Skip to content

On geo-distributed data management — Interview with Adam Abrevaya.

by Roberto V. Zicari on October 19, 2013

“Geo-distribution is the ability to distribute a single, logical SQL/ACID database that delivers transactional consistency across multiple datacenters, cloud provider regions, or a hybrid” — Adam Abrevaya.

I have interviewed Adam Abrevaya, Vice President of Engineering, NuoDB.

RVZ

Q1. You just launched NuoDB 2.0, what is special about it?

Adam Abrevaya: NuoDB Blackbirds Release 2.0 demonstrates a strong implementation of the NuoDB vision. It includes over 200 new features and improvements, making it even more stable and reliable than previous versions.
We have improved migration tools; included Java stored procedures; are introducing powerful automated administration; made enhancements to core geo-distribution functionality and more.

Q2. You offer a feature called geo-distribution. What is it and why is it useful?

Adam Abrevaya: Geo-distribution is the ability to distribute a single, logical SQL/ACID database that delivers transactional consistency across multiple datacenters, cloud provider regions, or a hybrid.

NuoDB’s geo-distributed data management lets customers build an active/active, highly-responsive database for high availability and low latency. By bringing the database closer to the end user, we can enable faster responses while simultaneously eliminating the time spent on complex tasks like replication, backup and recovery schemes.

One of the most exciting aspects of the Release 2.0 launch was the discussion about a major deployment of NuoDB Geo-Distribution by a customer. We were very excited to include Cameron Weeks, CEO and Co-Founder of Fathom Voice, talking about the challenges his company was facing—both managing his existing business and cost-effectively expanding globally. After a lengthy evaluation of alternative technologies, he found NuoDB’s distributed database is the only one that met his needs.

Q3. NuoDB falls broadly into the category of NewSQL databases, but you say that you are also a distributed database and that your architecture is fundamentally different than other databases out there. What’s different about it?

Adam Abrevaya: Yes, we are a NewSQL database and we offer the scale-out performance typically associated with NoSQL solutions, while still maintaining the safety and familiarity of SQL and ACID guarantees.

Our architecture, envisioned by renowned data scientist, Jim Starkey, is based on what we call “On-demand Replication”. We have an architecture whitepaper (registration required) which provides all the technical differentiators of our approach.

Q4. NuoDB is SQL compliant, and you claim that it scales elastically. But how do you handle complex join operations on data sets that are geographically distributed and at the same time scale (in) (out)?

Adam Abrevaya: NuoDB can have transactions that work against completely different on-demand caches.
For example, you can have OLTP transactions running in 9 Amazon AWS regions, each working on a subset of the overall database. Separately, there can be on-demand caches that can be dedicated to queries across the entire data set. NuoDB manages these on-demand ACID-compliant caches with very different use cases automatically without impact to the critical end user OLTP operations.

Q5. What is special about NuoDB with respect to availability? Several other NoSQL data stores are also resilient to infrastructure and partition failures.

Adam Abrevaya: First off, NuoDB offers a distributed SQL database system that provides all the ACID guarantees you expect from a relational database. We scale out like NoSQL databases, and offer support for handling independent failures at each level of our architecture. Redundant processes take over for failed processes (due to machine or other failures) and we make it easy for new machines and process to be brought online and added to the overall database dynamically. Applications that make use of the typical facilities when building an enterprise application will automatically reconnect to surviving processes in our system. We can detect network partition failures and allow the application to take appropriate measures.

Q6 How are some of your customers using NuoDB?

Adam Abrevaya: We are seeing a number of common uses of NuoDB among our customers. These range from startups building new web-facing solutions, to geo-distributed SaaS applications, to ISVs moving existing apps to the cloud, to all sorts of other apps that hit the performance wall with MySQL and other traditional DBMS. Ultimately, with lots of replication, sharding, new server hardware, etc., customers can use traditional databases to scale out or up but at a very high cost in terms of both time, money and usually by giving up transactional guarantees. One customer said he decided to look at alternatives to MySQL just because he was spending so much time in meetings talking about how to get it to do what they needed it to do. He added up the cost of the man-hours and he said “migrate.”

As I mentioned already, Fathom Voice, a SaaS provider offering VoIP, conference bridging, receptionist services and some innovative communications apps, had a global deployment challenge. How to get the database near their globe trotting customers; reduce latency and ensure redundancy. They are one of many customers and prospects tackling these issues.

———————-
Adam Abrevaya, Vice President of Engineering, NuoDB
Adam has been building and managing world-class engineering teams and products for almost two decades. His passion is around building and delivering high-performance core infrastructure products that companies depend on to build their businesses.

Adam started his career at MIT Lincoln Laboratory where he developed a distributed platform and image processing algorithms for detecting dangerous weather patterns in radar images. The system was deployed at several airports around the country.

From there, Adam joined Object Design and held various senior management positions where he was responsible for managing several major releases of ObjectStore (an Object database) along with spearheading the development team building XML products that included: Stylus Studio, an XML database, and a Business Process Manager.

Adam joined Pantero Corporation as VP of Development where he developed a revolutionary Semantic Data Integration product. Pantero was eventually sold to Progress Software.

From Pantero, Adam joined m-Qube to manage and build the team creating its Mobile Messaging Gateway platform. The m-Qube platform is a carrier grade product that has become the leading Mobile Messaging Gateway in North America and generated billions of dollars in revenue. Adam continued managing the mQube platform along with expanded roles after acquisitions of the technology from VeriSign and Mobile Messenger.

———

Related Posts

- On Big Data and NoSQL. Interview with Renat Khasanshyn. October 7, 2013

- On NoSQL. Interview with Rick Cattell. August 19, 2013

Resources

- Download NuoDB Pro Edition (Registration required) (NuoDB Blackbirds Release 2.0)

-ODBMS.org free resources on
Relational Databases, NewSQL, XML Databases, RDF Data Stores:
Blog Posts |Free Software | Articles and Presentations| Lecture Notes | Tutorials| Journals |

Follow ODBMS.org on Twitter: @odbmsorg

##

From → Uncategorized

No comments yet

Leave a Reply

Note: HTML is allowed. Your email address will not be published.

Subscribe to this comment feed via RSS