Data Synchronization

UDA
Gut
Teradata Employee

Data Synchronization

Currently Teradata offers two products that address data synchronization directly, Teradata Replication Services (TRS) and Data Mover (TDM). TRS is a highly available option that deploys a 2PC when updates occur on tables with TRS defined on them. Data Mover is a Table copy or table shipping utility that uses Teradata utilities such as fastexport, bteq and ARC to do table copy, full and partial tables can be copied this is an ansynchronous operation, therefore the secondary system if more vulnerable to getting out of synch when certain primaryt system faluresw occur. SRDF/SRDA technology is not a viable option with the Teradata DBMS. In any design, I will content that a sound table shipping foundation must be in place to begin with for all tables. There will always some cases where the table need to be restored from either copy of the data. The table shipping options that companies such as ebay and Meijer have put in place are sound strategies. What other customers have put in a general purpose table shipping function between multiple systems. (it already needs to be there from PROD to DEV/TEST for debugging purposes).  Besides making the method of replication invisible and optimized based on the customers preference, what do customers need in a data synchronization tool?

1 REPLY

Re: Data Synchronization

Gut,

I think it may depend on what you are using the copy of he data for. Are you updating a contingency site, which is lights out, and ready to take over after a short delay. Are you updating a different warehouse with data from another, ie where the databases are not physically the same.

Our systems traditionally use SRDF or SRDF-A for repliacation to another site, but they are DB2 systems, so any slight inconistantcy between logs and data can be fixed on database start.

I would be interested to know what the reasons are that SRDF/SRDF-A are not viable under teradata.

I suppose what people really need is confidence. TRS and TDM probably do not cover it all, as IMHO they do not provide true contingency , there could be too many points where human error interferes with what is copied/replicated and what is not, which may mean that the system is unusable should you lose the original computer centre in a fire or other disaster. This is not to say the software does not work as designed.

Some hardware based replication at disk level/ SRDF-like with software to make it work in contingency sceanarios that would be nice. a sort of fit and forget option.?

Random.