What if all transactions required strict global consistency?

My mum recently moved house. Being the dutiful son that I am I agreed to help her pack up her old house, drive to her new place and help unpack when we got there.

As it happens the most arduous part of the day did not involve packing, driving or unpacking but waiting: waiting for the various solicitors involved to confirm that the appropriate funds had been deposited in the appropriate bank accounts before the estate agents could hand over the keys.

It took hours, and was a reminder that while we might think of bank transfers as being instantaneous, there can be considerable delays involved in confirming that the correct amount has been debited from one bank account and credited to another.

I was reminded of the delays involved in this process the other delay while reading ODBMS Industry Watch’s interview with Basho CTO Justin Sheehy in which he compares financial transactions with the eventually consistent approach to database transactions.

“Traditional accounting is done in an eventually-consistent way and if you send me a payment from your bank to mine then that transaction will be resolved in an eventually consistent way. That is, your bank account and mine will not have a jointly-atomic change in value, but instead yours will have a debit and mine will have a credit, each of which will be applied to our respective accounts.”

The suggestion that bank transactions are not immediately consistent appears counter-intuitive. Comparing what happens in a transaction with a jointly atomic change in value, like buying a house, with what happens in normal transactions, like buying your groceries, we can see that for normal transactions this statement is true.

We don’t need to wait for the funds to be transferred from our accounts to a retailer before we can walk out the store. If we did we’d all waste a lot of time waiting around.

This highlights a couple of things that are true for both database transactions and financial transactions:

  • that eventual consistency doesn’t mean a lack of consistency
  • that different transactions have different consistency requirements
  • that if all transactions required strict global consistency we’d spend a lot of time waiting for those transactions to complete.
  • The Data Day, Three days: August 15/16/17 2012

    Symantec teams CFS with Hadoop. Informatica Cloud. And more

    And that’s the Data Day, today.