Re: Opinions on Informix

Tim - I think you have either been mislead or are possibly being slightly obtuse. > 1. The Oracle docs never dive deep into anything concrete, but rather treat > the reader as if they are nothing more than a sophisticated end-user. The book you are quoting is not a 'Oracle Doc'. It may be from Oracle Press, but it is not the official Oracle product documentation. It also sounds many years out of date. Found in a bargain bin, perhaps ? > 2. The Oracle data warehousing paradigm is nothing more than tying several > servers together to a shared disk ( probably NFS mounted as the docs don't > tell you). They emphasize raw disk usage but most of the docs talk about > using regular cooked files. Nothing specific. This is so wrong as to be laughable. Oracle provides many, many features for data warehousing. Lets start, tell me when you have had enough Parallel queries Parallel updates Parallel inserts Parallel deletes Parallel data load Parallel index creation Parallel statistic analysis Parallel table scans Parallel index scans Range partitioning Hash partitioning Composite partitioning Hash joins Parallel Bitmapped Star Joins Partitionwise joins Partition elimination Native CUBE operations Native ROLLUP operations Native Data Sampling Bitmapped indexes Functional indexes Reverse Key indexes Join indexes Materialized Views Summary Management Metadata Dimensions and Hierarchies Automatic Query Rewrite Transportable tablespaces The list goes on, and this only include the DW specific stuff - then there's all the other database features, plus Express, plus Discoverer, etc etc > 3. A really important note about Oracle Data Warehousing, from their own docs: > > Let's say that in Informix XPS, you set up four servers, and fragmented a > customer table across these four servers. You might have customer.1, customer.2 > customer.3, customer.4 dbspaces. You might fragment by hybrid on customer_no in > customer.1, customer.2, customer.3, customer.4. > > In Informix, on XPS, you can "select * from customer" and a result set comes back. > The user has no idea that the data is residing on 4 SEPARATE disk subsystems to > take advantage of real multiserver parallelism, and gets their result set back > nice and fast, in ONE set. If you didn't know any better, you'd think it's just > one server, and go about your business. > > In Oracle, the syntax would be as follows: > > select * from customer.1 > union > select * from customer.2 > union > select * from customer.3 > union > select * from customer.4> > And the result set comes back in pieces! > > Now, I have to ask myself, as just a dumbass end-user, why do I have to do all this > union stuff? Can't I just get the result back in one piece? Wrong again - this is the old way of doing this before we offered partitioning, or what you would call fragmentation. It's a technique called partitioned union views, and was a 'poor mans' version of partitioning before partitioning was available. Of course partitioned tables in Oracle are totally transparent to the application. > There is hardly any information on Oracle PQO their "equivalent" of XPS, so my above > observations are only from their book. ( PQO =Parallel Query Option ) PQO or Parallel Query Option is no longer an option (and hasn't been for many, many years). Parallelism is built into the product. Also, PQO was NEVER the equivalent of XPS. You are confused with the Oracle Parallel Server option (OPS), which is the distributed lock manager that is used when Oracle is installed on two or more nodes in a hardware cluster - for both Unix and NT. Large data warehouses use Oracle Parallel Server to increase scalability and the number of physical disks that can be addressed, but many, many Oracle Data Warehouses do not use the Oracle Parallel Server option, running instead on single SMP environments. And, unlike XPS, Oracle Parallel Server can also be used to increase OLTP scalability by allowing users to be spread workload across multiple nodes in a cluster > 4. Anyone can go buy the Oracle books I bought, and read up on the beast, and see what's > really available. The big question is, if *I* can do this, and understand the differences > SIMPLY FROM THE DOCS <<HELLO>> why can't Informix Marketing????? <<HELLO>>!. Breeeep. Wrong. Try again. You haven't determined what is really available, and you haven't understood the differences. Next caller please. > Informix Marketing, pay attention, you need to SELL your products, and rebut the competition. > You need to illustrate in clear terms simple differences like this. XPS should be selling > big if you simply compete. After all, data warehousing is one of your focused markets, not > just i.Sell. XPS should be on Linux, THE next great arena for clustered systems. Focus - interesting way of spelling niche. And why just Linux, what's wrong with NT ? Regards

Mark Townsend wrote: > > Tim - I think you have either been mislead or are possibly being > slightly obtuse. > Thanks Mark, it's probably a little bit of both. I honestly do want to learn more about Oracle. It actually looks like a lot of fun, using Oracle. After using Informix for years, the market for Oracle is big, and the work your products create looks like I'll find a rich environment of data base people who are real lightweights should I choose to work with Oracle. > > 1. The Oracle docs never dive deep into anything concrete, but rather treat > > the reader as if they are nothing more than a sophisticated end-user. > > The book you are quoting is not a 'Oracle Doc'. It may be from Oracle > Press, but it is not the official Oracle product documentation. It also > sounds many years out of date. Found in a bargain bin, perhaps ? > On the front of the books the words " The Authorized Oracle Press Editions". Books and their release year: Oracle Data Warehousing: 1997 Oracle 8 Architecture: 1998 Chapter 12: Oracle's Parallel Processing Option, page 322. Shows shared/NFS disks. Note, NFS mounting isn't confirmed, I'm just guessing. And I would imagine in this case, it would be recommended to use cooked files unless raw disks were set up with NFS--seems kinda nasty. Interesting that Oracle must continue to use SMP parallelism, while the Informix XPS engine is specifically designed for multi-server parallelism in an MPP environment. I cannot imagine that since the Oracle disk is shared across several servers that query processing would be distributed among instances, and in looking at these authorized docs, it does not indicate that multi-instances share the work of a query. As I understand it this would be a major flaw in Oracle, and several years behind in architectural design. Thus there would be an increased need for indexing and other "typical" SMP issues, but nothing in the way of MPP design. > > 2. The Oracle data warehousing paradigm is nothing more than tying several > > servers together to a shared disk ( probably NFS mounted as the docs don't > > tell you). They emphasize raw disk usage but most of the docs talk about > > using regular cooked files. Nothing specific. > > This is so wrong as to be laughable. Oracle provides many, many features > for data warehousing. Lets start, tell me when you have had enough > You're not showing me where I'm wrong here, please, offer me the proof. I just pointed this out from authorized Oracle press documentation. Anyway... Informix has all the following only in a real MPP environment, not just SMP: (As the docs show, the following must happen on ONE physical server in Oracle) > Parallel queries > Parallel updates > Parallel inserts > Parallel deletes > Parallel data load > Parallel index creation > Parallel statistic analysis > Parallel table scans > Parallel index scans > > Range partitioning > Hash partitioning > Composite partitioning > > Hash joins > Parallel Bitmapped Star Joins > Partitionwise joins > Partition elimination > > Native CUBE operations > Native ROLLUP operations > Native Data Sampling > > Bitmapped indexes > Functional indexes > Reverse Key indexes > Join indexes > Yes, in Oracle you need the indexes more because of the older single server SMP architecture. > Materialized Views > Summary Management > Metadata Dimensions and Hierarchies > Automatic Query Rewrite > > Transportable tablespaces > > The list goes on, and this only include the DW specific stuff - then > there's all the other database features, plus Express, plus Discoverer, > etc etc > It's a great marketing campaign you have, and I salute you for having old stuff repackaged. <...snipped...> > > There is hardly any information on Oracle PQO their "equivalent" of XPS, so my above > > observations are only from their book. ( PQO =Parallel Query Option ) > > PQO or Parallel Query Option is no longer an option (and hasn't been for > many, many years). Parallelism is built into the product. > Yes, SMP on ONE server, not MPP. It's still single-server-multi-CPU parallelism, not colocated parallelism. Until you actually see the difference I would be talking to somebody who is inexperienced as to understanding the difference. > Also, PQO was NEVER the equivalent of XPS. You are confused with the > Oracle Parallel Server option (OPS), which is the distributed lock > manager that is used when Oracle is installed on two or more nodes in a > hardware cluster - for both Unix and NT. Large data warehouses use > Oracle Parallel Server to increase scalability and the number of > physical disks that can be addressed, but many, many Oracle Data > Warehouses do not use the Oracle Parallel Server option, running instead > on single SMP environments. And, unlike XPS, Oracle Parallel Server can > also be used to increase OLTP scalability by allowing users to be spread > workload across multiple nodes in a cluster > XPS can be used in OLTP, but it's actually good that XPS is specialized for DSS. Using your older architecture engines which were designed for OLTP in a DSS environment is simply taking advantage of people who don't know any better. Again, great marketing, and I salute you. :-) > > 4. Anyone can go buy the Oracle books I bought, and read up on the beast, and see what's > > really available. The big question is, if *I* can do this, and understand the differences > > SIMPLY FROM THE DOCS <<HELLO>> why can't Informix Marketing????? <<HELLO>>!. > > Breeeep. Wrong. Try again. You haven't determined what is really > available, and you haven't understood the differences. Next caller > please. > Uh, Mark see above. As to understanding the differences, I really do. It's interesting that your posting has not pointed to one single fact to show that I am wrong. <...snip...> > Focus - interesting way of spelling niche. And why just Linux, what's > wrong with NT ? > > Regards Mark, If you don't know the difference why Linux is better than NT, you obviously are a lightweight. And I like that. I could have fun with Oracle. It might just be time...and it certainly would be fun. :-) Thanks, Tim - -- --- Tim Schaefer ---- tschaefe@mindspring.com --- http://www.inxutil.com -- -

PS Mark, I didn't engage in the discussion HERE only to have to chat about it off-line. If you want to chat about it, post it here for the rest of the world to read as well. Don't send me off-lines to talk about this. Post here what you sent me off-line and I'll respond. Thanks, Tim -- - -- --- Tim Schaefer ---- tschaefe@mindspring.com --- http://www.inxutil.com -- -