So normally I will not go controvercial on the blog. But here I go and I hope to do this to start dialogue that I can not seem to get the answers on. As a follow-up I will call both vendors (Teradata and Netezza) and interview them on their point of view. So here I go and remember, this is supposed to be the beginning and not the answer as some of these are my opinions which I have been trying to get validated.
Company / Product Age:
– Netezza: > 6 years (privately held)
– Teradata: >25 years (publically held through NCR Corp)
– Teradata: > 800 customers & 1200 implementations
– Netezza: > 65 customers & 130 systems
Revenue / Profitable:
– Netezza: Not known as private
– Teradata: $1,480 million in Revenue, $309 million in operating income (2005)
What Netezza Says:
– Terascale – DBMS software, hardware, high speed storage close together
– Assymetric Massively Parallel ProcessingTM architecture for complex query processing
– Intelligent Query StreamingTM design
– Integrated Appliance
– Industry Standard Interfaces
– Fully compatible with Marker-leading BI tools, applications and infrastructure – not certified that I could see
What Teradata Says:
Teradata provides integrated, optimized and extensible enterprise data warehousing technology and services for a single application-neutral repository of your current and historical data. It’s a combination that allows you to analyze business operations and drive better, faster decisions. Featuring high-performance parallel database technology, a full suite of data access and management tools, robust data mining capabilities and the industry’s most experienced data warehousing consultants, Teradata provides you with a complete view of your business.
— End of What they say —
– Teradata: version 2 release 6, but they have manu applications that they have that run on their platform
– Netezza: OS: Red Hat Linux server 4.0, SQL ODBC, SQL-92 compliant with SQL-99 extensions
Load (as no benchmarks all anecdotal observations based on customer comment):
– Netezza: very fast
– Teradata: more dependent on the ETL, but can be fast
Manipulations (our observations):
– Teradata: Mature tool set
– Netezza: We have found load and unload to be quick, but have found the tools to manipulate and program the system difficult and early stage. We have found undocumented features and issues with load error handling.
Not able to speak to this on a mixed load, but anecdotally we have found that on certain queries Netezza can be faster. Netezza uses zone maps and this means data needs to be in sequence. This will remain anecdotal as I have not seen any true benchmarks available
Approach (our experience):
– Netezza – No
– Teradata – Yes – very strong
Certified with other vendors:
– Teradata – Yes
– Netezza – Interoperable but not necessarily certified
Availability and Recovery (what they say):
Certified and well integrated with many different options
– redundency across the whole system
– more complex solutions in general
– Some identified issues
– PC class ATA hard drives
– RAID – not sure
– Interesting redundency, but how often and what is load hit and who swaps
– Archive recovery is off a Postgres ascii file, not integrated
Value / Cost (our experience):
I will not talk about value as that is in the mind of the beholder, but here are some numbers in Canadian on Cost:
– Netezza – $200 to $300 per Gb
– Teradata – $400 to $700 per Gb
In our experience and in our model that we have created we find that most of the platforms end up actually being about the same cost if you look at total cost of ownership – True TCO.
– Teradata – generally known as the easiest to support with dbas
– Netezza – needs more dba’s in general
Final thoughts and Summary:
In no way do I want anyone to think that we are one way or the other, we have clients with both platforms, both are happy with certain instances of the platforms and both have issues as you can imagine. Project X knows Teradata better than Netezza, but we are working on that as we speak.
So come on weigh in with your thoughts and let’s start a healthy dialogue. I am going to allow free flow of comments and will update the main body of the post as we go to sift through and keep the post as accurate as possible. I will be using my living room rule for moderation of comments and will try and base as much as I can on facts, so if I am wrong let me know and I will fix it.
Update: It was noted that my tense (first person…) was not consistent. I will try and fix that.
Netezza Twinfin is much better/faster. Cheaper than Teradata 55xx
Never heard of netezza before, anyhow nice comparison
Netezza can’t reach upto Teradata speed.
RE: ‘Netezza sucks….’
Sounds like someone decided to skip the planning and instead distribute everything on random keys.
Netezza sucks.. All of their marketing pitches are bunch of BS. The time you try to join two large tables, Netezza is dead. You need to build materialized views for every thing to get the performance you need.
I know that when we worked with some of the Netezza utilities 3 years ago there wasn’t anything there for SOA.
Can’t speak for today as I would hope that the tools have evolved.
Not sure though. There are some great middleware products that are ODBC compliant that should be able to work with it though.
Some say that the necessary software (middleware, etc,)to accomplish SOA with the Netezza product is not there as yet. True?
We implemented Netezza recently in our shop and I don’t see any advantages with Netezza over Teradata.
Netezza is still under development. Teradata is more mature product.
Netezza does not follow some of the RDBMS rules.
How ever, I am sure Netezza will catch up with Teradata in couple of years.
This is the seconed comment on this thread and would be willing to talk with people if there are good referencable examples. I have just not been able to find any with regard to DATAllegro. If you have any suggested cases let me know.
Good article but I really would suggest you look at DATAllegro, the performance and price is unbelievable compared to both Netezza and Teradata. Their customer wins are notable too now.
I think one cost that you might want to consider is the relative ease of maintaining a Netezza box. For the most part, you are looking at a power and network connection, both very standard.
From my understanding, this a lot more challenging on the Teradata side, requiring a lot more resources and effort to keep the box humming.
NoName: Thanks for mentioning DATAllegro, I have a colleague who has been talking about them with me for about a year. Not talked with anyone who has implemented it yet. Maybe a good future post there.
James: The Frontier example is great, I will try and get a hold of someone there to talk with them about it. We have a client who is very interested in open source DW so we are just learning about it’s pluses and minuses. Love more of your thoughts on it. Have you seen it much?
any sign of BizGres yet? the Frontier Airline example *really* solid.
Must Look at DATAlllegro. Real difference in price and performance.