We have designed and developed OceanBase, a distributed relational database system from the very basics for a decade. Being a scale-out multi-tenant system, OceanBase is cross-region fault tolerant, which is based on the shared-nothing architecture. Besides sharing many similar goals with alternative distributed DBMS, such as horizontal scalability, fault-tolerance, etc., our design has been driven by the demands of typical RDBMS compatibility as well as both on-premise and off-premise deployments. OceanBase has fulfilled its design goal. It implements the salient features of certain mainstream classical RDBMS, and most applications on them can run on OceanBase, with or without a few minor modifications. Tens of thousands of OceanBase servers have been deployed in Alipay.com as well as many other commercial organizations. It has also successfully passed the TPC-C benchmark test and seized the first place with more than 707 million tpmC. This paper presents the goals, design criteria, infrastructure, and key components of OceanBase including its engines for storage and transaction processing. Further, it details how OceanBase achieves the above leading TPC-C benchmark in a distributed cluster with more than 1,500 servers from 3 zones. It also describes lessons what we have learnt in building OceanBase for more than a decade.

In a traditional distributed database system, the partitions use two-phase locking (2PL) as the concurrency control protocol to ensure distributed read consistency. But the read-lock acquired by a read operation is incompatible with a write-lock, which undermines the performance of the system. While in a system at the snapshot isolation level, where partitions use Multi-Version Concurrent Control (MVCC) as the concurrent control protocol, distributed read inconsistency may occur. To achieve read consistency and guarantee the performance at the same time, we propose Distributed Multi-Version Concurrent Control (DMVCC). With DMVCC, the system can support snapshot reads, which do not block write operations, and ensure distributed read consistency. In this protocol, a transaction obtains a set of consistent snapshot version numbers at the startup time. The transaction then uses those numbers to read the corresponding data stored on each partition. The correctness of the protocol is strictly proved. We conduct a series of experiments to compare the performance of the system when using and not using DMVCC with a scaled TPC-C benchmark. We observe that our DMVCC based system outperforms the system using 2PL at both medium (up to 1.53x speed up) and high contention (up to 2.0x speed up) levels. Furthermore, when read/write ratio goes up to 1:1, the throughput of the DMVCC based system is 290% higher than that of the system using 2PL. The scalability of the system is also presented.

We propose a compact test generation method for transition faults based on a conflict avoidance driven scheme. A new measure is proposed to estimate the influence inputs, which is the subset of inputs to be specified, needed for detecting transition faults. The value requirements at the pseudoprimary inputs (PPIs) of the second frame of the automatic test pattern generation circuit model are partitioned into separate subsets. The sequential backtracing scheme backtraces the value requirements on the PPIs of the second frame subset-by-subset. Justification of the necessary value requirements at the PPIs in the second frame is completed using a conflict-driven procedure. With an influence input measure for transition faults under the launch-on-capture scan testing application scheme, a new dynamic test compaction scheme is proposed. The new test compaction scheme tries to compact as many faults as possible into the current test. Experimental results and comparison with existing approaches demonstrate the efficiency and effectiveness of the proposed method.

Two conflict-driven schemes and a new scan architecture based on them are presented to improve fault coverage of transition fault. They make full use of the advantages of broadside, skewed-load and enhanced scan testing, and eliminate the disadvantages of them, such as low coverage, fast global scan enable signal and hardware overhead. Test power is also a challenge for delay testing, so our method tries to reduce the test power at the same time. By the analysis of the functional dependency between test vectors in broadside testing and the shift dependency between vectors in the skewed-load testing, some scan cells are selected to operate in the enhanced scan and skewed-load scan mode, while others operate in traditional broadside mode. In the architecture, scan cells with common successors are divided into one chain. With the efficient conflict driven selection methods and partition of scan cells, fault coverage can be improved greatly and test power can be reduced, without sacrificing the test time and test data. Experimental results show that the fault coverage of the proposed method can reach the level of enhanced scan design.

Testing for small-delay defects requires fault-effect propagation along the longest testable paths. However, the selection of the longest testable paths requires high CPU time and leads to large pattern counts. Dynamic test compaction for small-delay defects has remained largely unexplored thus far. We propose a path-selection scheme to accelerate ATPG based on stored testable critical-path information. A new dynamic test-compaction technique based on structural analysis is also introduced. Simulation results are presented for a set of ISCAS'89 benchmark circuits.

In this paper, we propose a novel way to save test power, using the DFT based technique as basic method and post-generation filling as complementary. In this architecture, two methods of clustering flip-flops into scan chains are presented. One is clustering scan flip-flops into two parts to save capture power, and the other is clustering scan flip-flops of each part into scan chains to save shift power. By partitioning the scan flip-flops into two parts, the capture operation is cut into two sequential steps, which can effectively reduce the capture power. By partitioning flip-flops with common successors into one chain, we can make sure that, only one or a small part of scan chains are active during the shifting phase. For other scan chains, filling strategy is used as complementary method to further reduce test power. This architecture can effectively reduce the test time too. Experimental results show that average power reduction of 91.81% and average peak power reduction of 49.35% can be achieved, comparing to the ordinary full-scan architecture.

This paper presents two new conflict-driven techniques for improving transition fault coverage using multiple scan chains. These techniques are based on a novel test application scheme, in order to break the functional dependency of broadside testing. The two new techniques analyze the ATPG conflicts in broadside test generation, and try to control the flip-flops with most influence on the fault coverage.The conflict-driven method selects some flip-flops that work in the enhanced mode and distributes them into different chains. In the multiple scan chain architecture, to avoid too many scan-in pins, some chains are driven by the same scan-in pin to construct a tree-based architecture. Based on the architecture, the new test application scheme allows some flip-flops working in the enhanced mode, while most of others working in the traditional broadside mode. With the efficient conflict-driven selection method, fault coverage is improved greatly, which can also reduce test application time and compress test data volume. Experimental results show that fault coverage based on the proposed method is comparable to the enhanced scan.

We propose a compact test generation method for transition faults, which is driven by a conflict-avoidance scheme employed during test generation. Based on an influence-cone function for transition faults in broadside scan testing, two dynamic test compaction schemes, named selfish test compaction and unselfish test compaction respectively, are proposed. The selfish test compaction tries to compact as many faults as possible into the current test, while the unselfish scheme attempts to compact the tests of the hard-to-compact faults into the current test. Potential conflicts produced by the signal requirements at the pseudo-primary outputs in the first frame are avoided through the use of an input dependency graph. Experimental results and comparison with existing approaches demonstrate the efficiency and effectiveness of the proposed method.