Migration methodologies vary, but all require a level of analysis, configuration of the migration tool and testing. Using traditional techniques such as sampling implies that the first time data level testing is applied will occurs late in the process. If the testing is reliant on sampling, it is easy to imagine why cost and schedule overruns become all too familiar.
Valiance takes a different approach and we strongly advocate the following:
Still not convinced? Valiance has experience testing millions of records and we have compiled a list of error types and the types of techniques required for identification including:
Our testing techniques for 100% post migration verification have been evaluated for each of these types of error against black box testing of the migration tool, post migration sampling techniques. There are two significant outcomes; first, no technique is foolproof, and Valiance’s techniques for testing reduces the likelihood of error by orders of magnitude over the alternative approaches. Note that we maintain a list of errors and the effectiveness of these techniques which is available upon request.
Sampling Theory
The discussion of migration testing is not complete without any review of the most common approach, sampling.
Traditional migration testing techniques rely upon sampling or testing a subset of the migration outcomes. However, experience demonstrates that this type of testing yields an all too familiar process; to test via sampling, find one of more errors, repair the error and repeat only to find that subsequent testing cycles identify new errors. According to sampling theory, “It is assumed that nonconformities occur randomly and with statistical independence within and between units”. This implies that the probability of a nonconformity is uniformly distributed across the data being sampled, that fact this migration testing is iterative in nature, where each execution identifies new issues proves that this assumption is not valid
.
