope, we'll not see either in a mac for quite a while. But, if either (providing leading performance) could be had cheap, then either would be good.
There is no better design as far as we're concerned. Just faster and cheaper. Don't care how you do it, just do it.
[QB]
Well, powerful as it would be for Apple to cede it's hi end server offerings to IBM; what price would Apple pay for effecitvely giving away control of this market?
Would it really be worth it?
<strong>[QUOTE]Originally posted by Mike:
[QB]
Well, powerful as it would be for Apple to cede it's hi end server offerings to IBM; what price would Apple pay for effecitvely giving away control of this market?
Would it really be worth it?</strong><hr></blockquote>
That wasn't posted by me <img src="confused.gif" border="0">
<strong>Maybe there are too many people in here who know a little too much about electronics. So why don't I clarify the concerns of customers based on the understanding of someone who knows nothing about electronics.
"How fast?" is for us (users) measured in time to completed task. Whether such times are a result of software efficiency or hardware power, we don't care, we don't see that, we see only the speed with which our commands are carried out by the system.
"How expensive is it?" Can I afford to buy it? Sure, it might be twice as fast, but if it costs 10X as much, I'll take 2 or 3 or 4 of the cheaper system and work out a distributed solution.
Put those two questions together and you have the primary metric by which all processor designs are judged. The ONLY exceptions would be for highly specialized "cost-is-no-object" applications. A consumer, a pro, or a very high-end pro is each forced to evaluate a system in that way.
The next set of questions surround efficiency (but they are really just subsets of cost)
How much energy does it use? How much space does it take up? If a processor isn't the fastest out there but it's cheap enough and cool enough that I can cram a lot of them into 1 box and a lot of boxes into one space (rack, closet, room) then it is a superior design for some applications and not for others.
It really comes down to how fast and how much? People who know and love design might prefer a certain design over another because they find it intellectually gratifying/stimulating to apply certain philosophies, but I venture that the consumer doesn't care.
All the other issues come down to those 2 questions: "How fast?" and, "How much?" Support costs, ease of use, stability, available applications, trained personelle -- It's a balance that each buyer weighs a little differently, but on the whole shows a very homogenous presentation.
Does either Itanium or Power matter to the mac user? Nope, we'll not see either in a mac for quite a while. But, if either (providing leading performance) could be had cheap, then either would be good.
There is no better design as far as we're concerned. Just faster and cheaper. Don't care how you do it, just do it.</strong><hr></blockquote>
I **believe** that Apple, in the next year, will make a big change. I believe that Apple will make the iMac line their consumer line...even more so than it is now.
I also belive that Apple will make the PowerMac line more of a professional lineup. I don't care what people say on these boards, 99% of the people out there would be just fine with a new iMac.
<strong>There is no better design as far as we're concerned. Just faster and cheaper. Don't care how you do it, just do it.</strong><hr></blockquote>
Two more:
Useability
Reliability
I suppose they can be considered variations of cost and speed, but really they are worth considering on their own.
And then there is the nasty issue of style and appearance.
You're right that most people really only measure the product based on these criteria and not on its internal design... but damn I hate that!!
<strong>[QUOTE]Originally posted by Mike:
[QB]
Well, powerful as it would be for Apple to cede it's hi end server offerings to IBM; what price would Apple pay for effecitvely giving away control of this market?
Would it really be worth it?</strong><hr></blockquote>
What are you talking about? Apple is barely a player in the low-end server market, much less the high end sector. What control would they be giving away, exactly?
<strong>
What are you talking about? Apple is barely a player in the low-end server market, much less the high end sector. What control would they be giving away, exactly?</strong><hr></blockquote>
... that's certainly one way to see it, 100% control of nothing is still nothing; but the problem with that argument is it assumes there's a potential middle ground ... as if Apple can keep part of any future server market to itself while pulling off some sort of win-win deal with a corporate behemoth like IBM in that realm.
Well, maybe there is, it's certainly worth a thought, but before Apple goes merily liscencing away with the best of future market share intentions, it'd do well to strongly consider what kind of game it's really getting into.
Fools rush in where wise men fear to tread; and Apple's already learned humility the hard way after one OS liscencing fiasco.
I'm not saying it's a bad idea, I'm just saying they don't dare go into it without really knowing what they're doing.
And right now, I don't think they do.
<strong>
That wasn't posted by me
... you're right, sorry mike, quotes got screwed.
<strong>
... you're right, sorry mike, quotes got screwed.</strong><hr></blockquote>
No problem