Cloud disk provided to users by cloud providers has been a prevalent form of cloud storage. Assigning disk storage space for cloud disks among available data warehouses remains a challenging task, as the load characteristics of cloud disks are not known at the time of disk creation. Methods considering only subscribed capacity could be prone to result in resource under-utilization and load imbalance among warehouses. We propose a Smart Cloud Disk Allocation (S-CDA) approach, which uses clustering and classifying to predict the load information for new cloud disks, and then realizes multi-dimensional allocation based on Manhattan distance. Experimental results with realistic cloud workloads show that, compared with the existing one-dimensional allocation scheme, S-CDA increases the overall space/IOPS/disk bandwidth utilization, while decreasing the load imbalance.

Nowadays, SSD cache plays an important role in cloud storage systems. The associated write policy, which enforces an admission control policy regarding filling data into the cache, has a significant impact on the performance of the cache system and the amount of write traffic to SSD caches. Based on our analysis on a typical cloud block storage system, approximately 47.09% writes are write-only, i.e., writes to the blocks which have not been read during a certain time window. Naively writing the write-only data to the SSD cache unnecessarily introduces a large number of harmful writes to the SSD cache without any contribution to cache performance. On the other hand, it is a challenging task to identify and filter out those write-only data in a real-time manner, especially in a cloud environment running changing and diverse workloads.In this paper, to alleviate the above cache problem, we propose an ML-WP, Machine Learning Based Write Policy, which reduces write traffic to SSDs by avoiding writing write-only data. The main challenge in this approach is to identify write-only data in a real-time manner. To realize ML-WP and achieve accurate write-only data identification, we use machine learning methods to classify data into two groups (i.e., write-only and normal data). Based on this classification, the write-only data is directly written to backend storage without being cached. Experimental results show that, compared with the industry widely deployed write-back policy, ML-WP decreases write traffic to SSD cache by 41.52%, while improving the hit ratio by 2.61% and reducing the average read latency by 37.52%.

Solid State Drives (SSDs) are popularly used for caching in large scale cloud storage systems nowadays. Traditionally, most cache algorithms make replacement upon each miss when cache space is full. However, we observe that in a typical Cloud Block Storage (CBS) system, there is a great percentage of blocks with large reuse distances, which would result in large number of blocks being evicted out of the cache before they ever have a chance to be referenced while they are cached, significantly jeopardizing the cache efficiency. In this article, we propose LEA, Lazy Eviction cache Algorithm, for cloud block storage to efficiently remedy the cache inefficiencies caused by cache blocks with large reuse distances. LEA mainly employs two lists, Lazy Eviction List (LEL) and Block Identity List (BIL), which keep track of two types of victim blocks respectively based on their cache duration when replacements occur, to improve cache efficiency. When a cache miss happens, if the victim block has not resided in cache for longer than its reuse distance, LEA inserts the missed block identity into BIL. Otherwise, it inserts the missed block entry into LEL. We have evaluated LEA by using IO traces collected from Tencent, one of the largest network service providers in the world, and several open source traces. Experimental results show that LEA not only outperforms most of the state-of-the-art cache algorithms in hit ratio, but also greatly reduces the number of SSD writes.

Solid State Drives (SSDs) are popularly used for caching in large scale cloud storage systems nowadays. Traditionally, most cache algorithms make replacement at each miss when cache space is full. However, we observe that in a typical Cloud Block Storage (CBS), there is a great percentage of blocks with large reuse distances, which would result in large number of blocks being evicted out of the cache before they ever have a chance to be referenced while they are cached, significantly jeopardizing the cache efficiency. In this paper, we propose LEA, Lazy Eviction cache Algorithm, for cloud block storage to efficiently remedy the cache inefficiencies caused by cache blocks with large reuse distances. Specifically, LEA uses two lists, Lazy Eviction List (LEL) and Block Identity List (BIL). When a cache miss happens, if the candidate evicted-block has not resided in cache for longer than its reuse distance, LEA inserts the missed block identity into BIL. Otherwise, it inserts the missed block entry into LEL. We have evaluated LEA by using IO traces collected from Tencent, one of the largest network service providers in the world, and several open source traces. Experimental results show that LEA not only outperforms most of the state-of-the-art cache algorithms in hit ratio, but also reduces the number of SSD writes greatly.