Data Integrity

In the case of a block bit error, it is critical that this error is detected and masked by the storage system. DDS provides this functionally through its RAID Plus system. In all the testing and lifetime operation of DDS we have not yet seen a case where an error is detected but unable to be corrected. However, were this to happen, the DDS would report the error and the restore or read operation would fail. This is far preferable to the standard approach which is to simply return incorrect data.

While such events are rare, they do happen, many will be familiar with a strange file error or image corruption on an old memory card. The same effect happens with new memory cards (particularly MLC cards) and in fact the embedded controller is constantly fighting with the slow degradation of raw flash memory. Typically memory cards are designed to hold data for short periods, before it can be transferred to a permanent storage mechanism. Such cards are clearly unsuitable for production use and certainly not in a telco environment where reliability is critical.

Some exchanges provide data checksumming features which will detect errors and indicate a failure to restore. However in our experience most do not provide this features, or it is not capable of detecting errors in all of the data transferred. Generally there is no possibility of error correction, so a large restore may fail right at the end wasting hours of time. DDS's data integrity features detect such problems immediately and prevent them from affecting the transfer, by correcting the errors at source.

Long cable runs are another cause of data integrity problems, particularly when cables are old. While telephone exchanges are normally fairly controlled environments, in our experience this is not always the case. DDS detects cable problems using SCSI and Pertec parity features and reports these immediately to the user with an error symbol. It also causes a retry to make sure that correct data is received, before continuing. This can avoid data losses between the exchange and the DDS which are otherwise very hard to detect. Using this feature, cables which are well below specification can be used successfully with DDS (with slight performance degradation due to automatic retries), which indicates that a good margin of error is available.