SQL Server to Teradata

  • Ours is a dataware house environment currently using sql server and planing to migrate to Teradata as it is very robust and can handle our environment very effeciently.

    I would like to know if any DBA's here expereinced such migration and what would be a DBA's role in the process of migration. please share your experences.

    Thanks

  • Mike Levan (3/2/2010)


    Ours is a dataware house environment currently using sql server and planing to migrate to Teradata as it is very robust and can handle our environment very effeciently.

    I would like to know if any DBA's here expereinced such migration and what would be a DBA's role in the process of migration. please share your experences.

    Thanks

    Never got a chance, but as a DBA you'll be involved from begining to end. There is no Scope of Work defined for a DBA in such a situation.

    cross your fingers and get into this.... you'll learn a lot.

    Regards,
    Sarabpreet Singh 😎
    Sarabpreet.com
    SQLChamp.com
    Twitter: @Sarab_SQLGeek

  • Our company hired contractors to create our Teradata data warehouse and also create all the ETLs from the SQL Servers. I don't know if you are familiar with Teradata, but it was an eye opening experience for me. The Teradata import / export tools are very limited and there is no method to schedule jobs other than using Task Manager.

    The database structure is also very different. The initial size of the database is created at installation time (done by NCR vendors) and subsequent databases are created by taking space from the main DBC database and allocating it to the new database. Interesting. There ae no schemas, but to reference objects you just specify the ownership like normal (database.object). Tereadata is normally installed on UNIX and it seems similar to Oracle.

    All the ETLs were written in Python because it was the contractors code of choice--another unknown to me. They also used a data movement tool they created to import and massage the data.

    I was greatly surprised to learn that a number of the management tools required a separate database to collect data in order to check the system health. I was also informed that there would be little to no DBA maintenance. How untrue. There is no rebuilding of indexes, but statistics need to be updated regularly. There is no built-in GUI or scheduler to create such tasks so Python code was used to update the stats and also to back up the databases. Too much manual stuff and coding in my opinion.

    I was also told that the system is so fast that there would be no locking issues. Untrue again. I hope you like to read, because their documentation could choke a horse.

    This is just a snippet of my data warehouse experience with Teradata. Good luck and have fun 🙂

  • Well, I didn't want to write this much, but some things need correcting.

    I have been a Teradata DBA for over 2 years. The duties are very different than with SQL Server and Oracle. Before that, I worked on a conversion from Oracle to Teradata. Since learning about SQL Server and a little bit about Oracle, I have been amazed at the amount of work required by a DBA compared to my duties on Teradata. Once a Teradata system is set up and the database objects have been optimized, it requires much less hands-on interaction from a DBA. Even setting up a new database or user is a breeze compared to SQL Server.

    Having said that, there are some caveats:

    1. A little understanding of the system hardware and software goes a long way. Teradata architecture utilizes Massively Parallel Processing - much different than SQL Server or other RDBMSs. This goes a long way toward optimizing query performance and and space management. The system is split into AMPs (Access Module Processors): several virtual processors working in parallel on their own table rows with their own dedicated disk space. Similar to a clustered index in SQL Server but not the same, Teradata's Primary Index (mandatory on every table) determines how the data is physically distributed across each of its dedicated disk arrays; however, Teradata doesn't care how the data is ordered. It just rams each table row onto disk in whatever order it gets it. The ordering is done upon retrieval. This gives Teradata its speed.

    2. Setting up the tables to be stored with the best primary index, and having the queries written to take advantage of that is probably 80% of ensuring promised performance gains over other RDBMS's. Poor design can seriously impact query performance. I cannot stress this point enough.

    3. Unfortunately, like the previous contributor expressed, Teradata does not have many user-friendly tools or wizards, but its database management tools have improved considerably in the few short years since I started managing them. If you are on a Teradata 12 or 13 platform, these are the latest and greatest versions and you won't know how lucky you are. Teradata doesn't care about Primary Keys or Foreign Keys (While you can define those constraints if you want to, performance will not change).

    4. Collecting statistics is indeed important, but they must be collected at the proper time (generally after a table load) or they will not improve performance. Teradata gives an explain plan on each query to indicate if and how the query can be optimized and suggests columns and indexes on which statistics should be collected or updated.

    5. Using Workload Manager (TDWM) and other management tools, setting up users under different Priority Groups will allow the Teradata system to distribute the load of CPU and disk I/O in a way that allows the most important queries to run faster. This was the main are of my duties as a DBA.

    6. Yes, the toolset offered by Teradata pales in comparison to Oracle and Microsoft, but performance on a healthy Teradata database system is miles ahead of the equivalent data on SQL Server and Oracle. Besides, there are many third-party tools out there with wizards, and schedulers and GUIs. Teradata concentrates on performance and it delivers.

    Simply, Teradata is the fastest, most robust, most scalable DBMS on the market. Period. Check Gartner Group's website and you will see it is the most advanced technology in the business. I have seen a Teradata query bring back GIGABYTES of data in a few seconds. I know it works.

    In closing I would say "Your job just got easier" but there will be a learning period to ramp up and take advantage of that new knowledge. You can go to teradata.com and order a working version of the latest database system to learn on, that comes with documentation on the architecture and tools.

    And, no, I'm not a Teradata salesman. The facts speak for themselves. The largest, fastest databases in the world are Teradata systems. E-bay, AT&T and Wal-Mart all use Teradata, as well as several airlines and many other Fortune 500 companies.

    I hope this helps.

  • I will have to strongly disagree with the previous post. Teradata is complete garbage, it does not lead in anything and was not even one of the 20 mentioned in Gartners Review. I have used it its not fast at querying or opening a database, get used to waiting 5 minutes for a table to open just so you can see the columns. I queried over a million records with SQL Server 2008 R2 on my local server in less than a second, it took teradata on the network about 15 minutes the 5th time. No way of looking at an execution plan, the visual look is what you may expect to have seen in 1983. I would use Access or Excel and get more accomplished. If my argument does not sound convincing get on the job boards and count the number of companies looking for Teradata VS SQL Server 2008 R2 or SQL Server 2012 it only trails Oracle in RDBMS but leads in growth, and it leads in Business Intelligence features. I am sure by know you know the answer regarding Teradata's BI tools, non existent so after spending more money on just the RDBMS you will fork out more for an ETL tool, then a Reporting Tool and probably another reporting tool since this seems to be a trend, and then based on whats left maybe a water downed OLAP tool which by the way is MS SQL Server's most robust product it has been the number 1 OLAP tool for 9 plus years which is the direction of this profession with data growth. I work for a great company that uses both but after a couple of weeks I made it clear I am not going to develop with Teradata when SQL Server is available, big companies make a lot of money and yet spend on crap like this more often than I would have thought. FYI I would not be suprised if Teradata is even around in 3-5 years, or upgrades thier product. That is why rumors of being acquired or moving to a different data management profile have surfaced.

  • Couldn't agree more. $15MM server runs slower that using FoxPro with flat text files. Seriously.

    What used to take an hour to develop, 100 lines of easy to maintain code, and 5 seconds to run on a $150k SQL Server now takes 4 hours to develop, takes 500 lines of multiple redundant nested sql statements and runs in 25 minutes on a $15MM server.

    Teradata promises everything, delivers nothing, and when questioned, the answer is, "spend more money".

    We spent more money and it fixed nothing.

Viewing 6 posts - 1 through 5 (of 5 total)

You must be logged in to reply to this topic. Login to reply