Craig Mckenna, Director, IBM Storage, Asia Pacific
DataStorageAsean: Is there a place for the relational database in this new world of unstructured and big data?
Craig: While the attention has been on unstructured data and Big Data, the world of relational databases has not really gone away. While new databases such as Hadoop are mainly analytical in nature, there has been very little erosion in the transactional database environment. That's not to say the market has not been impacted but the growth has definitely slowed. New NoSQL databases have taken some of the growth in the web and mobile transactional segments leaving relational databases mainly in the data center with existing mission critical transactional applications.
In other words, the demands for high performance and low latency infrastructure for Systems of Record (SOR) has always been there. As we start to connect and integrate these SOR with Systems of Engagement, we exacerbate the demands on SOR environments making performance ever more important to maintain a positive user experience.
Storage architectures must evolve to support the dramatically changing and increasing volume and variety of data in "the new world" of big data. IBMs Software Defined Storage (SDS) solutions provide the speed, flexibility and efficiency needed in today's enterprise. Better yet, the operational and economic benefits of these SDS solutions can be overlaid on your existing infrastructure to leverage the investments you've already made whilst modernising them to support the demands that the constantly changing business landscape places on IT.
DataStorageAsean: How is cloud computing affecting the Database market - should companies be thinking of moving all of their databases to the cloud?
Craig: As most of the growth in the Database market has been in the space of new web and mobile applications as well as Big Data, the financial and technical value proposition of the Cloud have been extremely compelling. As customers look to specialized databases such as GraphDBs, JSON and KeyValue to meet their new requirements, the complexity of managing both the environments and the data has increased significantly.
Companies will move to (or more correctly adopt) cloud when the attributes of cloud (flexibility, speed etc.) satisfy the demands of the workload in question and don't introduce significant risk (performance, security etc.) Some database workloads are born or can be moved to the public cloud. Others will be born or moved to on premise or hosted private cloud. Some will remain in legacy deployment models. So, when companies are thinking about cloud they should not think of it in terms of should I use "public cloud" but rather “do I need the attributes of cloud?” If yes, then they should think about the right deployment option (private, public or hybrid). The most likely option is a hybrid cloud and the most important thing is to ensure they have a consistent management, provisioning and governance model irrespective of where the workload "lives" today or where it might move to in the future.
Cloud offers customers a low cost environment to try and test new technologies and the ability to have elastic scalability as they rapidly ramp up new environments. Some of the new cloud environments offer the ability for customer to mix and match their deployments allowing customer not just increased flexibility but also reducing integration issues by offering prebuilt integration across the various databases. Especially in the world of analytical workloads, many cloud service providers are offering managed databases as a service allowing customers to focus on their analytics leaving the technology to the cloud providers.
DataStorageAsean: How important is it these days for databases to have hooks into other products such as BI or even Big Data technologies?
Craig: As organizations adopt analytics into every part of their business, it’s all the more critical for database technologies to not just have hooks but have native support for BI and complex analytics running in-database. Being able to run complex analytics such as predictive analytics within the database offers both significant performance advantages as well as reduced complexity in managing data at the BI layers. Furthermore, with the growth of Big Data, organisations need the flexibility to be able to manage their data across various database systems i.e. storing high value data in high performance databases (usually relational with in-memory and columnar support) and storing archived or older data as well as raw data in low cost platforms like Hadoop. But in an analytical environment, customers need the ability to analyse their data across the systems without having to run queries multiple times across different databases or even having to move and join the data at the more expensive end. After all, customers should have the flexibility to be able to run their queries in their databases across data in other systems such as Hadoop and more importantly taking advantage of the low cost MPP architecture of technologies of Hadoop.
DataStorageAsean: What is the future for Relational Databases?
Craig: There will always be a future for Relational Databases in complex transactional workloads. Customers may adopt new open source relational databases to lower costs of proprietary databases for non-mission critical databases requiring a lower service levels. But they continue to invest in relational databases for such workloads. The cost and effort in building complex transactional applications on NoSQL databases is significantly higher. It also increases the resources required to support the environment.
DataStorageAsean: What is unique about your own company's database offerings?
Craig: IBM's unique approach is based on the theme of "Open for Data". We believe in offering customers the option to select the offering that suits them best.
- Open Deployment: Choice to deploy on appliances, custom hardware or in the cloud (both IBM and other leading cloud providers)
- Open Technology: Choice of varied range of databases from relational to NOSQL to Hadoop, giving customers the option to deploy their data on the platform that best meets their application requirements. Hadoop, for example, consists of many open source modules. One of the primary open source modules is the Hadoop Distributed File System (HDFS), which is a distributed file system that runs on commodity hardware.
HDFS has some good features but it also has its limitations. For example, it's not a POSIX compliant file system so integration with other systems (or backup products etc.) requires more work than it otherwise would. In addition, to provide robust data availability, 3 copies of data are typically recommended which can turn even "commodity hardware" into an expensive proposition, so it's worth at least evaluating alternatives.
IBM Spectrum Scale™, which is a scale-out distributed file system, offers an enterprise-class alternative to HDFS. IBM Spectrum Scale provides integration with Hadoop applications that use the Hadoop connector (so you can use IBM Spectrum Scale enterprise-level functions on Hadoop).
- Open Source: While IBM offers our own enterprise class databases such as DB2 and Informix for mission critical applications on all platforms, we also have embraced open source technologies in our offerings. Building enterprise class functionalities on open source databases such as CouchDB (Cloudant), Apache Hadoop (BigInsights) etc.
- Open Analytics: Our database offerings allow for in-database analytical functions. Customers can run complex analytics such as R in a wide range of our offerings.
- Managed Services: While customers can host IBM databases in the cloud, IBM also offers managed database as a service with offerings such as DashDB and Cloudant. With IBM's CloudData Services and Compose acquisition, IBM offers customers the ability to provision and manage IBM Database platforms as well various opensource platforms such as postgres, mongodb and redis.