Serverless computing has recently attracted a lot of attention from research and industry due to its promise of ultimate elasticity and operational simplicity. However, there is no consensus yet on whether or not the approach is suitable for data processing. In this paper, we present Lambada, a serverless distributed data processing framework designed to explore how to perform data analytics on serverless computing. In our analysis, supported with extensive experiments, we show in which scenarios serverless makes sense from an economic and performance perspective. We address several important technical questions that need to be solved to support data analytics and present examples from several domains where serverless offers a cost and performance advantage over existing solutions. CCS CONCEPTS• Information systems → Parallel and distributed DBMSs; Online analytical processing engines; Database query processing.

Key-Value Stores (KVS) are becoming increasingly popular because they scale up and down elastically, sustain high throughputs for get/put workloads and have low latencies. KVS owe these advantages to their simplicity. This simplicity, however, comes at a cost: It is expensive to process complex, analytical queries on top of a KVS because today's generation of KVS does not support an efficient way to scan the data. The problem is that there are conflicting goals when designing a KVS for analytical queries and for simple get/put workloads: Analytical queries require high locality and a compact representation of data whereas elastic get/put workloads require sparse indexes. This paper shows that it is possible to have it all, with reasonable compromises. We studied the KVS design space and built TellStore, a distributed KVS, that performs almost as well as state-of-the-art KVS for get/put workloads and orders of magnitude better for analytical and mixed workloads. This paper presents the results of comprehensive experiments with an extended version of the YCSB benchmark and a workload from the telecommunication industry.

Cloud-based data analysis is nowadays common practice because of the lower system management overhead as well as the pay-asyou-go pricing model. The pricing model, however, is not always suitable for query processing as heavy use results in high costs. For example, in query-as-a-service systems, where users are charged per processed byte, collections of queries accessing the same data frequently can become expensive. The problem is compounded by the limited options for the user to optimize query execution when using declarative interfaces such as SQL. In this paper, we show how, without modifying existing systems and without the involvement of the cloud provider, it is possible to significantly reduce the overhead, and hence the cost, of query-as-a-service systems. Our approach is based on query rewriting so that multiple concurrent queries are combined into a single query. Our experiments show the aggregated amount of work done by the shared execution is smaller than in a query-at-a-time approach. Since queries are charged per byte processed, the cost of executing a group of queries is often the same as executing a single one of them. As an example, we demonstrate how the shared execution of the TPC-H benchmark is up to 100x and 16x cheaper in Amazon Athena and bigquery than using a query-at-a-time approach while achieving a higher throughput.

The enormous quantity of data produced every day together with advances in data analytics has led to a proliferation of data management and analysis systems. Typically, these systems are built around highly specialized monolithic operators optimized for the underlying hardware. While effective in the short term, such an approach makes the operators cumbersome to port and adapt, which is increasingly required due to the speed at which algorithms and hardware evolve. To address this limitation, we present Modularis , an execution layer for data analytics based on sub-operators , i.e., composable building blocks resembling traditional database operators but at a finer granularity. To demonstrate the feasibility and advantages of our approach, we use Modularis to build a distributed query processing system supporting relational queries running on an RDMA cluster, a serverless cloud platform, and a smart storage engine. Modularis requires minimal code changes to execute queries across these three diverse hardware platforms, showing that the sub-operator approach reduces the amount and complexity of the code to maintain. In fact, changes in the platform affect only those sub-operators that depend on the underlying hardware (in our use cases, mainly the sub-operators related to network communication). We show the end-to-end performance of Modularis by comparing it with a framework for SQL processing (Presto), a commercial cluster database (SingleStore), as well as Query-as-a-Service systems (Athena, BigQuery). Modularis outperforms all these systems, proving that the design and architectural advantages of a modular design can be achieved without degrading performance. We also compare Modularis with a hand-optimized implementation of a join for RDMA clusters. We show that Modularis has the advantage of being easily extensible to a wider range of join variants and group by queries, all of which are not supported in the hand-tuned join.

Key-value stores propose a very simple yet powerful data model. Data is modeled using key-value pairs where values can be arbitrary objects and can be written/read using the key associated with it. In addition to their simple interface, such data stores also provide read operations such as full and range scans. However, due to the simplicity of its interface, trying to optimize data accesses becomes challenging. This work aims to enable the shared execution of concurrent range and point queries on key-value stores. Thus, reducing the overall data movement when executing a complete workload. To accomplish this, we analyze different possible data structures and propose our variation of a segment tree, Updatable Interval Tree. This data structure helps us co-planning and co-executing multiple range queries together, as we show in our evaluation.