NuoDB Scale-Out Performance Demo - YouTube NuoDB

This short demo video demonstrates how the NuoDB distributed SQL database scales elastically on demand, while maintaining full ACID compliance. Download the Free NuoDB software to experience consistency: http://dev.nuodb.com/download-nuodb/r... Transcription: NuoDB is a distributed database management system that allows you to dynamically add and remove active nodes without ever sacrificing transactional consistency. In this short video, we will demonstrate how the NuoDB database scales elastically on demand, while maintaining full ACID compliance. We are demoing an online application called “Storefront,” which uses the NuoDB database and connects to it through standard JDBC connections. Storefront simulates shoppers browsing products, adding purchases to a shopping cart, checking out, and writing product reviews. Now let’s switch to the Storefront demo screen to see how NuoDB is designed to scale dynamically - out and in - across host machines. In NuoDB, transactions are performed by in-memory processes called Transaction Engines or “TEs”. To increase database throughput add one or more TEs. If less throughput is required, you simply remove a TE. Although NuoDB operates at in-memory speeds it is a general purpose database. Similar to the transaction layer, the storage layer is designed to scale out for high-availability and redundancy. As you can see on the screen, we are running Storefront in the AWS US West datacenter at 100 concurrent users with 200 transactions per second or “TPS” -- shown also in the large graph below --- and a latency of around 20 milliseconds per transaction. As more shoppers – in our case 500 - get online, throughput will be decreasing below 170 TPS and latency will become unacceptably high. Since we want our customers to have an excellent online experience in our store, we’ll simply bring new TE’s online to handle the increased workload efficiently. To do this we will fire off 3 new TEs on separate host machines. As the TEs come online, transactions are now being distributed equally across all 4 TEs. Notice latency is rapidly decreasing while throughput is increasing in a linear fashion to around 900 TPS. (Wait) This on demand scale-out performance behavior is one of the key advantages of the NuoDB distributed architecture. Now, let’s reduce the number of concurrent users down to 200 while we continue to distribute the workload across all 4 TEs. As you can see the transactions per second decrease in lockstep with the number of concurrent users, which suggests that the database is overprovisioned. Because of NuoDB’s distributed architecture, we can stop TE processes without bringing down the database. As we cut the number of TEs to 2, you will see that the TPS and latency stay the same while using half the TE resources. This proves that the database was overprovisioned. This elastic scalability ensures that you use only the resources you need, when you need them. What we’ve seen in this demo - in real time - is that it’s easy to add or take away Transaction Engines to ensure the best database performance at the lowest latency. Just download NuoDB today and take storefront for a spin. Tech Blog: http://dev.nuodb.com/techblog Press Releases: http://www.nuodb.com/resources/press-... Follow us on: - Twitter (@nuodb): https://twitter.com/nuodb - Facebook: https://facebook.com/nuodb - LinkedIn: https://linkedin.com/company/nuodb

NuoDB Scale-Out Performance Demo - YouTube
16 Likes	16 Dislikes
5,230 views views	254 followers
Science & Technology	Upload TimePublished on 12 Dec 2014

>