\beginabstract The trend in the DBMS market is to migrate to the cloud for elasticity, high availability, and lower costs. The traditional, monolithic database architecture is difficult to meet these requirements. With the development of high-speed network and new memory technologies, disaggregated data center has become a reality: it decouples various components from monolithic servers into separated resource pools (e.g., compute, memory, and storage) and connects them through a high-speed network. The next generation cloud native databases should be designed for disaggregated data centers. In this paper, we describe the novel architecture of \name, which follows thedisaggregation design paradigm: the CPU resource on compute nodes is decoupled from remote memory pool and storage pool. Each resource pool grows or shrinks independently, providing \revon-demand provisoning at multiple dimensions while improving reliability. We also design our system to mitigate the inherent penalty brought by resource disaggregation, and introduce optimizations such as optimistic locking and index awared prefetching. Compared to the architecture that uses local resources, \name achieves better dynamic resource provisioning capabilities and 5.3 times faster failure recovery speed, while achieving comparable performance. \endabstract

We show that enforcing shift patterns significantly reduces the cost to construct grids without monochromatic rectangles. Additionally, we prove that all valid 3-colorings of a 10 by 10 grid are isomorphic.

With the increasing demand for real-time system monitoring and tracking in various contexts, the amount of time-stamped event data grows at an astonishing rate. Analytics on time-stamped events must be real time and the aggregated results need to be accurate even when data arrives out of order. Unfortunately, frequent occurrences of out-of-order data will significantly slow down the processing, and cause a large delay in the query response. Timon is a timestamped event database that aims to support aggregations and handle late arrivals both correctly (i.e., upholding the exactly-once semantics) and efficiently. Our insight is that a broad range of applications can be implemented with data structures and corresponding operators that satisfy associative and commutative properties. Records arriving after the low watermark are appended to Timon directly, allowing aggregations to be performed lazily. To improve query efficiency, Timon maintains a TS-LSM-Tree, which keeps the most recent data in memory and contains a time-partitioning tree on disk for high-volume data accumulated over long time span. Besides, Timon supports materialized aggregation views and correlation analysis across multiple streams. Timon has been successfully deployed at Alibaba Cloud and is a critical building block for Alibaba cloud's continuous monitoring and anomaly analysis infrastructure.

PolarFS is a distributed file system with ultra-low latency and high availability, designed for the POLARDB database service, which is now available on the Alibaba Cloud. PolarFS utilizes a lightweight network stack and I/O stack in user-space, taking full advantage of the emerging techniques like RDMA, NVMe, and SPDK. In this way, the end-to-end latency of PolarFS has been reduced drastically and our experiments show that the write latency of PolarFS is quite close to that of local file system on SSD. To keep replica consistency while maximizing I/O throughput for PolarFS, we develop ParallelRaft , a consensus protocol derived from Raft, which breaks Raft's strict serialization by exploiting the out-of-order I/O completion tolerance capability of databases. ParallelRaft inherits the understand-ability and easy implementation of Raft while providing much better I/O scalability for PolarFS. We also describe the shared storage architecture of PolarFS, which gives a strong support for POLARDB.

To increase the capacity of hard disk drive (HDD) with no significant cost, Seagate announced the world's first Shingled Write Disk (SWD) that employs Shingled Magnetic Recording (SMR) technology in 2014. The main drawback of SMR is that non-sequential write access to the disk is restricted due to overlap in the layout of data tracks. Seagate SWD mitigates these access restrictions by using an internal persistent cache (PCache). However, it encounters poor sustained non-sequential write performance due to high collection cost and low collection efficiency of its PCache, which makes it can only be used for backup/archive applications. In this paper, we propose a hybrid storage system HS-BAS to provide good sustained non-sequential write performance. In HS-BAS, we use SSD as the cache of SWD to reduce the effect of high collection cost of SWD PCache on system performance and implement three different cache collection policies based on band awareness of SWD to increase the collection efficiency of SWD PCache. Finally, our measurements on HS-BAS show that it provides superb sustained non-sequential write performance with a small capacity of SSD. For example, HS-BAS achieves 1.05 times and 0.26 times performance improvement compared with SWD and conventional HDD respectively for financial1 trace. By using conventional hybrid storage system Flashcache on SWD, we find that it only achieves little performance improvement and even worse performance in some traces. It is shown that HS-BAS performs better than Flashcache on SWD.

A cooperative caching system, using disks as its cache media, is proposed for network storage system. This system is called Blue Whale Cooperative Caching System (BWCC). Through sharing the cached data among clients of cluster file system, the load on the centralized storage server is lowered, therefore, the BWCC significantly enhances the scalability of the network storage system. The advantages of BWCC are as follows: 1) direct data positioning technology without the participation of the centralized storage server guarantees low latency of cooperative data access, 2) supporting several granularities of data sharing is applicable to multiple data access patterns, 3) a global cache management strategy for Video-on-Demand service is designed according to the characteristics of the video data access pattern. BWCC has been implemented as a module in Linux Kernel-2.6.32. The preliminary experimental results verify the effectiveness of BWCC.

In order to meet the challenges of significant storage and application growth, as well as shortened backup windows and limited IT resources, more and more organizations embrace Hierarchical Storage Management (HSM). Parts of SAN file systems provide the ability to support HSM, but they didn't provide more details in design considerations and how to implement. In this paper, we present design considerations in Blue Whale File System (BWFS) which is a high performance SAN file system and implement the system. Event, migration and attribute constitute the three aspects of data management mechanism in BWFS. As for READ/WRITE/TRUNCATE event, the problem is that a client directly accesses to storage devices containing file data, rather than through a metadata server. We monitor LayoutGet operation instead of real I/O path. BWFS is a multi-volume file system which can support single name space across multiple data volumes. When different storage devices are managed by BWFS, migration is between volumes in BWFS. We also implement migration between BWFS and external repository. We believe the solutions to the problem we encountered are useful to file system developers.

Deploying operating systems and services in an enterprise environment can be a time-consuming and cumbersome task. With the dramatic increase of the number of computers managed by the system administrator, the work of management is more complicated than before. Moreover, services provided by servers in data center change frequently, which requires a flexible service deployment system to achieve on demand deployment. To maximize return on investment (ROI) and stay competitive in today's fast-paced global economy, administrators must be able to deploy systems or services quickly and effectively with minimal downtime. This paper presents SonD, a system for providing dynamic, flexible, and quick service and operating system deployment in a LAN environment. The techniques used by SonD include an iterative snapshot mechanism based on an IP SAN system, which provides writable snapshots to achieve fast creation of the service, a dynamic service mapping method, and a network boot technology achieved by both software and hardware solutions. This design results in a system which can rapidly, flexibly, dynamically and reliably distribute OSes and services to many clients simultaneously. For example, SonD can deploy 50 windows XP systems in a few minutes on commodity PCs. We describe SonD's design and implementation, and review the key technologies for our system. We evaluate its performance and apply it to oil exploration. The results show that SonD fits well with the service deployment usage in cluster computing environment.