25 Countering Fragmentation in an Enterprise Storage System RAM KESAVAN, Google (work done while employed at NetApp, Inc.), USA MATTHEW CURTIS-MAURY, VINAY DEVADAS, and KESARI MISHRA, NetApp, Inc, USA As a fle system ages, it can experience multiple forms of fragmentation. Fragmentation of the free space in the fle system can lower write performance and subsequent read performance. Client operations as well as internal operations, such as deduplication, can fragment the layout of an individual fle, which also impacts fle read performance. File systems that allow sub-block granular addressing can gather intra-block fragmen- tation, which leads to wasted free space. Similarly, wasted space can also occur when a fle system writes a collection of blocks out to object storage as a single large object, because the constituent blocks can become free at diferent times. The impact of fragmentation also depends on the underlying storage media. This arti- cle studies each form of fragmentation in the NetApp ® WAFL ® fle system, and explains how the fle system leverages a storage virtualization layer for defragmentation techniques that physically relocate blocks ef- ciently, including those in read-only snapshots. The article analyzes the efectiveness of these techniques at reducing fragmentation and improving overall performance across various storage media. CCS Concepts: • Information systems → Hierarchical storage management; Information storage tech- nologies; Storage virtualization;• Software and its engineering → File systems management; Additional Key Words and Phrases: Storage system, fle system, fragmentation, fle system performance, snap- shot, deduplication ACM Reference format: Ram Kesavan, Matthew Curtis-Maury, Vinay Devadas, and Kesari Mishra. 2020. Countering Fragmentation in an Enterprise Storage System. ACM Trans. Storage 15, 4, Article 25 (January 2020), 35 pages. https://doi.org/10.1145/3366173 1 INTRODUCTION File systems typically allocate physically contiguous blocks in storage devices to write out logi- cally sequential data and metadata. This strategy maximally uses the write bandwidth available from each storage device, since more blocks can be written to it using fewer write I/Os, and it allows optimal performance when those data or metadata are later read sequentially. Common fle An earlier version of this article [18] appeared in the proceedings of the File and Storage Technologies Conference (FAST’19) in Boston, MA, 2019. Authors’ addresses: R. Kesavan, Google (work done while employed at NetApp, Inc.) 1600 Amphitheatre Pkwy, Moun- tain View, CA 94043; email: ram.kesavan@gmail.com; M. Curtis-Maury and V. Devadas, NetApp, 7301 Kit Creek Rd, Durham, NC, 27709; emails: {mcm, vdevadas}@netapp.com; K. Mishra, 1223 Crescent Terrace, Sunny vale, CA, 94087; email: km@netapp.com. Permission to make digital or hard copies of all or part of this work for personal or classroom use is granted without fee provided that copies are not made or distributed for proft or commercial advantage and that copies bear this notice and the full citation on the frst page. Copyrights for components of this work owned by others than ACM must be honored. Abstracting with credit is permitted. To copy otherwise, or republish, to post on servers or to redistribute to lists, requires prior specifc permission and/or a fee. Request permissions from permissions@acm.org. © 2020 Association for Computing Machinery. 1553-3077/2020/01-ART25 $15.00 https://doi.org/10.1145/3366173 ACM Transactions on Storage, Vol. 15, No. 4, Article 25. Publication date: January 2020.