Debugging of Performance Degradation in Distributed Requests Handling Using Multilevel Trace Analysis

Root cause identification of performance degradation within distributed systems is often a difficult and time-consuming task, yet it is crucial for maintaining high performance. In this paper, we present an execution trace-driven solution that reduces the efforts required to investigate, debug, and solve performance problems found in multinode distributed systems. The proposed approach employs a unified analysis method to represent trace data collected from the user-space level to the hardware level of involved nodes, allowing for efficient and effective root cause analysis. This solution works by extracting performance metrics and state information from trace data collected at user-space, kernel, and network levels. The multisource trace data is then synchronized and structured in a multidimensional data store, which is designed specifically for this kind of data. A posteriori analysis using a top-down approach is then used to investigate performance problems and detect their root causes. In this paper, we apply this generic framework to analyze trace data collected from the execution of the web server, database server, and application servers in a distributed LAMP (Linux, Apache, MySQL, and PHP) Stack. Using industrial level use cases, we show that the proposed approach is capable of investigating the root cause of performance issues, addressing unusual latency, and improving base latency by 70%. This is achieved with minimal tracing overhead that does not significantly impact performance, as well as O log n query response times for efficient analysis.

Lustre I/O performance investigations on Hazel Hen: experiments and heuristics

With ever-increasing computational power, larger computational domains are employed and thus the data output grows as well. Writing this data to disk can become a significant part of runtime if done serially. Even if the output is done in parallel, e.g., via MPI I/O, there are many user-space parameters for tuning the performance. This paper focuses on the available parameters for the Lustre file system and the Cray MPICH implementation of MPI I/O. Experiments on the Cray XC40 Hazel Hen using a Cray Sonexion 2000 Lustre file system were conducted. In the experiments, the core count, the block size and the striping configuration were varied. Based on these parameters, heuristics for striping configuration in terms of core count and block size were determined, yielding up to a 32-fold improvement in write rate compared to the default. This corresponds to 85 GB/s of the peak bandwidth of 202.5 GB/s. The heuristics are shown to be applicable to a small test program as well as a complex application.

An online service for analysing ozone trends within EOSC-synergy

<p>The European Open Science Cloud-Synergy (EOSC-Synergy) project delivers services that serve to expand the use of EOSC. One of these services, O3as, is being developed for scientists using chemistry-climate models to determine time series and eventually ozone trends for potential use in the quadrennial Global Assessment of Ozone Depletion, which will be published in 2022. A unified approach from a service like ours, which analyses results from a large number of different climate models, helps to harmonise the calculation of ozone trends efficiently and consistently. With O3as, publication-quality figures can be reproduced quickly and in a coherent way. This is done via a web application where users configure their queries to perform simple analyses. These queries are passed to the O3as service via an O3as REST API call. There, the O3as service processes the query and accesses the reduced dataset. To create a reduced dataset, regular tasks are executed on a high performance computer (HPC) to copy the primary data and perform data preparation (e.g. data reduction, standardisation and parameter unification). O3as uses EGI check-in (OIDC) to identify users and grant access to certain functionalities of the service, udocker (a tool to run Docker containers in multi-user space without root privileges) to perform data reduction in the HPC environment, and the Universitat Politècnica de València (UPV)  Infrastructure Manager to provision service resources (Kubernetes).</p>

Kreon

Persistent key-value stores have emerged as a main component in the data access path of modern data processing systems. However, they exhibit high CPU and I/O overhead. Nowadays, due to power limitations, it is important to reduce CPU overheads for data processing. In this article, we propose Kreon , a key-value store that targets servers with flash-based storage, where CPU overhead and I/O amplification are more significant bottlenecks compared to I/O randomness. We first observe that two significant sources of overhead in key-value stores are: (a) The use of compaction in Log-Structured Merge-Trees (LSM-Tree) that constantly perform merging and sorting of large data segments and (b) the use of an I/O cache to access devices, which incurs overhead even for data that reside in memory. To avoid these, Kreon performs data movement from level to level by using partial reorganization instead of full data reorganization via the use of a full index per-level. Kreon uses memory-mapped I/O via a custom kernel path to avoid a user-space cache. For a large dataset, Kreon reduces CPU cycles/op by up to 5.8×, reduces I/O amplification for inserts by up to 4.61×, and increases insert ops/s by up to 5.3×, compared to RocksDB.