Web service administrators must ensure the stability of multiple systems by promptly detecting anomalies in Key Performance Indicators (KPIs). Achieving the goal of "train once, infer across scenarios" remains a fundamental challenge for time series anomaly detection models. Beyond improving zero-shot generalization, such models must also flexibly handle sequences of varying lengths during inference, ranging from one hour to one week, without retraining. Conventional approaches rely on sliding-window encoding and self-supervised learning, which restrict inference to fixed-length inputs. Large Language Models (LLMs) have demonstrated remarkable zero-shot capabilities across general domains. However, when applied to time series data, they face inherent limitations due to context length. To address this issue, we propose ViTs, a Vision-Language Model (VLM)-based framework that converts time series curves into visual representations. By rescaling time series images, temporal dependencies are preserved while maintaining a consistent input size, thereby enabling efficient processing of arbitrarily long sequences without context constraints. Training VLMs for this purpose introduces unique challenges, primarily due to the scarcity of aligned time series image-text data. To overcome this, we employ an evolutionary algorithm to automatically generate thousands of high-quality image-text pairs and design a three-stage training pipeline consisting of: (1) time series knowledge injection, (2) anomaly detection enhancement, and (3) anomaly reasoning refinement. Extensive experiments demonstrate that ViTs substantially enhance the ability of VLMs to understand and detect anomalies in time series data. All datasets and code will be publicly released at: https://anonymous.4open.science/r/ViTs-C484/.

Persistent key-value (KV) stores are widely used by cloud services at ByteDance as local storage engines, and RocksDB used to be the de facto implementation since it can be tailored to a variety of workloads and requirements. In this paper, we provide key insights into local storage engine usage at ByteDance, explain why the combination of highly write-intensive workloads and stringent requirements on cost efficiency and point lookup tail latency may pose challenges to a general-purpose local storage engine such as RocksDB, and present the design and implementation of LavaStore , a high-performance cost-effective local storage engine purpose-built to address these challenges. LavaStore achieves its design goals by selectively customizing a few components of a RocksDB-based, general-purpose local storage engine, including a distinct KV separation design that decouples garbage collection from compaction, a specialized engine type for the commonly recurring Write-Ahead-Logging workload, and a customized user-space append-only filesystem. LavaStore has been deployed to production with hundreds of thousands of running instances, storing more than 100 PB of data and serving billions of requests per second, bringing significant performance improvements and cost reductions to customers over their original local storage engines. For example, a ByteDance proprietary distributed OLTP database service has experienced a reduction in average write and read latency by 61% and 16%, respectively, and a ByteDance proprietary caching service has gained an 87% increase in write throughput with no more than 6% space overhead.

With the escalation in the demand for privacy-preserving and tamper-resistant data management and processing on the public cloud, an increasing number of mainstream databases start to provide always-encrypted and blockchain-like features, including Microsoft SQL Server, MongoDB, and Alibaba PolarDB. The recent progress in Trusted Execution Environment (TEE) technology has enabled the deployment of the complete database engine within TEE. This implementation ensures that data stored in memory, cache, and registers is encrypted, thereby maintaining the confidentiality of information. In this paper, we present SecuDB, a multi-granularity privacy-preserving and tamper-resistant relational database by placing the entire RDBMS in Intel TDX. We propose a novel visibility control mechanism incorporating column masking, log masking, and statistics masking to realize fine-grained privacy preservation and devise an isolated TEE-endorsed temporal table method to support efficient data and query verifiability, without affecting insertion and selection performance. We evaluate SecuDB using Sysbench, TPC-C and TikTok copyright workloads. The results show that compared with a system without an enclave, SecuDB hits 84.7% and 94.7% of the performance when providing coarse-grained and fine-grained privacy preservation, respectively. While the overhead for tamper-resistance is less than 22.6%.

Blockchain-like ledger databases emerge in recent years as a more efficient alternative to permissioned blockchains. Conventional ledger databases mostly rely on authenticated structures such as the Merkle tree and transparency logs for supporting auditability, and hence they suffer from the performance problem. As opposed to conventional ledger DBMSes, we design VeDB - a high-performance verifiable software (Ve-S) and hardware (Ve-H) enabled DBMS with rigorous auditability for better user options and broad applications. In Ve-S, we devise a novel verifiable Shrubs array (VSA) with two-layer ordinals (serial numbers) which outperforms conventional Merkle tree-based models due to lower CPU and I/O cost. It enables rigorous auditability through its efficient credible timestamp range authentication method, and fine-grained data verification at the client side, which are lacking in state-of-the-art relational ledger databases. In Ve-H, we devise a non-intrusive trusted affiliation by TEE leveraging digest signing, monotonic counters, and trusted timestamps in VeDB, which supports both data notarization and lineage applications. The experimental results show that VeDB-VSA outperforms Merkle tree-based authenticated data structures (ADS) up to 70× and 3.7× for insertion and verification; and VeDB Ve-H data lineage verification is 8.5× faster than Ve-S.

In recent years, at ByteDance, we see more and more business scenarios that require performing complex analysis over freshly imported data, together with transaction support and strong data consistency. In this paper, we describe our journey of building ByteHTAP, an HTAP system with high data freshness and strong data consistency. It adopts a separate-engine and shared-storage architecture. Its modular system design fully utilizes an existing ByteDance's OLTP system and an open source OLAP system. This choice saves us a lot of resources and development time and allows easy future extensions such as replacing the query processing engine with other alternatives. ByteHTAP can provide high data freshness with less than one second delay, which enables many new business opportunities for our customers. Customers can also configure different data freshness thresholds based on their business needs. ByteHTAP also provides strong data consistency through global timestamps across its OLTP and OLAP system, which greatly relieves application developers from handling complex data consistency issues by themselves. In addition, we introduce some important performance optimizations to ByteHTAP, such as pushing computations to the storage layer and using delete bitmaps to efficiently handle deletes. Lastly, we will share our lessons and best practices in developing and running ByteHTAP in production.

ByteDance offers several massively popular products such as TikTok, Jinri Toutiao and Douyin for creating, sharing and discovering a variety of content, in which recommendation plays an indispensable role for helping billions of users to interact with highly personalized content. The personalized experience in products largely comes from the ability of sophisticated machine learning models to make accurate predictions based on users' interests and one key component in such systems is the user profile service.In this paper, we introduce Instance Profile Service (IPS), a large scale distributed system for managing unstructured profile data as well as serving various feature computations at ByteDance. Different products leverage IPS in many different ways and place various demands on the system, in terms of complex computation logic and latency requirements. One major challenge in the design of a large scale user profile system is how to strike the right balance among efficiency, scalability, reliability and versatility. With deliberated choices made on its design and implementation, we demonstrate IPS can provide a simple yet flexible solution to all these products while meeting the targeted high availability and performance goals. At ByteDance, IPS has successfully replaced many legacy profile systems and runs on thousands of machines. One of our largest production instances can process a hundred million feature queries and tens of millions writes per second.