AMD_Tracy | Hi, everyone. |
mike | hello AMD_Tracy |
mike | I am Mike Lydon with TopCoder |
AMD_Tracy | Nice to meet you. |
mike | likewise |
mpaulse | hey AMD_Tracy :) |
mike | I believe the topic today is "business intelligence" |
joelimlimit | Hello |
joelimlimit | Is any rep here? |
joelimlimit | Hello |
AMD_Sharon | Hello |
AMD_Tracy | We're here. |
joelimlimit | what does AMD do? |
joelimlimit | I mean, why do you need programmers for exactly? |
AMD_Sharon | AMD is a CPU manufacturer |
joelimlimit | yea that's about the only thing I know at this point |
AMD_Sharon | There are several ways to optimize your software code to take advantage of CPUs |
butler | Aren't you AMD folks based in Sunnyvale? |
AMD_Sharon | Today we will talk about business intelligence |
joelimlimit | which is? |
AMD_Tracy | butl |
nimisis | a contradiction in terms |
joelimlimit | butler: hey, you are here for the SRM too? |
mpaulse | lol |
AMD_Tracy | butler: amd is based in sunnyvale |
butler | joelimlimit: yup |
joelimlimit | butler: g'luck :) |
joelimlimit | So, do you hire any interns too? |
Pasto | What's exactly 'business intelligenc' ? |
Pasto | *intelligence |
shash | Pasto: common sense? ;-) |
joelimlimit | I was going to ask the same Q with Pasto |
vinay.emani | AMD_Sharon:hi sharon,does AMD have offices in India? |
Pasto | shash: hope not :P |
joelimlimit | haha becuz I sometime need to log in to two computers |
AMD_Sharon | vinay.emani: yes we have offices in india |
joelimlimit | at the same time |
dhavalgiani | AMD_Sharon: do ur india offices offer internships? |
vinay.emani | AMD_Sharon: thanks,are you hiring here(india)? |
joelimlimit | AMD_Sharon: Any Intern offer? |
AMD_Tracy | butler: Definition of business intelligence can mean different things to different people. Loosely, it is a collection of business processes for collecting and analyzing business information. |
mpaulse | So it refers to knowledge/expert systems? |
AMD_Sharon | dhavalgiani: Check our website on www.amd.com for the internships and careers available world wide |
butler | Or data mining? |
AMD_Tracy | butler: For some, business intelligence is synonymous with data warehousing. For others, business intelligence really means the analysis of data in the warehouse. |
dhavalgiani | AMD_Sharon: thanks |
atul.kulkarni | what is new with this Bussiness Intelligence> |
AMD_Tracy | mpaulse: I would say not, that's usually associated with artificial intelligence. |
rajkumar.iyer | AMD_Tracy: How does AMD fit in here ... for business intelligence ?? |
atul.kulkarni | rajkumar.iyer> AMD_Tracy: that was next question |
AMD_Sharon | Business intelligence is just one area that AMD is paying attention too |
AMD_Sharon | We decided to talk about that today in order to give you an idea that there really are many possibilities for software developers |
AMD_Tracy | rajkumar.iyer: I was mentioning how data warehousing is important. In that case, you are talking a great deal of I/O, memory consumption, and n-tier systems. AMD has great server systems that provide great value here. |
atul.kulkarni | AMD_Sharon: is AMD planing to write software for this? |
mpaulse | AMD_Tracy: That's interesting, coz one only thinks of AMD as just a CPU manufacturer and not a (should I say) complete server vendor (like IBM for example) |
AMD_Sharon | atul.kulkarni: AMD partners with other software and hardware companies to make sure that they take advantage of everything that our CPU architecture has to offer |
AMD_Tracy | rajkumar.iyer: Typically with data warehousing, there are multiple data sources from multiple transactional based systems. A data warehouse is an infrastructure to get data from those transactional systems. Then, you need to be able to analyze and |
AMD_Tracy | rajkumar.iyer: figure out what to do with the information that is buried in the data warehouse. |
atul.kulkarni | AMD_Sharon: oh ok! so what kind of profiles does this venture of your open for devs? |
AMD_Tracy | rajkumar.iyer: AMD is working with major OEMS such as Sun, IBM, and HP to provide world class scalable server systems. |
AMD_Sharon | atul.kulkarni: We have teams of software developers dedicated to performance optimization who work directly with our partners |
atul.kulkarni | AMD_Sharon: oh ok! |
shash | AMD_Sharon: what kind of optimizations are you talking about for the warehousing servers? |
rajkumar.iyer | AMD_Tracy: Are we here looking at something like dedicated system .. which are used specially to do to the intelligent data analysis.. |
nimisis | AMD_Sharon:When I hear "business intelligence" I think of scheduling problems that are often NP hard. So, if I was to come to AMD, might I be working on heuristics to solve such probs? |
nimisis | ...for example |
AMD_Tracy | rajkumar.iyer: It could be but not necessarily; for example you could have a partitioned n-way system with multiple tiers |
butler | Optimization would include writing code in low level languages to take advantage of MIMD instructions? |
AMD_Sharon | Tracy is the expert in Business intelligence, so I'll let him answer your questions |
AMD_Tracy | rajkumar.iyer: and maybe one partition on that system is dedicated to analysis phase and other partitions are used for the data store. |
rajkumar.iyer | AMD_Tracy: And all of them somewhere at the processor level and each processor interacting more intelligently then what they ar doing now is it ?? |
rajkumar.iyer | AMD_Tracy: *than |
vinay.emani | AMD_Sharon: is amd planning to make any GPUs? |
AMD_Tracy | rajkumar.iyer: Well, I was actually putting forth a virtualization example which is becoming more and more common in IT. |
AMD_Sharon | vinay.emani: No :) We are sticking to CPUs |
rajkumar.iyer | AMD_Tracy: I was just calling partition a processor... |
AMD_Tracy | rajkumar.iyer: sure, okay. So for example a case of more intelligent interactions is the Direct Connect Architecture. |
modenl | AMD_Tracy: Do you coporate with sql server 2005? |
modenl | AMD_Tracy: They have cool data minning tools. |
CoreDumped | AMD_Tracy: hi. what does amd think of Azul's 48 core processor? |
AMD_Tracy | rajkumar.iyer: The HyperTransport links and integrated memory controllers within the DCA allow for point-to-point connections between the various components on the system such as your i/o devices, CPU and memory. |
shash | AMD_Tracy: what you're describing is something like a more generalized DMA? |
AMD_Tracy | modenl: We have a very strong partnership with Microsoft and we work with them on everything from SQL Server, to Visual Studio, to XP and Server OSs. |
butler | AMD_Tracy: Is HyperTransport scalable for large clusters? Or is it only for 1-8 way CPU systems? |
AMD_Tracy | modenl: The neat thing about their data mining is the relative seamlessness of it; nice little extension to the sql language in a fairly easy to understand way. |
AMD_Tracy | shash: DMA? |
butler | AMD_Tracy: direct memory access |
shash | AMD_Tracy: direct memory access |
AMD_Sharon | butler: Hypertransport is a chip to chip interconnect, so it will only be used in each system |
AMD_Tracy | shash: yes, the amd CPUs now have the memory controller directly integrated. This means there isn't a single memory controller hub bottlenecking the system in a traditional front side bus design. |
shash | AMD_Sharon: chip to chip as in, not on a bus? |
AMD_Tracy | butler: also keep in mind that HyperTransport is now in consortium status. |
modenl | AMD_Tracy: Could you give us a solid example what you've done for Speed up data minning? |
AMD_Sharon | shash: It's an internal bus, so as long as the chips on down on the motherboard, HyperTransport is an excellent solution |
butler | AMD_Tracy: consortium status? |
AMD_Tracy | modenl: AMD processors have great floating point performance and that's what you are going to see in data mining. One aspect of this is that AMD processors are good at x87 sequenes and also support SSE2 instructions. |
AMD_Sharon | shash: when you talk about clusters, they would be connected via GigEthernet or something like that, I believe |
shash | AMD_Sharon: i guess... |
shash | AMD_Sharon: so what does that mean for the processor? |
AMD_Tracy | modenl: In addition, AMD has the AMD Core Math Library (ACML) which contains optimised BLAS routines that would be used in data mining algorithms. |
AMD_Sharon | shash: The main advantage of hypertransport is that it allows us to connect memory to each processor in a system, throught our integrated memory controller |
AMD_Sharon | shash: As well as connecting the I/O to CPUs |
modenl | AMD_Tracy: Any comparision for ACML with similar libs? |
shash | AMD_Sharon: so, each processor has a memory controller, which all work together? |
AMD_Sharon | shash: You got it! |
i_am_here | what is the topic of disscussion today |
butler | i_am_here: business intelligence |
rajkumar.iyer | AMD_Tracy: How are u planning to have this intergerated with the existing Business Intelligence Software which are already in market ?? |
rajkumar.iyer | AMD_Tracy: take for example MS-SQl data Warehousing utility itself |
butler | Wouldn't AMD's target be the larger scale systems? A completely different product? |
AMD_Tracy | rajkumar.iyer: I cannot get into too many specifics because this is a new area for AMD but we are learning all we can. |
AMD_Tracy | rajkumar.iyer: So, is the ms sql data warehousing utility something you are using? |
AMD_Tracy | butler: Can you clarify your question? |
shash | what does all this mean for small-scale businesses? |
rajkumar.iyer | AMD_Tracy: No I work on Database Engine (OLTP) |
butler | AMD_Tracy: I was trying to understand what you were saying. Is AMD targeting the large scale systems offered by IBM, Sun, etc, rather than the MS-SQL data warehousing utility offered currently? |
rajkumar.iyer | AMD_Tracy: I was just trying to make our whar are we looking here from AMD ...and can that be of any use in traditional database engines.. |
rajkumar.iyer | AMD_Tracy: *make out |
iapx86 | I optimize MMX/SSE by generating opcodes at run-time with my own special classes and methods. Any advice on AMD documents that help more than the obfuscatory Intel ones? |
AMD_Tracy | rajkumar.iyer: Sure, in traditional database engines you have an I/O component as well as a CPU component, and a memory component. These can all be bottlenecks. AMD pays attention to everyone. |
AMD_Sharon | iapx86: You should definitely check out our set of Architecture Programmer's Manuals |
AMD_Sharon | iapx86: We also offer an optimization guide |
AMD_Sharon | iapx86: you can find them on http://developer.amd.com |
AMD_Tracy | iapx86: the optimization guide is one of my favourite bits of reading material. |
iapx86 | Thanks. I just registered there. Will look after SRM296 is over. |
AMD_Tracy | butler: I can say that the work with HP / IBM / Sun is a natural thing for us to do and to understand. The MS-SQL thing is more of a challenge to us but we are keenly interested in any pain points you have |
AMD_Tracy | iapx86: there's good stuff in the opt guide on latencies; our strategy is to document the latencies as of the latest production round of chips. I know the guy who does the guide and he does a terrific job. |
rajkumar.iyer | AMD_Tracy: Thanks for the info Tracy :) |
Parchandri | Is anything under discussion yet? |
dexy | ikabiljo: a sada? |
AMD_Sharon | Tracy is working on a response to a question |
AMD_Tracy | rajkumar.iyer: to clarify more about the Business Intelligence view of the world, in a nutshell what often happens is to take a snapshot of data, offline to a different system, and process |
AMD_Tracy | rajkumar.iyer: and analyze it to make tactical and strategic decisions for the business. |
sipani | jeevesalterego: hi |
nikzzz | do you guy have kind of trick to write a bug-free and effective program? |
Parchandri | AMD_Tracy: sorry, but what does offline mean as a verb? |
nikzzz | Parchandri: hi |
AMD_Tracy | Parchandri: okay, so what I mean is that the data will go to a different system where it will no longer be live; but can be processed without affecting ongoing transactions. |
AMD_Sharon | nikzzz: We offer a profiling tool which will help optimize and tune your software code for better performance |
AMD_Sharon | nikzzz: it is called CodeAnalyst. It is availalbe for free download on www.amd.com/devcentral |
shash | AMD_Sharon: is it linux-friendly? |
Abednego | oh, cool. I didn't know CodeAnalyst was free. |
AMD_Sharon | shash: absolutely! |
AMD_Sharon | shash: it's similar to OProfile |
shash | AMD_Sharon: right... thanks |
Bi11 | maybe a free trail |
AMD_Sharon | bill: the full version is completely free |
senthilruban | AMD_Sharon: That is great. i will try now that profiler |
Bi11 | There are some others are really expensive |
Bi11 | so why it's free ;) |
shash | Bi11: it undercuts the competition... :) |
butler | quite a bit |
AMD_Tracy | Parchandri: Another comment about business intelligence; often the need is to process large amounts of data. Because of amd64 addressability, there's more headroom for crunching more data and bigger files. |
AMD_Sharon | We aren't in the business of making money off of our developers ;) |
senthilruban | AMD_Sharon: yes. i also felt like Bill that it is not free |
Parchandri | AMD_Tracy: So is Business intelligence a sort of data-processing/computational-specialised business consulting service? I don't really know anything about it, I joined the chat late |
shash | AMD_Tracy: in essence, amd64 provides a bigger hammer... :) |
AMD_Tracy | Parchandri: loosely speaking it is a set of ideas and processes to process, report, summarize, and finally analyze your business's data. |
AMD_Sharon | lol |
AMD_Tracy | Parchandri: For example, suppose you'd like to know where people are buying chili flavored popcorn the most; what part of the country? |
Abednego | are there 128-bit integers on amd64? |
AMD_Tracy | Parchandri: if you can figure that out, then you can make sure you are sending that product to the stores that need it in advance of the store |
Parchandri | AMD_Tracy: OK. So, you would want to combine all your information about popcorn sales, and pull out relevent bits. |
Excilus | Are consumers that predictable ;) ? |
AMD_Tracy | Parchandri: calling the home office to say we ran out of popcorn, send us more. Over time and a big area, this represents a lot of money at stake. |
AMD_Tracy | Parchandri: Right. |
Abednego | AMD_Sharon: so it's machine learning? |
Parchandri | AMD_Tracy: What makes a solution to this problem a AMD Business Intelligence solution? |
AMD_Tracy | Excilus: The idea with business intelligence is to make those predictions possible. |
shash | Excilus: statistically, to a large extent, yes |
AMD_Sharon | Before we go, I just want to sum up some of the main points that people were interested in with respect to AMD's Developer Program: CodeAnalyst is free on http://developer.amd.com, we offer a great optimization guide and architecture manual there as well |
AMD_Tracy | Parchandri: the important thing is our architecture is scalable and high-performing and that's what si needed in BI. |
kxyzlx | How i Know my room |
AMD_Sharon | Business Intelligence is just one area that AMD looks at that can take advantage of our architecture |
Parchandri | Well SRM's starting, thanks. |