Dean Jacobs

Predicting in-memory database performance for automating cluster management tasks

In Software-as-a-Service, multiple tenants are typically consolidated into the same database instance to reduce costs. For analytics-as-a-service, in-memory column databases are especially suitable because they offer very short response times. This paper studies the automation of operational tasks in multi-tenant in-memory column database clusters. As a prerequisite, we develop a model for predicting whether the assignment of a particular tenant to a server in the cluster will lead to violations of response time goals. This model is then extended to capture drops in capacity incurred by migrating tenants between servers. We present an algorithm for moving tenants around the cluster to ensure that response time goals are met. In so doing, the number of servers in the cluster may be dynamically increased or decreased. The model is also extended to manage multiple copies of a tenants data for scalability and availability. We validated the model with an implementation of a multi-tenant clustering framework for SAPs in-memory column database TREX.

Extensibility and Data Sharing in evolving multi-tenant databases

Software-as-a-Service applications commonly consolidate multiple businesses into the same database to reduce costs. This practice makes it harder to implement several essential features of enterprise applications. The first is support for master data, which should be shared rather than replicated for each tenant. The second is application modification and extension, which applies both to the database schema and master data it contains. The third is evolution of the schema and master data, which occurs as the application and its extensions are upgraded. These features cannot be easily implemented in a traditional DBMS and, to the extent that they are currently offered at all, they are generally implemented within the application layer. This approach reduces the DBMS to a ‘dumb data repository’ that only stores data rather than managing it. In addition, it complicates development of the application since many DBMS features have to be re-implemented. Instead, a next-generation multi-tenant DBMS should provide explicit support for Extensibility, Data Sharing and Evolution. As these three features are strongly related, they cannot be implemented independently from each other. Therefore, we propose FLEXSCHEME which captures all three aspects in one integrated model. In this paper, we focus on efficient storage mechanisms for this model and present a novel versioning mechanism, called XOR Delta, which is based on XOR encoding and is optimized for main-memory DBMSs.

Towards Analytics-as-a-Service Using an In-Memory Column Database

For traditional data warehouses, mostly large and expensive server and storage systems are used. For small- and medium size companies, it is often too expensive to implement and run such systems. Given this situation, the SaaS model comes in handy, since these companies might opt to run their OLAP as a service. The challenge is then for the analytics service provider to minimize TCO by consolidating as many tenants onto as few servers as possible, a technique often referred to as multi-tenancy.

Strict SLAs for Operational Business Intelligence

Today, SLAs for SaaS business applications usually lack stringent service level objectives and significant penalties. Moreover, Operational Business Intelligence features of modern business applications, like analytic dashboards, result in mixed workloads which make it even more difficult to predict execution times accurately due to resource contention. In contrast to the traditional three-tier architecture, an architecture for SaaS business applications should combine application and database layer to allow for processing business transactions and queries according to a queuing approach which enables strict SLAs with stringent response time and throughput guarantees. With stricter SLAs it would be easier to compare different cloud offerings with on-premise solutions and thus cloud computing could become more attractive for potential customers.

Towards enterprise software as a service in the cloud

For traditional data warehouses, mostly large and expensive server and storage systems are used. In particular, for small- and medium size companies, it is often too expensive to run or rent such systems. These companies might need analytical services only from time to time, for example at the end of a billing period. A solution to overcome these problems is to use Cloud Computing. In this paper, we report on work-in-progress towards building an OLAP cluster of multi-tenant main memory column databases on the Amazon EC2 cloud computing environment, for which purpose we ported SAPs in-memory column database TREX to run in the Amazon cloud. We discuss early findings on cost/performance tradeoffs between reliably storing the data of a tenant on a single node using a highly-available network attached storage, such as Amazon EBS, vs. replication of tenant data to a secondary node where the data resides on less resilient storage. We also describe a mechanism to provide support for historical queries across older snapshots of tenant data which is lazy-loaded from Amazons S3 near-line archiving storage and cached on the local VM disks.

Operational Business Intelligence: Processing Mixed Workloads

Modern business applications have operational business intelligence features that require processing analytical queries and business transactions together at the same time and on the same data. This results in mixed workloads, which are a big challenge for current database management systems. Cloud computing fosters the development of data management systems optimized for specific application scenarios. The authors propose an architecture for software-as-a-service business applications that helps process these mixed workloads with a special-purpose main-memory database system, while meeting strict service-level objectives with stringent response-time and throughput guarantees.

Cloudy skies for data management

Cloud computing is a fast emerging computing paradigm for which no commonly accepted definition has yet emerged in the research community. Often, cloud computing denotes a business or service model where a hosted software or infrastructure is made available on a subscription or pay-peruse basis. Due to the massive parallel scale of the hosting environment, cloud computing is often also associated with highly scalable, fault-tolerant, and adaptable computing on large clusters of off-the-shelf computers. The simplicity as perceived from a users perspective greatly distinguishes Cloud computing from grid computing. This is achieved by a much simpler API at a higher level of abstraction, which is used to deploy and to execute software and to access entire infrastructure stacks as services; the user pays for the service based on its usage. Another advantage of a Cloud computing platform is that the amount of resources required for performing a task does not have to be specified upfront; Cloud computing platforms are explicitly able to handle rapid scale-out. This flexibility allows for economic usage of resources based on demand. Currently, there is an active community evolving in the space of data management on cloud platforms, spurred to a large extent by the the Map/Reduce paradigm and the open-source framework Hadoop. Researchers explore query languages on top of the map/reduce paradigm, e.g., PigLatin, JAQL, Hive, optimization of map/reduce programs, and investigate and even teach map/reduce as a programming paradigm for web-scale computing. Researchers also explore multi-tenancy and the impact of the services model on infrastructure and application development.