Are Cloud database applications automatically limited

Do relational databases in a Cloud environment have a pe-determined ceiling before running into problems?

If this is the case what have Cloud developers done to address this issue? Alternatively, is the percieved transaction volume with these applications not regarded as high enough to make this an issue?

Here is an interesting article

http://dbmsmusings.blogspot.com/2010/08/problems-with-acid-and-how-to-fix-them.html

 

Comments
mkcdavies's picture

What's the real issue behind this question?

mkcdavies | | Permalink

I have been involved with cloud-based accounting applications for nearly 4 years (not including Hotmail and the like, which have been around a lot longer).  Well designed and well managed cloud applications are massively scaleable, meaning that the vendor is easily able to add resources (e.g. additional processors, servers or data storage) as transaction volumes grow. 

In 2001, Twinfield processed 3,350 transactions during the year.  In 2009 we processed 3,350 transactions per minute.  This growth simply hasn't been a problem.  By keeping an eye on performance and planning ahead for upgrades, users will never experience any performance degradation.

The cost of adding extra capacity keeps driving down (see Moore's Law) so adding extra data storage and/or processing power becomes more easily affordable as the number of paying users grows.

In our case, we have over 40,000 users managing the accounts for over 80,000 companies.  This number of users, while at the forefront of online accounting, is a mere drop in the ocean compared to some of the well known web-based e-mail providers and social media networks.  I simply can't see this being an issue providing the cloud-vendor has the right management processes in place.  I'm certainly not aware of any ceiling.

daveforbes's picture

the article

daveforbes | | Permalink

The article from my reading was about data that was not easily partitionable.

You quote 4,000 transactions per minute spread over 80,000 companies. You can naturally partition your data onto 80,000 servers if the urge took you and so you can use lots of cheap hardware.

The article was about high transaction volumes that were not easily partitionable. Say you had many thousands of transactions per second all all needing to go into the same table (say tesco's pos). Then the article was saying it is much more difficult to use large collections of cheap hardware to achieve this.

David

cverrier's picture

Why does this matter to the user?

cverrier | | Permalink

AccountingWeb probably isn't the forum for discussions about how to scale SaaS systems (although there are people on here who can do that if you want).

The point is - it's not a problem, and the technical aspects are well understood by the people who need to understand them.  As long as you design the system right in the first place.

There are any number of examples of INSANELY large cloud products that tick along just fine. Amazon, Salesforce, Hotmail itself, Google Apps.   Most SaaS bookkeeping apps will never come close to the transaction volumes that these systems handle every day.

If there are any SaaS providers who get it wrong, then they'll come a cropper and go out of business pretty fast.

 

mkcdavies's picture

It affects the price to the practice/client

mkcdavies | | Permalink

One of the business benefits of cloud computing is the ability to spread cost over many customers, leading to a relatively low price per user, a hallmark of the most successful online accounting systems.  If a separate partition or instance is needed for each customer or for a small number of customers (which could be how Sage 50 Online is set up), the cost per user increases significantly.

As the article says, it can be difficult to scale a system that wasn't designed with scaleability in mind.  That's one reason why it is no simple matter to take an existing product and re-engineer it for cloud computing.  Hence none of the traditional offerings have yet made it.  It's more likely that the likes of Sage will either build something new (they've already tried and failed once) or buy up one of the new players and Sage-ify it in some way to come up with a market offering that avoids decimating their revenue streams from traditional software licenses and SageCover - no easy task!

Add comment
Log in or register to post comments