Teradata's New DW Appliance Begets Wrath from Netezza
Data warehousing -- remember when that was a hot buzzphrase about seven years ago? -- remains a growing sector of IT data storage, and the generally acknowledged market leader continues to be Teradata. For those of you who might not be clear on the definition: A data warehouse is a repository of an organization's electronically stored data designed to facilitate reporting and analysis. An efficient data warehouse is able to retrieve and analyze data, to extract, transform and load data, and to manage dictionary data -- so that it can be used for business intelligence. Traditionally, data warehouses are big, hulking systems used for large enterprises. Not so anymore. Like the newer SAN and NAS storage systems, they are being slimmed down for smaller businesses and becoming easier to use. Prices are coming down, too, for these more compact systems. Newswise, the Miamisburg, Ohio-based corporation yesterday introduced a scalable new entry-level data warehouse in appliance form, one that can grow into an enterprise-level setup, using the Teradata 12.0 database engine. This must be a significant new entry into the DW playing field, because a Teradata competitor -- Netezza -- was eager to speak out about the development. "Teradata has finally recognized the impact of data warehouse appliances, entering this market segment eight years after Netezza invented it and at the back of the line of a handful of other vendors," Netezza President and COO Jim Baum told The Station via email. "From our position as the market leader, Netezza views this as validation of the current data warehouse appliance market success, and more importantly, a concession that the present and the future of the data warehouse market lies in the expanding role of appliances." Let the DW debate begin. What's your take on this: Is a data warehousing appliance something you might be interested in acquiring? |
Comments (3)
Obviously Netezza is in pain.
Actually, Teradata invented the 2nd database appliance -- the DBC/1012 back in 1984 -- which were specific purpose processing nodes. There were separate processor boards for different database functions (IFP, COP, AMP). Sounds like Netezza, eh? AMP=SPU. Later, Teradata bought Britton Lee (called Sharebase) which was the first real DW "appliance". In the 1990s, Teradata bailed out on custom hardware going to a standard "server" design excepting their BYnet.
Actually, I think the "appliance" word is all hype. REAL tech appliances emerged in networking, search, and security where a 1U or 2U rack mount device was plug in and go (aka Network Appliances Company. Contrary to Netezza's claims, you cannot plug in their product and have business users running queries in an hour. You still need to do a lot of data modeling, ETL, security set up, and so on. So Netezza's claim to be an appliance should be put in the "great marketing" category but its hardly an appliance the way 95% of the IT industry looks at appliances. The only reason the term stuck was customers wanted a RedBrick-like hot box for data marts.
Netezza sells data marts and does it with custom hardware. The 20-30 or so custom hardware companies that attacked Teradata and IBM over the last 20 years all died. Netezza is the only one to survive more than 2 years. Go find nCube, Convex, KSR, SGI COMA, Sequent, and all the others.
I didn't see Teradata claiming the 2500 is an appliance or a data mart. So why is Netezza in such pain? Probably because customers can now buy a top end data warehouse for Netezza prices. That's gotta hurt.
Validate is Jim Baum's self-serving word for it. Still good Marketing from Netezza.
Posted by Dan the Man | April 23, 2008 6:34 PM
Thanks, Dan, for the history lesson and for your perspective on this topic. That puts the ball back in Netezza's court. Jit or Phil, are you around? Might you want to add your response here? Thanks in advance.
--The Station
Posted by Chris P | April 24, 2008 12:13 AM
Having slipped Netezza into some significant research sites for scientific DW I can personally attest to the simplicity and performance claims that Netezza makes.
If there is an operational data warehouse with data models, etc., all you need to do is load the data model, load the data at 500GB/hour, point your BI/Analysis tools at the NPS and you have a DW that is orders of magnitude faster. The bigger the data set the faster Netezza will be. FPGA's next to the disk/storage media in a massively parallel architecture are a wonderful thing.
If you want to build a dw from scratch, that is a completely different story for any vendor/user. To shorten that process you need to understand the data that you have first - that requires a data profiling tool. Understanding the data first will prevent significant pain. Once your understand what data you have you can then model the data, clean it and ETL it, or ELT to Netezza and clean it (this works really well on Netezza).
As a middleman/systems integrator we could choose to sell any DW hardware/software solution. Especially if we wanted to gouge customers with massive DBA service charges to keep the DB tuned. You do not need to tune Netezza because it does not have indexes!
Real world benchmark 1 - Teradata ($2M) Netezza ($500K) - 800GB data - Netezza 20 to 30 times faster for all queries but one. That simple query was a dead heat. If a Netezza system with twice as many Snippet Processing Units was used the benchmark results would have been twice as fast. Performance on a Netezza box in linear.
Real world Benchmark 2 - Netezza versus Oracle/MySQL on IBM Cluster($4M) Netezza ($1M) - 5TB Data - Netezza on average 350X faster. It took one person five days to setup the Netezza system, load the data model and load the data.
As far as Netezza being in pain - just look at their financial performance - its public - sustained and controlled growth, quarter over quarter, year over year.
Posted by Big Data Daddy | April 26, 2008 11:56 AM