Thank you, Dr.Allen for the explanation. But it makes one new question:
if the postprocessing is so huge, why not include it in distributed computing?
I cannot understand why this work cannot be split to be analyzed by our computers. May be it can be done the additional run for e.g. S5R2_pp etc.
Thank you, Dr.Allen for the explanation. But it makes one new question:
if the postprocessing is so huge, why not include it in distributed computing?
I cannot understand why this work cannot be split to be analyzed by our computers. May be it can be done the additional run for e.g. S5R2_pp etc.
Though the question is meant for Bruce, I hope you don't mind that I'll try to answer it:
1. The post-processing is basically combining the information in the results that we got back from an Einstein@home analysis run, so it's not easily possible to split up this task - what we were able to split up already has been done in setting up the run.
2. The computational power might be large enough on Einstein@home, but the data transfer bandwidth is not. You'd spend more time down- and uploading data than for the computation. This is unacceptable both for you and for our servers.
3. The software and parameters we use in for post-processing is under continuous refinement based on previous results. Setting up a new Einstein@home analysis run typically takes 4 weeks minimum, getting a new application ready for BOINC takes months.
4. For comparison: Holger runs a complete S5R1 post-processing "run" (given a set of filter parameters) in a bit more than a day (on ATLAS with a data distribution scheme specifically developed for his pipeline. This mean that the longest job runs that long, most jobs finish way earlier). With the new "Hierarchical Search" Application we are using since S5R2 a complete post-processing "run" (with a set of filter parameters) takes about two hours on ATLAS.
Bottom line: it's much faster to do this on our own clusters than it would take to just set it up as a "run" on Einstein@home.
I notice that the Atlas processors show a fairly high amount of cache memory. 4000+ kb. I hope this is not too out of place, but I wonder are the Atlas processors an off the shelf type or a specially designed processor. I'm not up to date with all the processors on the market these days, there seems to be too many for me to keep up with. But 4000 kb of cach seems like one of the higher cache amounts I've seen.
I notice that the Atlas processors show a fairly high amount of cache memory. 4000+ kb. I hope this is not too out of place, but I wonder are the Atlas processors an off the shelf type or a specially designed processor. I'm not up to date with all the processors on the market these days, there seems to be too many for me to keep up with. But 4000 kb of cach seems like one of the higher cache amounts I've seen.
If the CPU info is correct, they are off the shelf
I notice that the Atlas processors show a fairly high amount of cache memory. 4000+ kb. I hope this is not too out of place, but I wonder are the Atlas processors an off the shelf type or a specially designed processor. I'm not up to date with all the processors on the market these days, there seems to be too many for me to keep up with. But 4000 kb of cach seems like one of the higher cache amounts I've seen.
If the CPU info is correct, they are off the shelf
I notice that the Atlas processors show a fairly high amount of cache memory. 4000+ kb.
If the CPU info is correct, they are off the shelf Intel Xeon X3220
processors. One for approx. $200 (times 1342....)
One might add that 2.4 GHz Kentsfield with 4 Mbyte L3 cache also describes the Q6600 popular with many Einstein and SETI contributors, especially since they hit a popular price point about July 2007. Each a pair of Conroe-class chips in a single package, running the same stock clock rate. The base chip is, I believe, identical, with a few blown fuses making the personality change to Xenon.
The Xenon-ness can mean more support for a single motherboard bearing more than one socket, but so far as raw compute is concerned, these start equal to a Q6600, and often lag in Einstein service, as many Q6600 owners see fit to overclock--I suspect that is rarer in Xenon-drivers.
Confession: I'm not expert on Xenon-ness, would welcome additions or corrections.
Newegg's prices often track the 1000-piece list price from Intel pretty closely. If you want much discount, you may have to place a much bigger order than required to populate ATLAS.
I never understood the difference between a Pentium and a Xeon. Is it simply a marketing stunt or is there a real difference? Same question for the Opteron/Athlon couple. I have bought a SUN WS with an AMD Opteron 1210 while most of my AMD powered wingmen have Athlons.
Tullio
I never understood the difference between a Pentium and a Xeon. Is it simply a marketing stunt or is there a real difference? Same question for the Opteron/Athlon couple. I have bought a SUN WS with an AMD Opteron 1210 while most of my AMD powered wingmen have Athlons.
Tullio
The difference is in the core. Xeon cores designed escpecially for server side functions. Thus it has a different socket and bigger L2 cache than usual Pentium. Moreover usually only Xeon cores has a possibility to work in multiprocessor mobos (I mean more than 2 cores per motherboard, not per processor). It is usually does not have enough power for overclocking, but usually more stable than Pentiums and mobos for it, escpecially for the ECC checking, memory with buffers and so on. And that is why Pentiums are significantly cheeper than Xeons.
The main secret of this is that in usual realtime business processes it doesn't really needed at all. But nobody talks about it. For e.g., most of my servers, working with trade databases for supermarkets and other trade companies (that I support this time), are usual Pentiums and rarely Core2. Moreover, most of them working hard primarily on E@H, because SQL and fileserver functions doesn't really use even 1% of processor power and this is a big area for E@H.
So, for home or office use - don't even mind to buy a Xeon (only on Ebay like Bikeman does), but for the realtime applications that needs a processing power with stability of calculations (for e.g. - clusters like ATLAS) - Xeon is a solution.
Thank you Stranger7777 for answering. But I am satisfied of my Opteron at 1.8 GHz and, when the warranty period of one year for my WS has passed, I am thinking of substituting it with an Opteron of higher clock frequency on the same AM2 socket, maybe with a more powerful fan. But I am not an overclocker. My WS looks as a kind of car radiator in the front panel, I mean it is open, and the airflow is good, with a fan on the PSU and a bigger one on the back panel. Thanks.
Tullio
Thank you, Dr.Allen for the
)
Thank you, Dr.Allen for the explanation. But it makes one new question:
if the postprocessing is so huge, why not include it in distributed computing?
I cannot understand why this work cannot be split to be analyzed by our computers. May be it can be done the additional run for e.g. S5R2_pp etc.
RE: Thank you, Dr.Allen for
)
Though the question is meant for Bruce, I hope you don't mind that I'll try to answer it:
1. The post-processing is basically combining the information in the results that we got back from an Einstein@home analysis run, so it's not easily possible to split up this task - what we were able to split up already has been done in setting up the run.
2. The computational power might be large enough on Einstein@home, but the data transfer bandwidth is not. You'd spend more time down- and uploading data than for the computation. This is unacceptable both for you and for our servers.
3. The software and parameters we use in for post-processing is under continuous refinement based on previous results. Setting up a new Einstein@home analysis run typically takes 4 weeks minimum, getting a new application ready for BOINC takes months.
4. For comparison: Holger runs a complete S5R1 post-processing "run" (given a set of filter parameters) in a bit more than a day (on ATLAS with a data distribution scheme specifically developed for his pipeline. This mean that the longest job runs that long, most jobs finish way earlier). With the new "Hierarchical Search" Application we are using since S5R2 a complete post-processing "run" (with a set of filter parameters) takes about two hours on ATLAS.
Bottom line: it's much faster to do this on our own clusters than it would take to just set it up as a "run" on Einstein@home.
BM
BM
I notice that the Atlas
)
I notice that the Atlas processors show a fairly high amount of cache memory. 4000+ kb. I hope this is not too out of place, but I wonder are the Atlas processors an off the shelf type or a specially designed processor. I'm not up to date with all the processors on the market these days, there seems to be too many for me to keep up with. But 4000 kb of cach seems like one of the higher cache amounts I've seen.
RE: I notice that the Atlas
)
If the CPU info is correct, they are off the shelf
Intel Xeon X3220
processors. One for approx. $200 (times 1342....)
Michael
Team Linux Users Everywhere
RE: RE: I notice that the
)
Would one be able to purchase large quantities at a bulk rate?
RE: RE: I notice that the
)
One might add that 2.4 GHz Kentsfield with 4 Mbyte L3 cache also describes the Q6600 popular with many Einstein and SETI contributors, especially since they hit a popular price point about July 2007. Each a pair of Conroe-class chips in a single package, running the same stock clock rate. The base chip is, I believe, identical, with a few blown fuses making the personality change to Xenon.
The Xenon-ness can mean more support for a single motherboard bearing more than one socket, but so far as raw compute is concerned, these start equal to a Q6600, and often lag in Einstein service, as many Q6600 owners see fit to overclock--I suspect that is rarer in Xenon-drivers.
Confession: I'm not expert on Xenon-ness, would welcome additions or corrections.
Newegg's prices often track the 1000-piece list price from Intel pretty closely. If you want much discount, you may have to place a much bigger order than required to populate ATLAS.
Thank you, Bernd. I didn't
)
Thank you, Bernd.
I didn't got in mind that postprocessing is concerned with huge amounts of data transfer. BTW, how much data is in S5R1-S5R4 results database?
I never understood the
)
I never understood the difference between a Pentium and a Xeon. Is it simply a marketing stunt or is there a real difference? Same question for the Opteron/Athlon couple. I have bought a SUN WS with an AMD Opteron 1210 while most of my AMD powered wingmen have Athlons.
Tullio
RE: I never understood the
)
The difference is in the core. Xeon cores designed escpecially for server side functions. Thus it has a different socket and bigger L2 cache than usual Pentium. Moreover usually only Xeon cores has a possibility to work in multiprocessor mobos (I mean more than 2 cores per motherboard, not per processor). It is usually does not have enough power for overclocking, but usually more stable than Pentiums and mobos for it, escpecially for the ECC checking, memory with buffers and so on. And that is why Pentiums are significantly cheeper than Xeons.
The main secret of this is that in usual realtime business processes it doesn't really needed at all. But nobody talks about it. For e.g., most of my servers, working with trade databases for supermarkets and other trade companies (that I support this time), are usual Pentiums and rarely Core2. Moreover, most of them working hard primarily on E@H, because SQL and fileserver functions doesn't really use even 1% of processor power and this is a big area for E@H.
So, for home or office use - don't even mind to buy a Xeon (only on Ebay like Bikeman does), but for the realtime applications that needs a processing power with stability of calculations (for e.g. - clusters like ATLAS) - Xeon is a solution.
Thank you Stranger7777 for
)
Thank you Stranger7777 for answering. But I am satisfied of my Opteron at 1.8 GHz and, when the warranty period of one year for my WS has passed, I am thinking of substituting it with an Opteron of higher clock frequency on the same AM2 socket, maybe with a more powerful fan. But I am not an overclocker. My WS looks as a kind of car radiator in the front panel, I mean it is open, and the airflow is good, with a fan on the PSU and a bigger one on the back panel. Thanks.
Tullio