There are two main types of scalability: vertical and horizontal. Vertical scalability consists of potentiating the hardware specs of a given server. This is typically done by increasing the number (and to a minor extent, speed) of the processors, adding more RAM, and so on. Commodity hardware tends to impose a strong limit on the resources that can be augmented. There is only so much that can be stuffed into your typical x64_86 Dell server, even replacing the entire machine with the top of the line model (having the same architecture). For this very reason, many customers opt to purchase more expensive hardware that has the advantage of facilitating vertical scalability, by allowing much more headroom for additional resources. This is why Unix-based solutions like IBM System p + AIX and Sun’s Solaris + SPARC are common choices in demanding environments.
For all their merits, even premium hardware solutions can only do so much in terms of vertical scalability. They may postpone the problem, by allowing a large number of processors and a huge amount of RAM to be installed, but they are still not limitless. And that’s when it becomes cheaper and much more feasible to attempt horizontal scalability, or the process of adding extra servers that share data and the overall system load.
Commercial databases, particularly DB2, excel at vertical scalability, which incidentally is one of the weakest aspects of open source RDBMS like MySQL. When it comes to horizontal scalability though, even commercial databases tend do struggle a little. Scalability is achieved, but a lot of work is required to configure complex setups involving masters, slaves, replication, failover servers, and so on. Until now.
IBM has just introduced a new technology for DB2 called pureScale, which accomplishes three incredible feats:
- Horizontal scalability: Virtually unlimited capacity by simply adding new nodes to the cluster.
- Application transparency: No code changes needed.
- Continuos data availability: regardless of which server fails, the system will continue to operate albeit with a proportionally reduced throughput.
DB2 pureScale applies to DB2 running on Unix (IBM System p running AIX only, for the moment), the same principles and rock solid technology employed by DB2 for z/OS Data Sharing and SYSPLEX. This technology happens to be employed by some of the largest companies in the world and it’s considered the golden standard when it comes to easily create OLTP database clusters, providing virtually unlimited scalability capabilities.
You can read more about it here and here. If you are going to IOD this year, don’t forget to attend this overview by one the architects of DB2 pureScale. And even if you can’t head over to Las Vegas next week, you can still attend online our Chat with the Lab today at 11:30am EST. Database scalability is a hot topic and this is going to be one of the most popular chats with the lab ever, so register for free now.