本文選題：RAID + 存儲(chǔ)容災(zāi)��； 參考：《工程科學(xué)與技術(shù)》2017年03期

【摘要】：針對(duì)RAID存儲(chǔ)容災(zāi)系統(tǒng)中數(shù)據(jù)存儲(chǔ)的可靠性和擴(kuò)展性等問(wèn)題,提出一種具有較高容災(zāi)能力且易擴(kuò)展的存儲(chǔ)容災(zāi)方法,稱之為隨機(jī)陣列碼。通過(guò)研究GF(2)上隨機(jī)矩陣列滿秩的性質(zhì),并將其應(yīng)用在RAID存儲(chǔ)容災(zāi)方案中。首先,依據(jù)RAID存儲(chǔ)系統(tǒng)的環(huán)境配置和容災(zāi)需求設(shè)置條帶參數(shù);其次,構(gòu)建相應(yīng)規(guī)模且滿足特定性質(zhì)的隨機(jī)矩陣作為編碼矩陣;最后,將原始數(shù)據(jù)等分成塊,利用編碼矩陣將其編碼并折疊存儲(chǔ)到不同磁盤(pán)上。當(dāng)發(fā)生磁盤(pán)損毀、扇區(qū)失效等原因造成數(shù)據(jù)丟失時(shí),可依據(jù)相應(yīng)的校驗(yàn)矩陣及剩余的編碼分塊進(jìn)行失效數(shù)據(jù)的高概率譯碼恢復(fù),從而,實(shí)現(xiàn)了數(shù)據(jù)高效、可靠地容災(zāi)存儲(chǔ)。實(shí)驗(yàn)驗(yàn)證及理論分析表明:1)GF(2)上的隨機(jī)高矩陣,在隨機(jī)概率p=0.5,矩陣行列差δ≥15時(shí),即具有高概率列滿秩的性質(zhì);2)隨機(jī)陣列碼的編碼參數(shù),不再受到素?cái)?shù)或有限域規(guī)模的限制,可靈活設(shè)置,其容災(zāi)能力也可根據(jù)容災(zāi)需求進(jìn)行擴(kuò)展,并可實(shí)現(xiàn)較多的容錯(cuò)模式;3)隨機(jī)陣列碼由于基于XOR運(yùn)算,在均勻隨機(jī)時(shí)與RS碼、CRS碼相比,具有較高的編譯碼速率,特別是在較大規(guī)模的編碼構(gòu)造中表現(xiàn)良好;4)隨機(jī)陣列碼隨著規(guī)模的增長(zhǎng),可趨于近似MDS碼,具有較高的存儲(chǔ)空間利用率�；陔S機(jī)陣列碼高效,可靠,易擴(kuò)展等特點(diǎn),可實(shí)現(xiàn)一般化RAID存儲(chǔ)容災(zāi)方案的構(gòu)造,此外,也可與其他存儲(chǔ)容災(zāi)技術(shù)結(jié)合使用,共同構(gòu)建特定需求下的RAID存儲(chǔ)容災(zāi)系統(tǒng)。
[Abstract]:Aiming at the problems of reliability and expansibility of data storage in raid storage disaster recovery system, a storage disaster recovery method with high disaster tolerance capability and easy expansion is proposed, which is called random array code. By studying the property of full rank of random matrix on GF (2), we apply it to raid storage and disaster recovery scheme. Firstly, according to the environment configuration and disaster tolerance requirements of raid storage system, the strip parameters are set up; secondly, the random matrix with corresponding scale and satisfying certain properties is constructed as the coding matrix; finally, the raw data is divided into blocks. The encoding matrix is used to code and fold it to different disks. When the data is lost due to the damage of the disk and the failure of the sector, the high probability decoding and recovery of the failure data can be carried out according to the corresponding check matrix and the remaining coding blocks, so that the data can be stored efficiently and reliably. Experimental verification and theoretical analysis show that the encoding parameters of random high matrix on 1: 1) GF (2) are no longer restricted by prime number or finite field size in random probability p0. 5, matrix row difference 未 鈮，

本文編號(hào)：2101172