Monthly Archives: May 2014
The presentation about tools and frameworks at GitHub for gaining confidence on changing codes. Test suite provides certain protections against regressions, but there are other spots to cover.
According to the presentation, they are accumulating various statistical metrics, and also automating them through HUBOT commands, along with standard development/deploying workflow on GitHub features.
The above tools are interesting to verify the impact of code change on production load, without changing behavior. It sounds a little similar as Cookpad’s chanko which provides isolated production deploy, but dat-science seems more focusing on experiment than deployment.
Nice quick demo to configure simple web servers with load balancing capability on OpenStack.
This long version covers basic OpenStack introduction through the installation settings and operational demonstrations.
From the demonstration, dashboard features seems being enhanced rapidly, and it sounds getting closer to the AWS Management Console for the “basic” PaaS capability. The demo is only the main-path and managing the underlying physical hardware might involve tedious error handling and troubleshooting yet (since GUI is very simple). But, once it gets matured along with the current growth-pace, it would like a great option to control the virtual environments.
A little old one, but it’s a nice panel discussion about NoSQL databases. Each participant represents a specific NoSQL database product, but having Martin Fowler additionally helps maintaining the balance between the NoSQL and SQL in a practical way. NoSQL world sometimes criticizes SQL, but many people are accustomed to SQL, and are using them along with certain amount of complaints (maybe as similar way as Java getting criticism, in some sense).
A lot of new data-store options are available lately, along with the various ‘BigData’ requirements and technical transition to distributed systems. As mentioned in the discussion, it’s getting more difficult to choose a data-store which fits all use cases. Even a single application can require multiple different requirements.
NoSQL databases are relatively new and they still require specialists and experts to design into the system properly, due to the different type of complexity. It is largely imposed by the nature of distributed systems, in contrast to their simple data model. It’s pretty much tough to catch up, but this kind of discussion helps understanding the underlying concepts.