物理分析平臺建設(shè)研究2013年度報告

王萌 李衛(wèi)東 何苗 齊法制 石京燕

摘要：2013年，大亞灣實驗已經(jīng)進入穩(wěn)定取數(shù)階段，物理分析平臺的任務(wù)是針對不斷增加的數(shù)據(jù)量，提高計算能力和存儲能力；根據(jù)物理分析的需求，開發(fā)或改進探測器刻度算法和重建算法，進行數(shù)據(jù)和模擬的比較，并提高物理分析的效率。 在計算環(huán)境方面，我們擴大了CPU和硬盤資源，開發(fā)了計算資源監(jiān)視系統(tǒng)，建立了計算環(huán)境監(jiān)控管理系統(tǒng)。在數(shù)據(jù)存儲方面，本年度建成了一個574TB的磁盤存儲系統(tǒng)。系統(tǒng)基于在高性能計算領(lǐng)域廣泛應(yīng)用的開源Lustre分布式文件系統(tǒng)，完全采用商品服務(wù)器和存儲設(shè)備構(gòu)建。課題組針對大亞灣現(xiàn)場網(wǎng)絡(luò)和計算環(huán)境設(shè)計并實現(xiàn)了大亞灣現(xiàn)場網(wǎng)絡(luò)系統(tǒng)運行支撐平臺。在大亞灣現(xiàn)場和高能所都部署了傳輸系統(tǒng)的數(shù)據(jù)緩沖模塊。并添加了傳輸監(jiān)視系統(tǒng)。 在軟件算法方面，通過對同時穿過不同探測器、實驗廳的繆子事例的觸發(fā)時刻延遲分析，得到了各探測器之間的時間延遲。通過強制觸發(fā)數(shù)據(jù)得到了阻性板室探測器的層噪聲率和模塊噪聲率。開發(fā)了MC Tuning軟件框來管理模擬參數(shù)，通過xml文件和txt文件對不同的參數(shù)進行分組管理。根據(jù)大亞灣物理分析的特點，設(shè)計了新的軟件框架，確定了數(shù)據(jù)模型的定義，并設(shè)計、實現(xiàn)了滿足分析需求的事例緩存機制。

關(guān)鍵詞：大亞灣；計算；存儲；網(wǎng)絡(luò)；軟件框架；刻度；模擬

2013 annual report of the physics analysis platform

Abstract：In 2013， the Daya Bay experiment is taking data stably. The physics analysis platform aims at improving the ability of computing and data storage according to accumulated data volume， detector calibration， event reconstruction， MC tuning， and analysis efficiency. The computing resources and hard disks have been increased. A computing resource monitoring system and a computing environment management system were developed. A 574TB Lustre data storage system was established. According to the Daya Bay network and computing environment， an onsite network support platform was designed and implemented. The data buffer module as well as the monitor system was deployed at both onsite and IHEP for better data transfer. For the offline software， a calibration algorithm was developed using cosmic muon events to get the time latency between different detectors. The layer noise rate and module noise rate of the resistive plate chambers were obtained with the forced trigger events. A MC Tuning software framework was developed to manage simulation parameters. A new lightweight analysis framework was proposed and the data model was defined as well as the event buffer mechanism.

Keywords：Daya Bay； computing； storage； network；software framework； calibration； simulation

閱讀全文鏈接（需實名注冊）：http：//www.nstrs.cn/xiangxiBG.aspx？id=48308&flag=1