(This blog was adapted from the June Continuent Newsletter’s “Word from the CEO.”)
Continuent is a bit of a different bird in the open source community. Yes, our logo is Whooper Swans (the national bird of my native Finland), but that does not explain the difference. There are two key elements (which often separate us from many other open source operators, such as Percona) to our MySQL HA/DR and Continuous Operations solutions:
- Use of asynchronous replication and
- Offering a complete, proven, fully-tested and 24/7-supported solution under a commercial license (versus home-grown patchwork solutions that use open source components with only ad-hoc support)
The choice of the asynchronous technology is by design. Asynchronous approach is just a better way to do highly-available global database clusters.
That said, the design choice and decision was made by us only after two ‘failed’ attempts to solve the MySQL high-availability and disaster-recovery problem using synchronous replication. To be fair, these attempts did not fail completely as both Continuent m/cluster (available 2003-2006, the predecessor of Galera) and Continuent uni/cluster (2006-2009, predecessor of Tungsten Connector, aka Tungsten Proxy) had a large number of customers using them successfully in business-critical environments, but the limitations of synchronous replication became painfully clear in too many cases where the solution did not perform adequately.
Among my friends I am known to be an 'eternally optimistic’ guy, but even I am not optimistic enough to really believe that the ‘optimistic conflict resolution’ used by the current synchronous cluster solutions would truly work. This lack of belief comes to me especially easily after having tried twice with our own synchronous solutions… With synchronous clusters there are always issues with performance and geographic scaling. The higher the number of transactions, the larger the size of the transactions, the larger the cluster, and the longer the distance between clustered databases, the worse synchronous replication performs, and often causes the database performance to slow to a crawl.
We often also hear that asynchronous replication creates a possible data loss window, thus it is not an acceptable solution for certain types of applications. Yes, this is true. In theory, that is. In practice, we have developed the Tungsten Cluster solution to the point that the data loss really does not happen in real life. Please read this blog to understand why not!
When it comes to the commercial license for an open source database solution, that is a bit trickier.
The open source purists do not accept Continuent's presence, and we have frequently been shut out from talking at Percona Live and similar events, even if we enable MySQL servers for use cases that no other solution currently available on the market place can match. Thus, in our mind, we clearly have made significant contributions for the open source realm by enabling MySQL to become a strong alternative for Oracle and MS SQL Server in the enterprise computing space and beyond. But no, some people would rather use sub-par DIY solutions with expensive consulting services attached to them. And in many cases just because of policy reasons, not based on what could be the best solution. Go figure.
We could consider open-sourcing Tungsten Clustering to expand the market reach and the rate of adoption.
Would that really work for the benefit of our customers and open source users in general? I really doubt that. What would likely happen is that Peter the Great and his troops would swoop in, raid our technology, rename it, and offer it free of charge, but without the very fast and experienced 24/7 support needed for business-critical deployments. This could effectively destroy our revenue stream and kill any future development of Tungsten Cluster. Not sure how this would help our current customers and MySQL adoption in business-critical cases in general.
Despite the opposition from some MySQL corners, we believe that the work we have done since the early days of MySQL, (i.e. since 2002) has been highly beneficial for the MySQL community as a whole and we will stick with the course we have chosen!
Smooth sailing, steady as she goes!
Eero Teerikorpi
CEO & Founder, Continuent
Comments
Add new comment