All products uses the same principles that i outlined above to handle transactions processing in a scale-out environment.

So from a pattern perspective the way all products deals with XTP is pretty much the same. We all gone through various competitive situations and were forced to converge on those terms simply since most customers want the same thing when it comes to transaction processing and most customers want to at a position where they would be able to compare all various solutions.

I think the question is where to draw the line. I am personally very interested in the new models, and think they have a lot of potential.

Yes, please let us know if you think we haven’t covered the material correctly. It’s always good to hear another point of view!

Thanks,

Eric

I understand your point of view even though i may have different perspective wrt to the maturity of the new transactions processing models , (XTP) as they are referred to by Gartner.

I’ll have a deeper look at your book and let you know if i have any further comments.

Good luck anyway!
Nati S.

Yes, we are well aware of the in-memory replication products starting to emerge from what I would call the “scale out” space, which I would also characterize as pioneered by the large web site infrastructures. And I know personally about the adoption of some of these new products, including yours, in the financial services industry to reduce trading latency etc.

Coherence, GigaSpaces, and ExtremeScale are examples of products that are starting to emerge, but we could not really say that any one of them is either a complete TP environment, or widely enough adopted, for us to use as the basis of examples that illustrate the principles and concepts in the book.

The focus of the book is on explaining how it all works. We have an entire chapter on replication, which covers the topic you reference in your comment.

Eric

I believe that if your referring only the internet companies as a reference then your probably right. Based on my own personal experience majority of the application in the financial world (Order managemetn) are not using the traditional 2pc model and there are products that are well established in that space already and are used to run those financial applications in production in the past few years. A good example is obviously the company that i work for (GigaSpaces) but IBM (ExtremeScale) and Oracle (Oracle Coherence) have a fairly good solution on that space as well.

The pattern that i often use to describe this model is named Space Based Architecture which can be summarized as follows:

# Rule #1 – Build your data model from a set of independent data items (Tuples)
# Rule #2 – Partition your different data items based on your data affinity (Customer-id, account-id,..)
# Rule #3 – Execute transaction as a set of individual steps scoped within a single partition.
# Rule #4 – Minimize the amount of moving parts involved under the context of a transaction

You can read full details here

Cheers
Nati S.

We do recognize and understand the trends away from SQL and 2PC, especially for large web site systems. However when we investigated them, we discovered that everyone currently creates a lot of custom code, and we could not identify common patterns among the applications using some of the techniques you are referring to.

The book is primarily focused on current practice, meaning we need to be able to include examples of how developers would create programs using widely adopted product APIs to include these concepts and principles into TP applications, and our conclusion is that such products do not yet exist.

To be clear, some products are beginning to emerge, but the trends are not yet mature enough to clearly predict which of them will achieve widespread adoption – such as SQL, Java EE, or .NET, to name a few.

The industry is in transition, and it’s too early to focus the book on the new trends, regardless of how interesting they are – and things like Stonebreaker’s paper, and Helland’s papers, and the Dynamo paper are very interesting indeed.

One reviewer asked us to please invent a new type of transaction coordination for loosely-coupled TP application architectures. As much as we’d like to, proposing a new design is one thing, having it adopted and used in large TP deployments is another.

We are as interested as anyone in what’s going on with the new database designs, compensation-driven transaction patterns, relaxed consistency, and in-memory replica updating and synchronizing. And these things are all included, yes, but the main focus of the book was and remains firmly on current practice.

I’m looking forward to your feedback on the book!

Thanks,

Eric