Re: With the upcoming IIUG conference, lets put some ideas on the table...
On 14 Apr, 16:04, Ian Michael Gumby <im_gu...@hotmail.com> wrote: > Wow. > For those who are going to attend the conference, please lets try and > keep grounded and focus on a couple of things. > > 1) We learned from Diane Fraiman that 'Stealth Marketing' doesn't > work. Its really code for 'My company doesn't want to waste any money > on yet another unsuccessful marketing campaign.'. And that's really a > comment about the sales and marketing management team's inabilities > and not the fact that IBM doesn't have money to spend on marketing. > > Of course we all know that short term profitability over long term > stability only helps to line the current CEO's pockets with bonuses > and vesting stock options. Right Phil White? ;-) Sorry if I'm a bit > jaded. Even though I made a decent profit by selling out of the money > covered calls every month, I still think that Phil cooked the books to > hide the fact that his purchase of Illustra, although encouraged by > the Board, was a bit tougher to digest than originally thought. Unlike > IBM where Sam's just trying to squeeze as much as he can from the > company until he retires. ;-) > > Yet I digress. > > The point I was trying to make is that IBM's Information Management > hasn't done anything to promote any of their brands. Of course DB2 is > still synonymous with IBM and no one but Informix users and ex- > Informix users know who or what IDS is... > > So when you meet with Alyse and Inhi, please feel free to ask them the > tough questions that won't be asked in the session. (You can bet > Stuart et al will filter the questions and none will be taken from the > floor. ) Ask them what they've done besides the useless all hands > calls to 'edjucate' the sales teams. Have the changed the IT > Specialists' PBCs or their FLMs' PBCs? (Didn't think so) > > Ask them how they've put some skin in the game to attract new > customers and shrink the ever growing 'white space'... > > 2) TPC-C benchmarks. > We know that IBM will never do an Informix IDS TPC-C benchmark because > IBM does DB2 TPC-C benchmarks. Don't want to open the door to > potential FUD. And I can appreciate that. But then there's the TPC-E > which I guess all hardware vendors want to avoid a database war so we > won't see an IDS TPC-E benchmark either. > > But that's ok. > What's not OK is the fact that IBM hasn't questioned Oracle's pricing > of 10g on their latest yet to be audited TPC-C benchmark. Oracle's > pricing is pretty impressive. Even with a 28% maintenance tax, > Oracle's 3 year price for 10g is still less than half the cost of IDS > on the same hardware. > > But I'm not a guru on pricing and I'm only going from the information > posted on the TPC.org's website where Oracle and HP post their > detailed information. And of course which is what the El Reg's > reporter also read. > > So where is IBM or one of their Business Partners coming out and > saying "Gumby, you're not comparing Apples to Oranges. Oracle's base > 10g compares to IDS Express and the prices or X." Or "Gumby, good > catch. Looks like Oracle was trying to pull a fast one where they > priced out an express version and they didn't price out the full > Oracle 10g." > > 3) Solid. > Oh yes. Did you listen in on the Solid and Informix call? I asked a > couple of questions off line from Denise. He's the guy who's been > supporting Moto going back to the good ol days at Informix. > > Seems that Solid doesn't do Time Series ("TS"). So the integration > with IDS isn't that good for financial or LBS type of work. Not to > mention that IBM's new push is the idea of streaming data: (Check out > this...)http://www.theregister.co.uk/2009/04/10/ibm_system_s_super/ > > So if you were to do some sort of streaming, you'll of course want to > have some form of time series reference. I mean it would be helpful... > like if you wanted to stream all of the gold stocks, futures, > commodity prices from different markets around the world in real time. > Then of course check them against other precious metals... Time Series > would be a big help. Too bad Solid doesn't do Time Series, nor does > DB2.* > > Ok, I did a quick google on 'IBM DB2 Time Series Data Type' and found > this quote from a 2001 ACM paper... "IBM DB2 for example implements > support for time series using data-blades." Here's a link to the > abstract:http://portal.acm.org/citation.cfm?id=375808 > > And to be fair... > I did check IBM's site for 'Time Series' and DB2. Nothing, nada. No > information on an IBM extender for time series, except if you look at > third party products and even there, the information appears to be out > of date. > > So I guess IBM has lost focus on time series within DB2 and hence > since IDS supports TS, I guess it will be a while for Solid to support > it. > > Of course since we're talking about an 'in memory' database, time > series doesn't really offer much of a competitive advantage, until you > start to look at tying a query in to both Solid and IDS. > > I think that Kevin Brown should have kept working on his TS and his > linkage between shared memory buffering data until it hits the disks. > He could have spun it off in to a Solid/TimeTen like in memory > database which would have really close links to IDS. In fact, couldn't > one also take advantage of the VTI to do this too? (I don't know, I > didn't really play with VTI for a couple of reasons.... ;-) > > But hey! These are just some food for thought. > > And yeah, I've got to put in the obligatory "But hey! What do I know?" > I'm just a guy asking the questions. > ;-) > > -G I believe (from my sources in IBM Tech Support) that DB2 Timeseries uses regular tables as DB2 does not have VTI/VII to allow custom storage . So how would IDS (or DB2 or Oracle or Microsoft) compete with a dedicated in-memory + on-disk database such as KDB+ http://kx.com/? "A scalable relational database for high performance as data volumes grow" "The ability to treat streaming and disk-based data as a single database for analysis without loss of speed " See http://kx.com/papers/Kdb+_Whitepaper_2009.pdf "There is almost no discernable effect on the performance of kdb+ as you scale, even with hundreds of CPUs." "In addition, you can divide kdb+ across servers and operating systems, making use of your existing systems" "a 2-CPU server running kdb+ can handle continuous feeds involving a few billion transactions a day" "Kdb+ queries in-memory data at the rate of over 200 million records per second per CPU and updates data in memory at the rate of 1 million inserts per second." What is IBM's response to this?
david@smooth1.co.uk wrote: > > What is IBM's response to this? Have you heard about TimeSeries and the Real Time Loader? -- Cheers, Obnoxio The Clown http://obotheclown.blogspot.com The following is a disclaimer and a protest at the collection, retention and sharing of my personal mail by the morally bankrupt state. By adding a string of key words, it will guarantee that each and every mail that I send will now need to be manually viewed as it is picked up by the auto scan software. If every person in the UK does exactly the same, then the entire system will quickly become so unmanageable, so unwieldy that it will become unworkable. Key words: bomb, assassinate, president, brown, Osama, Obama, Sarkozy, Merkel, government, target, location, rocket, grenade, al-Qaeda, Pakistan, India, Afghanistan, UK, America, guns, jets, bombs, machine-gun, terrorists, MP's, pigs, troughs, France, Germany, Italy, nuclear, Korea. PPS If you are reading this because you've picked up the above keywords: I hope your parents are proud of what you have become. -- This message has been scanned for viruses and dangerous content by OpenProtect(http://www.openprotect.com), and is believed to be clean.
> From: david@smooth1.co.uk > Subject: Re: With the upcoming IIUG conference, lets put some ideas on the table... > Date: Tue, 14 Apr 2009 13:51:44 -0700 > To: informix-list@iiug.org > [SNIP] > > I believe (from my sources in IBM Tech Support) that DB2 Timeseries > uses regular tables as DB2 does not have VTI/VII to allow custom > storage . > Well I'm not sure how DB2 would set up time series. I mean you could create an extender that is built over a blob for each time series. > So how would IDS (or DB2 or Oracle or Microsoft) compete with a > dedicated in-memory + on-disk database such as KDB+ http://kx.com/? > IDS already does compete with the real time loader, at least with respect to the inserts. As the data is inserted its first in a shared memory segment and then gets written to disk asap. While its in memory, you can query the data as if its in the table or so I'm told. Its after the data is written to disk that you don't see the shared memory as a cache. I guess Kevin Brown could have taken his work a step further but then the IBM takeover happened. Its a bummer in that all the "great press" with Solid working in concert with IDS is old hat. Informix could have had it developed back in 2001-2002 time frame. Sucks, doesn't it? I don't know if this is really an issue. While in memory, you could just keep the data within smart objects and not have to worry about an in memory database. Of course you're talking about some custom code and people who want this sort of speed can afford the cost of development. I mean for LBS you're talking about creating a map context in memory and then as you modify it, you would keep a sort of history, with maybe some form of i/o stream to an underlying database. I've seen the prototype of this and its impressive. The real kicker is the amount of memory you'll need for a basic map context and then you could create a mash up of the different layers of data as well as multiple live copies of the map context on different servers. Kind of interesting. _________________________________________________________________ Express your personality in color! Preview and select themes for Hotmail®. http://www.windowslive-hotmail.com/LearnMore/personalize.aspx?ocid=TXT_MSGTX_WL_HM_express_032009#colortheme
On 14 Apr, 23:37, Obnoxio The Clown <obno...@serendipita.com> wrote: > da...@smooth1.co.uk wrote: > > > What is IBM's response to this? > > Have you heard about TimeSeries and the Real Time Loader? > Yes. > -- > Cheers, > Obnoxio The Clown > > http://obotheclown.blogspot.com > > The following is a disclaimer and a protest at the collection, retention > and sharing of my personal mail by the morally bankrupt state. > > By adding a string of key words, it will guarantee that each and every > mail that I send will now need to be manually viewed as it is picked up > by the auto scan software. If every person in the UK does exactly the > same, then the entire system will quickly become so unmanageable, so > unwieldy that it will become unworkable. > > Key words: bomb, assassinate, president, brown, Osama, Obama, Sarkozy, > Merkel, government, target, location, rocket, grenade, al-Qaeda, > Pakistan, India, Afghanistan, UK, America, guns, jets, bombs, > machine-gun, terrorists, MP's, pigs, troughs, France, Germany, Italy, > nuclear, Korea. > > PPS If you are reading this because you've picked up the above keywords: > I hope your parents are proud of what you have become. > > -- > This message has been scanned for viruses and > dangerous content by OpenProtect(http://www.openprotect.com), and is > believed to be clean.
On 15 Apr, 00:31, Ian Michael Gumby <im_gu...@hotmail.com> wrote: > > From: da...@smooth1.co.uk > > Subject: Re: With the upcoming IIUG conference, lets put some ideas on the table... > > Date: Tue, 14 Apr 2009 13:51:44 -0700 > > To: informix-l...@iiug.org > > [SNIP] > > > I believe (from my sources in IBM Tech Support) that DB2 Timeseries > > uses regular tables as DB2 does not have VTI/VII to allow custom > > storage . > > Well I'm not sure how DB2 would set up time series. I mean you could create an extender that is built over a blob for each time series. But doesn't IDS Timeseries use custom storage (containers)? I never did understand how they worked in terms of storage exactly and the Timeseries/RealTime loader manuals do not go into any details... > > > So how would IDS (or DB2 or Oracle or Microsoft) compete with a > > dedicated in-memory + on-disk database such as KDB+http://kx.com/? > > IDS already does compete with the real time loader, at least with respect to the inserts. KDB used compressed files and IDS doesn't do compression for timeseries... > As the data is inserted its first in a shared memory segment and then gets written to disk asap. While its in memory, you can query the data as if its in the table or so I'm told. > > Its after the data is written to disk that you don't see the shared memory as a cache. I guess Kevin Brown could have taken his work a step further but then the IBM takeover happened. Its a bummer in that all the "great press" with Solid working in concert with IDS is old hat. Informix could have had it developed back in 2001-2002 time frame. Sucks, doesn't it? > Yes. > I don't know if this is really an issue. While in memory, you could just keep the data within smart objects and not have to worry about an in memory database. Of course you're talking about some custom code and people who want this sort of speed can afford the cost of development. > > I mean for LBS you're talking about creating a map context in memory and then as you modify it, you would keep a sort of history, with maybe some form of i/o stream to an underlying database. I've seen the prototype of this and its impressive. The real kicker is the amount of memory you'll need for a basic map context and then you could create a mash up of the different layers of data as well as multiple live copies of the map context on different servers. Kind of interesting. > > _________________________________________________________________ > Express your personality in color! Preview and select themes for Hotmail®.http://www.windowslive-hotmail.com/LearnMore/personalize.aspx?ocid=TX...