Once upon a time, I was known to the world (that cared) as 'olap genius'. On Yahoo, a forum was maintained in which I engaged in outsized bloviation and championship (as opposed to FUD) about the Arbor/Hyperion products, industry and technology. I've read about five or six entries of this bloviation and I must say that I'm rather shocked. I now see quite clearly why people called me arrogant. But, I did know what I was talking about for the most part, and it was very reasonable for my old boss Dan Druker to have me turn off the spigot about the new stuff we were doing in eCRM at Hyperion.
I see now in retrospect how well I should have fit into what was then the burgeoning open source movement. As I have read the book, it's clear how my communication style was influenced by the spirit of the times - this is just the kind of fractious and fanatic attention to detail and passion that generates top quality stuff from Type A personalities. This was precisely the kind of spirit that energizes software technology on the cutting edge, which was where we were during the OLAP wars. I don't miss, nor do I want to romanticize those days. That was then. But I did want to put all of this stuff out there on the record. I mean, I said it in public then, there's no reason for me to not republish it now. That said, I haven't really reviewed it and I have no idea how embarassing it might be to me now.
f? by: olap_genius (30/M/California) 01/22/99 08:58 pm Msg: 1402 of 12794
ok so rbdms can do a ragged hierarchy. you use a circular method, now i remember the guys at dell do it in oracle. but guess what, they are an essbase shop and they threw out sap...
essbase can do read/write on ragged hierarchies and allocate data loaded high in (user defined) fractional amounts to multiple levels by naming generations and/or attributes. (with security) show me the sql to do that and you can join joe celko on the standards committee.
2. i don't have to read ralph kimball's book because sagent is a hyperion partner. the people at sagent confide in us and understand essbase's unbeatable strengths.
3. yes it is huge. we do it with security. ask the guys at ibm santa teresa labs how they implemented it on top of db/2. they used essbase source code and bypassed all the sql interpretive stuff and talked to the storage manager layer. (roughly speaking)
plato integration by: olap_genius (30/M/California) 01/22/99 09:12 pm Msg: 1403 of 12794
i feel sorry for the grunts out there who are bent on trying to get plato to work with excel. especially considering the abdication of bruce mckinney, the former god of visual basic, in light of all the reinvention of database access the msft is attempting. the ugly gets uglier. i wouldn't touch it with a 10 foot pole. cognos deserves the desktop, except for one thing ---
hyperion wired for olap is twice as nice, plus it does java automatically, plus it talks to essbase and plato.
express should have been gone 3 years ago. just ask the ex oracle people at hyperion.
budgeting planning and consolidations, as we agree, are much better served by mddbms. but i believe these are just the tip of the iceberg.
centralization and scalability by: olap_genius (30/M/California) 01/22/99 08:45 pm Msg: 1401 of 12794
are important issues. i just hope you're not the same person who was talking about plato, which will never run on unix. anyway, it's common knowlege by now that essbase can calculate a 2 GB cube on a garden variety 4 banger pentium 2 nt server in about 30 minutes. it's also common knowledge that there are 10GB essbase cubes in production all over the place. furthermore it's common knowledge that hyperion is a partner with acta, sagent and informatica who do everything with sql that mere mortal relational dba's wouldn't dare on their own.
furthermore it is common knowledge that latest version of peoplesoft budgets is built on top of essbase. and it is common knowledge that both ibm and showcase (the kingpin of the as400 space) have built their olap strategies on essbase.
it's something we geniuses don't sweat. but it's sometimes disappointing to know how slowly common knowledge trickles out to the masses.
there really is very little, marketwise, that doesn't fit into the performance envelope of essbase. it's just that a lot of people keep throwing around figures like one terrabyte of data, as if they really knew what kind of applications actually use a terrabyte of data.
< Previous | Next > [ First | Last | Msg List ] Msg #: Reply Post Recommend this Post Ignore this User | Report Abuse IT Gauntlet by: olap_genius (30/M/California) 01/23/99 04:22 pm Msg: 1411 of 12794
dead right about the hysw pitch. and of course those companies remain happy. sometimes it's hard for people who come out of cs degree programs in the 90s to understand how america managed it's money before the days of 200MB hard drives. then again, think about it this way: how much space and computing power would you guess it takes
to keep the stock prices for the s&p 500 going back 10 years? that's palm pilot stuff, guys. lots of great decisions can be made with very little compute power. it's all about implementation.
nevertheless, the data warehouse idea refuses to die and a dumptruck load of freshly minted dbas play gleefully in fantasies of galactic repositories. the hub and spoke architecture is one of the first things that makes commercial sense, but a glance at kimball's private email distribution makes it clear how many people are still just getting their minds around that, and how many are still clueless about dss. be that as it may, presumptions about scalability persist.
i think that there are several good reasons for that, and they all have product names. holos, express and gentia. on gentia: i can remember when (two years ago), hotshots from cambridge technology partners were mouthing off about how full-cycle object oriented
databases and development environments were going to devour the market. well. it didn't happen. but scalability was an issue there. (company scalability too (!)). with holos and express, the finest of the 4GLs which survived the 80s, scalability had also been a problem, and especially with express, product integration.
if i were in the relational business as a dba, i would have had the same cocksure attitude. i can specifically recall marketing information applications which came out of proprietary querying systems which choked db/2 on the biggest 390s back in the days when F500 IS departments didn't trust UNIX for mission critical apps. but that stuff was 'only' 1-3GB. so let me quantify it this way: in 1993, it was damn near impossible to hit a 6way star in db/2 enterprise on a 390/600 with 10 concurrent users, unless you did very tight query management and worked with static binding.
redbrick did that kind of stuff, but very few people knew how to implement that. so companies outsourced.
i've got to go but the bottom line i want to express here is that the capacity of essbase has just about outstripped the ability of corporate IT to invent mission critical DSS. there are very few (if any) systems that make strategic sense to build which cannot be handled by essbase and the tools which work with it. IT always asks for 'detail', end users never query it.
OLAP Boring? No Way! by: olap_genius (30/M/California) 01/29/99 01:37 am Msg: 1430 of 12794
You know what? I don't care about the stock market so much. Some folks just never get it. What I really care about is that HYSL has the number one products in the market. even if the market doesn't blow everyone's mind, whatever, we will be the leader.
real soon, people are going to realize 2 things. one, they will finally get their transactional systems in place or die. that's what all this fuss is over right now. if you haven't done it by y2k then fire your cio and outsource because you're pathetic. client/server transaction systems are a 90s problem.
number 2, now you need to *do* something with all that data. it's going to be all gravy for hyperion by then. why? simple. everyone will have the web - and plenty people will have high speed access. they're going to start asking common sense questions like, how come i can get a query back from yahoo in 5 seconds when it searches the entire web, and i can't get profitability by product from my company system *at all*? or better yet, how come i can get webtv at circuit city but i can't get intranet data from my corporation? time's up for baby olap. it's time for the grown-up industrial strength stuff. you have 11 months... tick, tick, tick.
oh by the way, those of you with sap can use sagent to automatically build essbase cubes. erp to olap plug and play, baby! that's exciting.
elliptically speaking.. by: olap_genius (30/M/California) 02/03/99 05:02 pm Msg: 1469 of 12794
imagine you were a hot little software company full of very sharp programmers who got bought out. it's heaven as you count you pennies. then you realize you don't have control and warm&fuzzies you are merged into a company x times your size. how do you leverage the benefits of being a very tight knit group within an international corporation? how do you work with 'competitors' in the same industry?
subsidiary. why spinoff?
<censored> by: olap_genius (30/M/California) 02/03/99 10:59 pm Msg: 1474 of 12794
i don't want to be responsible for saying anything more than i've said, because i've actually heard tell within the company of a potential alliance with this other company back in the arbor days.
i'll say this, it's not a current partner, but it is a well known company. the synergy is not immediately obvious, but also not a stretch for a, say 600-700 million dollar software company. it would also piss oracle off big time.
like i said before - i'm not an official insider by any stretch. and i don't know anything. but i'm going to keep my mouth shut because i did hear this name tossed about before the hyperion-arbor merger.
kiss by: olap_genius (30/M/California) 02/04/99 02:58 pm Msg: 1479 of 12794
i bought a pile at 15. sold some higher priced stuff and lost about 2% last fall. but i still have a good stake around 25 too. so i'm mostly happy but i was expecting a lot more. i think it can do 40 easy right after jan 1, 2000.
i try to stay cool about the fact that i missed BCST by one day, and bought CPQ instead of DELL. but overall my stocks are doing great. it's hard for me to complain. (well except for my one dog, GMGC - but i'm holding that forever. they are going to *own* voicemail in 2-3 years) besides, i bought it with gravy from INKT which i got at the ipo.
another secret i like is a certain infrastructure provider who builds PCS and cellular
towers. i'm also anticipating portal wars in the movie production business.
does anybody here know how i can get my hands on some subaru stock? auto mergers man!
Fries with that... by: olap_genius (30/M/California) 02/11/99 03:26 pm Msg: 1522 of 12794
SPSS is already a partner. Their whole extended analytic package works with the essbase excel front end today. i haven't seen it work, but i imagine that people can do anything with it. the difficulty with regard to putting any more effort into that is that most of the fortune 500 is not *that* analytical. most folks minds boggle at about the 7th dimension of data, add box-jenkins or least squares regression onto that and you get serious eyeglaze.
acquisitions are very different that mergers. the appsource acquisition was almost a no- brainer. i expect that a well integrated product which already has some market share out there would probably make for easy picking. especially if they don't come top heavy with management and shareholders as did hysoft. the wholly owned subsidiary thing seems to work just fine.
as for engineers leaving hysl. well, i hear that's true too. but really, what is there left to build? essbase is baked. (and they just cranked out a new release 2/5 which at long last includes the 32bit application manager).
all that's left to build are turnkey applications. relatively speaking, that's child's play as compared to building a serious olap engine. all the hot analytic applications with marketshare are either already built on (fast) essbase, or (sluggish) relational platforms - sales&marketing, balanced scorecard, activity based costing, category management, asset management... hell, people even do that stuff on as/400. now it's just a matter of putting all the pieces together and executing.
who's your waiter? by: olap_genius (30/M/California) 02/11/99 05:01 pm Msg: 1525 of 12794
if essbase is being served up to you half baked, don't blame the chef.
1. parallelism is achieved through the partition manager. you can break up a single application across cubes (each cube owns an independent process & memory space) and across machines.
parallelism within calcs would be nice, if anybody needed it, considering that on a 4 processor nt box you can calculate and write 2GB/hour.
2. yeah you're right. 3. been there, done that in version 4.
4. transaction rollback can be done with the db2 olap server. it's all on db2. again, as if anybody really needed it.
5. integration server talks metadata with acta, sagent, ibm visual warehouse and informatica. it's the defacto industry standard olap metadata. no other vendor even has that. whenever the industry standard repository is done, integration server will talk that language. 'massive' is relative. ibm visual warehouse handles all metadata back through ETL, very little of that is required for olap architects, cube builders or end users.
6. huh? do you mean allow end users to select their favorite query language, or that there are queries which can't be done against essbase?
erps by: olap_genius (30/M/California) 02/12/99 12:55 pm Msg: 1542 of 12794
i'd hate to be in the erp space right now, stockwise. i think the glow is off. irl, implementations are nightmarish - only reflecting the goldberg state of the pre-client server world. i hear sap has a new quick-start implementation cycle. but we've all seen what's happened to peoplesoft.
what i have seen is the beginning of some of the olap technology begin to creep backwards into erp. these 3000 table monster baan impls may be disintermediated. same with jde. that takes some IS moxie, but it can happen. plus i think more enterprise software vendors are building links into erp. so as tools proliferate which talk to warehouses, various erp schemas and olap tools, the entire space will get squishy. that can only serve to confuse things which means, in the end that peoplesoft, baan, jde, sap will no longer seem pure plays.
on the other hand that might limit hysl as it grows. right now it doesn't seem to be traded in the same portfolios as the big erp vendors even though it takes hits with oracle. also cognos is growing up. so i'd probably look for more analysts to groupthink hysl and cognos with the erps as those two companies grow.
fwiw...
mint jelly and chili sauce.. by: olap_genius (30/M/California) 02/12/99 01:18 pm
Msg: 1543 of 12794 ..the continuing saga of essbase features
1. parallelism. i'd just make one qualification on that 50GB input data bit. i can't imagine anybody doing that much incremental data per cycle. what kind of apps do that? if that was 50GB/month, i'd definitely build a new cube and expect a weekend to crunch it. but i'd also definitely have that size on a 6 to 8 way ES5000 or bigger Sun box. and for complex calcs, essbase is the acknowledged champ (except for simultaneous equations as the express boys will gladly crow).
2. enterprise security. dead horse. i concede. however, esspro does allow you to tie essbase to nt domain security so users don't have to login twice, and the new wired allows you to import and export user lists. and there is no integrated security for drill-through to relational.
3. cached retrievals since v4 you could always change the size of your data cache which is where blocks retrieved from disk hang around in memory. there are some select events which flush and commit that cache. in v5 you can change your commit rate back to disk. all that is independent of your storage type {dynamic, dynamic & store, store}.
4. rollback. when you think about rollback in the context of essbase olap, it only makes sense at one point. that is on batch inputs which *don't* overwrite, but add/subtract against existing data *and* are *not* marked as adjustments. from a strict atomicity standpoint, a good dba wouldn't allow such transactions. generally you would just re-run the load. again almost nobody does real-time loads against essbase so re-running the job is a no-brainer.
so that leaves just one place for exposure which is end-user updates. you can either trap that in your front end, OR you can use the excel audit trail, which has been standard since v4. it creates a log of all lock&send operations by user. ugly to roll back, but possible.
(cont'd)
..cilantro, salt & pepper.. by: olap_genius (30/M/California) 02/12/99 02:47 pm Msg: 1544 of 12794
..the continuing saga of essbase features/futures..
5. metadata storage yes, large outlines (>50K member/dim) have been a pain in the 16bit appmanager. in 5.0.2 the new 32bit app manager is said to significantly reduce that pain. i haven't got my mitts on it yet, but i hear that certain light restructures are (it couldn't have finished
already) fast. 6. pass, awaiting further clarification
7. drillback no, that assumption is mistaken. integration server works just like you would expect a non brain-dead development org to build it. there is no redundant storage, just a context sensitive mapping from the 'bottom' of the cube back to the 'top' of the warehouse detail. in this way it makes essbase a true hybrid. essbase does no processing when a drillback is called, other than formulating the sql. plus end-users can modify the query (under desiger control).
8. extended attributes. funny you should mention that. next release you can do attribute cross-tabs. i've heard the term 'virtual dimensions' tossed around. we'll see.
9. spss has so many products, i can't remember which one works with essbase. but hey, they're only a market cap of 150million. hmmm.
10. plato. gimme some elbow room...
i hear what you're saying about stock options and all that. but i think (and hope) that sooner or later these *engineers* get real about the so-called killings they can make in the software business. maybe some of them need a dose of reality, working for eds or something.
..& hemlock by: olap_genius (30/M/California) 02/12/99 03:03 pm Msg: 1545 of 12794
plato.
i suspect that the olap world will be big enough for plato and essbase. but i consider plato the 'other side of the tracks'. from a high-tech bigot point of view, plato is mostly cheap, microsoftian, stuff.
let me put it to you this way. there are lots of F500 companies out there who have people who program MSAccess in on their datawarehouse planning committees. such companies will probably go with microsoft. and through marketshare, microsoft will remain a player. but ms will not play hardball in the bigs. they won't go unix, they won't go as/400.
there's plenty of things microsoft can do, but one thing they definitely *won't* do is lower the price point for serious olap. consider the fact that oracle for nt has a bigger marketshare than sql server. and if it comes down to war, hysl can always play dirty and change their pricing to be platform or processor dependent.
i say let microsoft have the low-end of the market. so what? has sql server even caught up to sybase yet?
MSTR Market by: olap_genius (30/M/California) 02/12/99 10:33 pm Msg: 1555 of 12794
well please help me out here. this stuff that microstrategy has been 'handling for years'. what is it? what kind of applications are we talking about because i just don't get it.
i have a very difficult time reconciling the very concept of multidimensional analysis to >10GB input streams/month. i have a hard time visualizing what kind of person looks at that much data at a detailed level. or maybe i don't understand why such volumes aren't mined ahead of time.
and i also don't understand what value this level of detailed scrutiny provides over the long term. it seems to me that once you notice something wrong at the transaction level that you immediately abstract the problem and never need to look at that detail again.
again olap is pointed at the business planning cycle. the presumption is not that you just find something in the detail, but that you slap a kpi on it and monitor it. again, i have a hard time picturing any management attention at that level.
contextualizing what i mean might be helpful. for example. if i sold 2 million vehicles a year and tracked them down to the VIN, i could do a 7 dimensional model in essbase, keeping 2 years of history in a cube smaller than 15GB total. the only way to make that cube much bigger is to add more demographic attributes to the customer dimension. and to make it huge, i could map all 2 million VINs to all the individual customers. but my razor is this: how detailed a marketing plan do i actually make? i could build a model which will tell me how many women over 30 with mortgages in seattle buy a particular option package of camry. after i crunch my 205GB hybrid olap cube, the answer is 47. so who makes a promotion plan for 47 women? (except maybe the seattle area dealers, who don't need national stats) you see my point?
the tiniest bit of thoughtful sampling and abstraction makes all hybrid olap superfluous except for audit purposes. but that doesn't negate the need for transacation processing - admittedly another market.
plato and vb by: olap_genius (30/M/California) 02/12/99 11:04 pm Msg: 1556 of 12794
what i really didn't want to do is get into a holy war. but what the hell? here's my gut. *the* front-end to plato is going to be vb. #2 will be access stuff. #3
will be excel/vba. now everybody in the vb world is starting from ground zero with the new msft 'oda' data interface. (is that the name?) but basically, it's all new and untested. meanwhile the essbase api has been out there for years. (check out all the tools partners) http://alliances.hyperion.com/partners/tools.html
considering msft's track record with vb, it's going to be a long time before plato interfaces are going to be robust. the popular wisdom says that all read/write to plato is kludgy. and it's pretty much a sure bet that microsoft is not going to get domain expertise and build analytical apps. they'll just energize the vb grass roots and wait for a killer app. good luck selling that to corporate IS.
there's another bottom line here too. that is that microsoft will *not* outspend hysl on r&d.
what i'm waiting for is a plato run at the apb-1 benchmark which essbase has run publicly 3 or 4 times. when they show up with a respectable score, then i'll shutup.
one word by: olap_genius (30/M/California) 02/14/99 01:13 am Msg: 1563 of 12794
one word about the theory that cheap database technology would take over the apps world because developers love it.
watcom. i rest my case.
straight queries by: olap_genius (30/M/California) 02/18/99 10:58 am Msg: 1586 of 12794
just back from another jaunty jaunt. sometimes i stand incredulous at the lack of sophistication with which organizations one would consider brilliant actually implement their IS infrastructure. and understanding this gives me chills, and a certain amount of respect for what must obviously be the marketing strategy of the baby olaps.
"just put it in oracle and query it"
from now on, i will be strictly bigoted to reverse engineer all data warehouses from the business end backwards. relational dbas are useless. ok well not useless, severely challenged. but, then again you can't blame 'em. they've got a hammer.
the biggest problem i've ever had in solving all a company's business monitoring
problems for management is finding the data. the baby olaps sell their products as if that was a trivial exercise.
turnkey apps vs olap tech by: olap_genius (30/M/California) 02/20/99 03:02 pm Msg: 1604 of 12794
the point of essbase is to enable an enterprise wide architecture of 'speed of thought' response time. that's what hyperion calls the analytic platform. they are after all of the realtime interactive programs that are currently built on relational technology under the rubric of 'data warehouses'. however most people don't get it, and so market penetration resembles a bell curve.
at any rate, there are other products out there - synon, holos, a program called invision (or envision) that works with peoplesoft, tm1, metacube, sybaseIQ, and a half dozen other products which perform an order of magnitude slower than the two market leaders, essbase and express. plato is a wildcard - they haven't submitted to an audited benchmark yet. in the meantime essbase has improved its performance over the past 3 years by a factor of 42!
so there are lots of people who have consolidations (85% of the us market is hyperion enterprise) and budgeting solutions from hyperion who still are living in the past with regards to having a true enterprise scalable analytic platform. some do it for logical reasons, apple for example has endorsed seagate holos as its dss strategy - they're in bed together, others are just not ready yet, or haven't spent enough time and money to build serious systems.
whatever. the bottom line is that if you don't already understand olap because you are working with it in your place of business, they way you might understand 'oracle' or 'netscape', then technologically speaking, you are behind the times. (the word i prefer to use is 'retarded', but i'm not trying to flame).
here's a customer story from real life that i discovered in the past week. a certain automobile manufacturer has it's own credit acceptance company. it tracks 24 months of payments on every vehicle it finances in america. (funny, we just talking about this kind of example). and in a 9 dimensional model can create a loss triangle in 2 minutes. now unless you understand the credit analysis business, a 'loss triangle' may sound greek to you, but it is basically the industry standard method of answering the question 'over the life of a loan, for every 100 dollars we lend we will lose 1.98 in defaults'. it used to take 4 days for a department to create ONE, for any particular credit rating, loan or lease term, make and model, etc etc. now it takes minutes. again, this is not a packaged application like pillar or enterprise. it was built from scratch based on the needs of this one company using essbase. that's a real olap application.
i specifically asked about the size of the data, since relational types are always talking
about how big they are. these guys crunch 6G of historical data once a month. they cleanse, calculate and verify this data in a 20 hour period, lights out on a monthly basis completely rebuilding 2 years of history. they do so on 2 compaq 6000s and then have 160GB of live queryable data for users nationwide. response time is immediate - they have up to 100 users on the system simultaneously.
if you can't query 160GB of data for online realtime analysis, then you have inferior olap technology. there are a large number of businesses out there who cannot conceive of building these types of applications. but the olap market is predicted to be about 3- 5 billion in a couple years. hyperion is the market leader with half a billion in revenue projected for this fiscal year. we're still in the area where the majority of people don't 'get it'. a big reason for that is relational fud. another reason is that people actually believe that ERP systems will give them this capability, others are pissing away dollars on y2k and cobol...
momentum by: olap_genius (30/M/California) 02/20/99 03:48 pm Msg: 1605 of 12794
i'm not aware of what this product is. i am aware that peoplesoft budgets 7.5 has an interface to essbase / that it is built on top of essbase. but i have a suspicion that third parties may even do better in integrating ERP systems with olap.
that said, peoplesoft is a hyperion partner. they may have some non-comptete agreement. i don't know. does anybody know what ERO stands for?
reverse engineering by: olap_genius (30/M/California) 02/20/99 04:07 pm Msg: 1606 of 12794
i'm mostly serious, and that is because i have encountered few warehouse architects who understand olap as well as they do relational. but i have met many cfos and contollers who have had to pull teeth to get transaction systems disciplined and on the whole would rather not deal with IS - that is until they get essbase, and start explaining what all the data *means*...
at any rate, i'm a bigot and that's fine. i always have to work with end-users and relational dbas both. it is inevitable and without question that the efficiencies of essbase olap will never be matched by relational engines. hyperion integrates the two better than anyone - so the design is cooperative. BUT...
more and more customers are buying into the concept of purpose-built warehouses. so there's a financial data warehouse, a manufacturing data warehouse, a distribution data warehouse, etc etc. these can and will be reverse engineered. i'll bet you a nickel. the reason they will be is because IS departments will eventually become olap savvy, and they will be able to deliver the analytic reporting platform. jad sessions will become more productive.
i mean really, if cios can get the budget, all the best of breed tools are there and obvious. i could lead a team to scratch build any type of warehouse (given ETL, ERP, RDBMS, OLAP, & thin clients) in 90 days. --
i think the fixation on dillon is like the fixation on bill gates. perhaps that is rightfully so. but the company runs because of dillon - why the stock market runs is anybody's guess. i'm loading up too.
dillon rich? by: olap_genius (30/M/California) 02/20/99 06:45 pm Msg: 1609 of 12794
he's not the insider selling off huge chunks of stock. that's perakis. my understanding that all of the hysw sales execs who split were already millionaires. anyway, the business is growing still. i wonder if the ceo of peoplesoft hears this kind of acrimony...
who are the big shareholders anyway? amerindo? former principals at arsw? also, to this day, i've never heard any hyperion employee gripe about their salary.
a buyback? i thought about that last week. it kinda makes sense, especially if ibm were a suitor. the arbor folks were glad that they were able to survive the original ibm agreement without getting swallowed. i wonder how they would feel about it today if they could get their options repriced and ibm paid 1.5b for the whole company and assumed the miniscule bond debt.
bill binch, current head of ww sales is an old ibmer. he'd know the rules. but i think the channels org would be savaged. that's a big ugly. international would be diluted. that too would be a big ugly. the new sales force is just starting to gel. scrambling their turf again would destroy the company. you'd turn everybody from an employee to a shareholder. greed isn't always good.
i wonder how the old tivoli folks feel.
PSFT & Analytics by: olap_genius (30/M/California) 02/20/99 06:50 pm Msg: 1610 of 12794
agree 100%. but who can affort psft? they'll dry up slowly, like baanf or maybe merge with catp & sybase. that would be an interesting turn in the industry. imagine a company like unisys + shl systemshous + redbrick in partnership with pricewaterhousecoopers. hmmmm.
2000 at 14
by: olap_genius (30/M/California) 02/20/99 10:20 pm Msg: 1617 of 12794
i think you can't lose. my personal opinion is that hysl is a bargain at anything under 25. i'm throwing money at it at this level. then again, i'm just an amatuer investor. the company is in good shape and is doing more big deals. revenue is still lead by new licenses and olap technology. for some reason this freaked out a lot of the hysw value folks.
anyway there's no accounting for market psychology. i know if enough people with enough money hate hysl performance, then the price will fall. the bottom line is that this company is the market leader in this space, and the space is growing. my nickel says that as soon as they post a quarter which beats the street, all will be forgiven. then we're back in the 20s. when will that be? who knows, but it will definitely be before next year. this is definitely a post y2k company.
this and that by: olap_genius (30/M/California) 02/21/99 10:45 pm Msg: 1630 of 12794
y less than 25-30 bucks which is the low end of the average shareprice last year. it's absolutely rediculous to presume that a company which generates $377 million in revenue per year at software company margins can only be worth 500 million which is about the current mindless valuation.
ibm is going to bed with all the baby olaps because they are desparately attempting to buy a clue. believe me when i tell you that hyperion has been doing all the brainwork in their joint sales ventures. once ibm is rolling and its marketing gets rolling, it's all over for oracle. they missed out on the ERP market, they refuse to miss out on business intelligence.
what you don't seem to understand is that brio is to ibm as powerbuilder is to sybase. just another front end. the money is in the database business, not in the front end business. besides, in case you haven't noticed, the trend in the industry is for front- ends to get cheaper and cheaper. front ends become commodities over time, database technology never becomes commodified. so who cares if ibm is sleeping around with a lot of baby olaps? that's inevitable and ultimately trivial. try to remember that wired for olap, the premier front end tool for essbase and plato was developed by a company <50 employees. and seagate is giving its front end away for free.
good shape by: olap_genius (30/M/California) 02/21/99 11:01 pm Msg: 1632 of 12794
alysts any credit for understanding the business - not for a company this size, like analysts of gm or chrysler understand the auto biz.
the olap report, the datawarehousing gurus, IDC, the meta group, gartner (to a lesser degree), these are the people who understand what hyperion brings to the market.
i worry seriously about what would happen to a hysl below 12. very ugly things could happen to the entire industry if the company got swallowed, but there would have to be a huge bidding war first. a lot of hysl people have been in the dss industry a long time, there are lots of comshare, irms, metaphor, holos, pilot and other folks who have survived. i believe a lot of them would be very pissed at any monkey business.
startup by: olap_genius (30/M/California) 02/23/99 01:02 am Msg: 1642 of 12794
well, i exaggerate. it takes 90 days get 90% of the F500 companies to make up their minds about what they need to know. and it takes another 90 days to get the end-users and IT people to stop squabbling and agree to cooperate. but i do not kid about the technology.
i have a cousin who builds telemetry databases for a living. he won't tell me any more - it's all hush hush. but he assures me that building 50TB databases is not difficult if you have the right people. of course he spends years going through the gov't procurement process.
now think about bloomberg. there are no corporations in the f500 who have the kind of data requirements of investment bankers and other capital markets players do. they have pushed the envelope way beyond what your ordinary corporation deals with. go to any major corporation and ask them what it would take to do a weekly p&l by product. their finance departments and is departments would freak. but that's just because they are living in dilbertville, not because the capability doesn't exist. bloomberg has proven that he can crunch the analytics of the world's most demanding users, and now most of his tech is just two steps above commodity. he's being undercut in price by a factor of 10. can any of you remember how hard it was to get 15 minute delays on stock quotes for say 50 stocks in 1993? today it's free.
y2k money is starting to free up. the f500 will start to have lots of processor cycles and free space on their big iron, not to mention bucks.
i'll tell you what's going to do it. more people will start to see things like hyperion's live demo over the web, think about the things that search engines do and start asking very tough questions about their data warehouses. then they're going to see some second generation olap implementations like integrated sales & marketing, supply & demand and supply chain. then they are going to ask very simple questions like. why can i browse the web and find out about anything at yahoo in 2 minutes, but can't even get my corporate reports in 2 days?
maybe heads have to roll.
btw, my prediction for the next big thing: continuous budgeting cycles, instead of one big budgeting mess once a year. really when you think about it, this is what they do in capital markets every day. the f500 has just decided to take 15 years to catch up. (psst - it's because IT managers were afraid of UNIX, remember those days?)
something scary by: olap_genius (30/M/California) 03/05/99 10:05 pm Msg: 1716 of 12794
I can't talk about it other than to allude to the fact that I have just recently seen something scary. it was essbase running on a HUGE machine. it's making me think on a whole new level.
i'm also beginning to appreciate the skepticism of relational bigots when they speak of monstrous amounts of data to warehouse. i kinda feel sorry for them - they honestly don't know what to do with all that data. this in no way impedes upon the olap space, but i'm beginning to understand the headaches associated with managing upwards of 25000 relational tables in an enterprise.
anyway, the guy i was having this extended conversation with loves essbase. he was expressing what he believes is a common problem - and that is how does one determine *which* data to take out of production systems and stage in a warehousing environment. tricky question when you are planning years in advance of end-user reporting and analytical requirements. when the volumes are massive, the number of tools one has
to work with are few. what amazed me was the number of trivial(?) data stored in transaction systems (for this company).
at any rate, it is clear that the hyperion integration server goes a long way in managing this problem. as i speak with more ETL & DW design folks in the next couple weeks, i'll share with the board.
any tech geeks around here heard of sand technology? they're on the cover of dm review. interesting stuff.
commodity my butt by: olap_genius (30/M/California) 03/16/99 04:46 pm Msg: 1822 of 12794
the more i learn about data warehousing, the more money i believe hysl is going to make. i've gotten an opportunity to get my mitts on informatica, which is one of the leading
ETL tools in the market. pretty hot company too. the product is elegant and powerful, and it sends metadata over to essbase such that you can build cubes.
the short of this is that now i think i have a handle on what relational dbas go through in trying to decided how to create a datamart. (and what scrubbing a production system is like, YIKE) without a tool like informatica's power center, it has GOT to be a nightmare. even with power center, it's no picnic. on the other hand, with essbase's application manager, it's a walk in the park.
what a LOT of people out there in the d/w industry obviously don't understand is how easy it is to build a datamart directly with essbase - they are bogged down with far, far too many issues that they don't realize are truly no-brainers if you have essbase. now that's not an easy message to market, and that is because of the 'olap schmolap' ideology wars, myths of relational scalability, the 'detail' question, and the success of the relational query tool market.
the answer is actually truly simple, and that is to buy into the analytical platform idea. clearly essbase is the leader, technologically speaking to work as the analytical platform. maybe if we called it the end-user query platform, warehouse guys would get it. because the bottom line is that there is no query that the warehouse needs to expose, except those which are batched onto the analytical platform. if more d/w architects would realize
this, they'd spend more time smartly mining production systems and building flexible repository strategies. the problem is that few of them have the tools for and experience of delivering end to end. microsoft is NOT going to get them there, Oracle's strategy is nothing but marketing cheese, and Hyperion has just begun to coordinate with IBM to get that message through. When you think about it, the HYSL/IBM combination is downright scary.
the idea that commodity pricing of essbase is going to hip the market to Hyperion and secure their gorilla status is nonsense. there's one simple reason why. everybody is outgrowing the tools which made up a bulk of this market just 5 years ago. all the pilot lightships, holistic systems holos, tm1, and those object oriented guys whose name i can't even remember.. they're all being tossed. nobody in their right mind is going into the 21st century with 4GLs. SQL is the only 4GL which will survive, and maybe ADABASE.
regarding the big investment in MSOS, that's ultimately not going to be much bigger than the investments in front-end development. all the back-end can easily be re-routed to essbase. that's what sagent, informatica, and ibm visual warehouse do. but yeah if you are an MDX programmer, you're probably a waste of your company's money in the long term. that is iff they are interested in robust multi-user, read/write application environments. (that's why it's *analytic* platform, not just end-user query platform).
(i don't capitalize because i spent many years working with paper tape, 300 baud modems, ttys and ADM 320s. when RSTS came, i thanked the digital gods for lower case (rest the soul of DEC). )
singing like tina turner by: olap_genius (30/M/California) 03/20/99 03:57 pm Msg: 1841 of 12794
i'm not coming to grips with the relational world, i'm pitying its shortcomings. if i haven't said so, i jumped on the enterprise database bandwagon back in 1990 with db2 on mvs. at the time i was trying to reconcile the performance of static binds with the demands of end-user queries. this was in the days before odbc, back when the huge majority of fortune 500 IS departments wouldn't touch unix with a ten foot pole.
scubbing data is a huge problem because corporate IS is dilbertville and they don't have the guts (or budget) to hire people brainy enough to deal with the problems. y2k. need i go on?
at any rate, i'd like to comment on the msos partner sheet. i'm going to check out these datawatch folks. if a third of the plato partners were as bold as those guys, i'd worry. but you'll notice the paucity of analytic applications vendors. believe it or not, that's where the money is. it's only rational. comshare is a large, but clumsy player - they show potential here. think of olap this way - who makes more money in an erp installation, oracle or sap? sap of course, even though they are building everything on top of oracle. hysl will have lots of pull in getting organizations like lawson, jd edwards and baan to build on top of essbase. (and of course hysl will build on top of essbase). that will leave only 2 oem players, ms and hysl. even though comshare builds on top of oracle express, i doubt many others will.
-- hierarchy maintenance is a pain, but it obviously can be done in relational. the proof of this is in the fact that metadata for essbase is stored in relational tables for the functions of the hyperion integration server. but a bit of hands on with both tools (integration server & essbase application manager) clearly shows the app. manager to be more flexible. for example, a lot of bugs have just been worked out of HIS with regard to self-joined tables used for recursive parent child relationships.
no brainers in essbase include things like 1. period to date calculations in rolling period calendars. 2. handling balance aggregations like cumulative assets or inventory or headcounts 3. relative & absolute share calculations across single or multiple dimensions. (a simple example would be market share of los angeles among the cities in california, a more involved one would be product-market share of SUVs sold in california vs all truck-class products in north america)
'brainer' things in essbase include things that have relational types losing sleep & declaring impossible, and not only them but SAS jockeys as well. like adjusted headcounts across purchase categories in retail.
mo' thunder
by: olap_genius (30/M/California) 03/20/99 03:57 pm Msg: 1842 of 12794
'microsoft is not going to get them there' is just a simple acknowledgement of the fact that MS publishes a buttload of 'standards' and expects everybody to support them, but is not in the business of supporting them themselves. companies like unisys stay in business because they SERVICE corporate accounts. computer professionals don't go to college so they can run a bunch of wizards. anyway i simply ask you to consider the net impact of ActiveX on the internet and extrapolate from there. besides MS is going to be disemboweled by the justice department.
yes i know this sounds like FUD, but the real bottom line is this: (and i make the analogy to linux). if and when MSOS *performs* on the APB-1 benchmark, then i'll shutup. until then, there is no practical reason to consider them a first tier player, no matter how much they talk. lots of people buy inferior systems and/or don't care about performance. fine. i have no reason to believe that plato is first rate.
the second bottom line is that the plato *architecture* is not designed to support analytical applications. it's too narrowly conceived. the very idea that writeback is a patched on afterthought completely destroys their credibility for serious apps development. it's like aol's browser (anybody remember that?) there's no way it would support frames and java, because they didn't consider seriously the world beyond simple html.
-- hysl is going to retain high profit margins because of the simple fact that its channel business is done by sales people that it doesn't have to pay.
CRM on the horizon? by: olap_genius (30/M/California) 03/20/99 04:50 pm Msg: 1843 of 12794
CRM (customer relationship management) sounds like a decent area for growth. there's nothing in it whatsoever which leads me to believe that it shouldn't be built on an essbase platform. while i admit that customer *profitability* provides a more compelling business case, you really can't control what trendy buzzwords the industry is willing to swallow.
nabisco is the story hysl hypes these days w/ regard to customer profitability. they now have the infrastructure in place to re-incentivise their salesforce to sell more profitably to customers rather than just pushing product. the simple idea that you have your salesforce oriented more towards profit than revenue makes perfect sense. but only a few companies (evidently) see their way clear of the bureacracy and systems snafus which forestall such an orientation. in my experience the #1 problem has been "well salespeople can't mouse their way out of a paperbag", which is a shorthand way of saying current relational query front-ends are still too complicated and unusable. hysl doesn't have such problems.
if it werent' for the people at ACTA, i would tend to believe the hype about SAP BW. but i don't believe SAP is anywhere near getting OLAP. again, lots of people can scrub data and build warehouses. but delivering added value tied to the *business planning cycle* is what Hyperion knows better than anyone else. If that were not the case, there's no way they would own 85% of the F500 financial consolidations market.
HYSL is smart to finally start taking back consulting from small-time hacks out there. it's time for coordinated vision. any integration partner who doesn't understand how to build CRM applications on essbase, or doesn't have the credibility with customers to let them try doesn't deserve to be a partner.
QWST + HYSL =? by: olap_genius (30/M/California) 03/20/99 04:56 pm Msg: 1844 of 12794
i'm still trying to make sense of something i read this week about Qwest Communications making a partnership with HYSL. Does anybody remember Dillon saying anything about an analytical portal. QWST has been one of my better picks, and i'm tending to get (as a shareholder) out of software and into telecom. i'm just scratching my head about the synergies.
snap crackle pop, but no thunder by: olap_genius (30/M/California) 03/23/99 05:27 pm Msg: 1863 of 12794
i stole this from the olap newsgroup. it's just what i expected to see. ---
From: "Scott Wheeler" <[email protected]> Subject: Balances over time in MS OLAP Date: Wed, 17 Mar 1999 14:46:32 -0600 Organization: http://extra.newsguy.com
We're having trouble implementing a cube in Microsoft SQL 7.0 OLAP Services. We can't figure out how to accomplish a rollup of an inventory balance over time.
Our inventory levels are counted on a montly basis, we have a cube dimension for inventory level (aggregation method = sum) with an entry for each store for each month. The problem is that when we summarize the months into quarters, the quartly inventory level is the sum of the monthly inventories rather than the inventory at the end of the month.
Example for Store A: January inventory = 10, February inventory = 15, March inventory = 13. This results we're getting is a Q1 inventory of 38, we need the Q1 inventory to be 13.
This feature is available in our current OLAP tool (Cognos) and is refered to as "time-based rollup".
Does anyone know how to accomplish this in Microsoft SQL 7.0 OLAP Services?
Scott Wheeler Senior Consultant
meaninglessness by: olap_genius (30/M/California) 03/25/99 01:18 am Msg: 1875 of 12794
i agree with the poster who mentioned that essbase will be a logical upgrade path for plato. that is the common wisdom out there. MS has a great deal to prove on NT *first*, before the bigger systems integrators will even consider plato.
so talk about market activity is valid. in fact it hurts microsoft because if and when MS OLAP scales, there will always be one simple fact, which is that the biggest customers of olap, current ERP and DW users, are by and large NOT standardized on SQL Server. i think that the likelihood is very slim that they will change, over to SQL Server just to save some dollars on plato. the chance is nil if they are already on UNIX.
speaking of market activity, it has come to my attention that a very large manufacturer was gung ho on standardizing on plato. but they already had an essbase customer profitability application inhouse. so what happened? hyperion challenged the microsoftians to replicate the functionality of the essbase app, and all bill gates' horses and men couldn't do it.
there is a very very good argument in support of essbase regarding cost of ownership, which is the point someone made very clear. you can hack around in MDX and the new database interfaces of VB or you can develop quickly in essbase.
my understanding is that it's on the plate to open up portions of the calculation engine to 3rd party functions, and i'm pretty sure that some new native functions (some, long awaited) are in the next release. --
in other news, i have heard glowing reports of BMC's Patrol product's Essbase systems management and monitoring tools.
i also talked to some EDS people today. internally, they are very hyped on essbase.
nothing but good news.
hate the stock, but buy it cheap by: olap_genius (30/M/California) 03/25/99 01:30 am Msg: 1876 of 12794
acta sucks? that's a new one. i'll watch them closer. certainly informatica and sagent have been pumping up their SAP compatibility. it could get ugly.
call center management is really a no-brainer to implement in essbase. i've seen it done in pilot lightship 5 years ago - even before lightship server. i used to wonder why arbor didn't do that a long time ago. there's probably not much money in it.
one more thing by: olap_genius (30/M/California) 03/25/99 01:51 am Msg: 1877 of 12794
regarding buying outdated applications with no upgrade path.
that really depends on how you look at it. for multinationals with multiple gls to consolidate, there really is very little that enterprise doesn't handle. where do you go after you have 85% market share? with silly little microsoftian upgrades?
pillar has been repriced to go downmarket, so there's more revenue to come. upmarket of that has always been essbase, the essbased based competition and comshare budget plus. as i see it, there's only a few reasons to grow through the cieling of pillar, and they are all very well known to pillar customers. if you think hyperion is stupid enough to lose those customers, then there's ample reason to complain.
fixation by: olap_genius (30/M/California) 03/25/99 11:09 pm Msg: 1889 of 12794
oh. i'm just having fun agreeing with gates. ms faces serious competition in a large number of markets in which there is no guarantee that it will dominate, lead or really even compete.
besides, i think they are going to get broken up.
also i didn't really mean 'own' the high end of olap or analytical apps. caught up in my own rhetorical excess. let me stick a number on it. ms will have 20% of high end w/ very little headroom.
a broken up microsoft would provide a much bigger threat, somewhere around 2003, to companies like hyperion, peoplesoft, i2 technologies, new era of networks, et al. right now they're too fixated on the desktop and wizards, satellites and mass markets to have the kind of focus it takes to build and market enterprise software. i think that's stating the obvious.
au contraire, more web & wan than anyone by: olap_genius (30/M/California) 03/26/99 07:54 pm Msg: 1895 of 12794
essbase works with tcp/ip or named pipes so there's your basic wan and lan.
the hyperion.com site has a live web demo. so there's your extranet. the product is essbase web gateway, cgi based html.
wired for java is 100% sun certified. the applet works out of the box with no programming.
3rd parties include painted word which has wrapped the essbase api in java, and has made a custom jdk specifically for essbase. plus there's partner alphablox which makes a java component-based application development environment.
furthermore, there's a thing called essbase objects which is in full microsoft activex compliance throughout their object model which is sold by hyperion.
and there's also a web based report-writer / smart agent delivery system called spiderman. jeez. maybe i should work for these guys. i could probably sell the hell out of it. hmmm..
apple computer by: olap_genius (30/M/California) 04/10/99 06:36 pm Msg: 1975 of 12794
1. CEO sets no vision for the company. 2. CEO blames outside issues for poor performance. 3. Key executives sell stock and leave the company. 4. Key engineers leave the company. 5. Wall Street down grades the campany often.
bidding war by: olap_genius (30/M/California) 04/10/99 07:00 pm Msg: 1976 of 12794
hmmm. who could buy hyperion? i don't think there's any doubt that if insiders were willing to roll over on hyperion stock, the price would have to be at least 50. ibm would be the logical choice, although tivoli folks have war stories.
hyperion under sap would be a world beater, but i could see lots of blood in terms of company culture.
The Real World by: olap_genius (30/M/California) 04/21/99 11:06 am Msg: 2037 of 12794
in the real world, 100GB single cube applications don't exist. 100GB applications don't very often exist but when they do, essbase handles it. but let me help you understand two things:
1. essbase scales infinitely because IT IS A MULTIDIMENSIONAL DATABASE MANAGEMENT SYSTEM. it scales the same way relational databases scale. if a table is too big, you split it into two tables. if an essbase database is too big, you split it into two. essbase is not 'a cube'. it's a SYSTEM.
2. sears discounts every item on display in every store in america using essbase TODAY.
money for nothing, express for free by: olap_genius (30/M/California) 04/21/99 11:23 am Msg: 2038 of 12794
express bundled with 8i. that makes sense, but it doesn't mean that hysl will not become the gorilla. what we know about express is that it is shelfware. now it will just become shelfware on more shelves.
obviously oracle has decided to give express away, but that won't make it a better software package to work with. it only points to the probability that it won't be further developed. there is no way that the express people at oracle are going to outgun the 8i people for resources. so that suggests that express is
going to go away and that features will gradually be absorbed into 8i.
that puts oracle at a short to mid term disadvantage as a platform for analytic applications while the world tries to figure out what's going on. in the long term it can work to their advantage if serious
multidimensional capability is built into 8i (how?).
none of that changes the fact that as far as i know, only comshare has built a packaged analytic application on top of express: budget plus. to put it simply, budget plus does NOT compete well against
the essbase based packages like peoplesoft budgets, walker horizon, and the lawson products. hyperion is already the gorilla in budgeting software which is the number one analytic application in the market. neither 100GB capability or bundling from either express or plato is going to change the years of advantage and sales experience of hyperion's leverage in this real world market.
name one
by: olap_genius (30/M/California)
04/21/99 07:24 pm Msg: 2046 of 12844
13 dimension model that makes business sense. i dare ya. better yet, name one 13 dimensional star schema in production.
Posted as a reply to: Msg 2045 by apps_dev
schmolap sucker..
by: olap_genius (30/M/California)
04/21/99 07:29 pm Msg: 2047 of 12844
"It just can not compete with DOLAP, HOLAP, and ROLAP. "
spoken like a true zealot. but what does that mean? you have no idea what it means. the very idea of a 20GB DOLAP destroys your credibility.
Posted as a reply to: Msg 2045 by apps_dev
Why Buy Sapling?
by: olap_genius (30/M/California)
04/26/99 11:49 am Msg: 2095 of 12844
i'm guessing buying wasn't a good idea if market reaction is likely to follow the kind of sniping in this room. but the agreement says reselling and co- development, that's about as much product control as you get.
looking good...
Posted as a reply to: Msg 2094 by MemysselfandI
Comments