-
公开(公告)号:US11481121B2
公开(公告)日:2022-10-25
申请号:US16727537
申请日:2019-12-26
Applicant: Amazon Technologies, Inc.
Inventor: Phillip Peterson , Leonid Baryudin , Daniel Sladic , Sujan Biswas
Abstract: An indirection mapping data structure can maintain a mapping between logical block addresses used by a host computer and physical data storage locations on a solid state drive. Changes to the indirection mapping data structure can be stored in journals. When a journal is full, the journal can be stored to a predetermined location on the cluster block determined based on the number of entries stored by the journal, leading to a number of journals scattered throughout the cluster block at predetermined locations. Each physical chunk of media, whether written with data or marked as defective is journaled. Such a journaling scheme, where the journal locations are predetermined and each physical chunk of media is journaled is referred to as physical media-aware spatially coupled journaling. During replay the spatially coupled journals can be retrieved from the predefined locations within cluster blocks and used to rebuild the indirection mapping data structure.
-
公开(公告)号:US10303397B1
公开(公告)日:2019-05-28
申请号:US15605729
申请日:2017-05-25
Applicant: Amazon Technologies, Inc.
Inventor: Leonid Baryudin
IPC: G06F3/06
Abstract: Management of a nonvolatile medium, such as a flash memory, is disclosed. The nonvolatile medium can be organized as a plurality of storage units. The storage units used to store user data in the form of logical units (sectors). For purposes of read disturb handling, the logical units need to be copied out of the storage unit and thus invalidated prior to hitting a read disturb limit. Rather than waiting until the read disturb limit is reached and copying out all of the logical units at once, the logical units are moved gradually over a period of time based, in part, on a number of valid logical units. A number of reads can also be used to determine when logical units are moved. Distributed movement of the logical units avoids an avalanche effect wherein multiple logical units need to be moved at once, which slows the overall system.
-
公开(公告)号:US11606104B1
公开(公告)日:2023-03-14
申请号:US17545846
申请日:2021-12-08
Applicant: Amazon Technologies, Inc.
Inventor: Avigdor Segal , Leonid Baryudin , Erez Izenberg , Erez Sabbag , Se Wang Oh , Noga Smith
Abstract: The integrity of transmitted data can be protected by causing that data to be transmitted twice, and calculating protection information (PI) for the data from each transmission. The PI can include information such as a checksum or signature that should have the same value if the data from each transmission is the same. If the PI values are not the same, an error handling procedure can be activated, such as may retry the transmission. For write operations, the data can be transmitted twice from a source to a storage destination, while for read operations, the data can be transmitted to a recipient then sent back from the recipient to the storage device, with PI calculated for each transmission. A component such as a storage processor can perform at least this comparison step. Such approaches can also be used for network transmission or high performance computing.
-
公开(公告)号:US10289321B1
公开(公告)日:2019-05-14
申请号:US15588488
申请日:2017-05-05
Applicant: Amazon Technologies, Inc.
Inventor: Leonid Baryudin , Phillip Peterson , Daniel Sladic
IPC: G06F3/06 , G06F9/4401 , G06F13/42 , G06F12/02
Abstract: Spatially coupled journals include information for every portion of physical media, including defective, or “bad”, blocks. Because data cannot be stored to bad blocks, a bad block table is needed before a solid state drive (SSD) can be accessed. Using the information already stored in the journals, the bad block table can be rebuilt following a loss of power. To ensure the journals can be located, a small seed can be stored in off band storage. The seed can include information pointing to a boot catalog stored on the SSD. The boot catalog can be used to determine the locations of journals, which may vary from their predetermined locations depending on the bad blocks of the drive. By storing a small seed, rather than an entire bad block table, the size of external storage needed to maintain the bad block table is reduced.
-
公开(公告)号:US11347413B2
公开(公告)日:2022-05-31
申请号:US17013435
申请日:2020-09-04
Applicant: Amazon Technologies, Inc.
Inventor: Christopher Nathan Watson , Leonid Baryudin , Tyler Huston Doornenbal , Truong Nguyen , Phillip Peterson , Wenzhou Chen , Christopher J. Douglass
IPC: G06F3/06 , G06F9/50 , H04L41/0806 , G06F9/455
Abstract: An opportunistic storage service, or system, identifies currently unused storage capacity on a plurality of physical storage components of computing devices dispersed throughout a provider network. In some embodiments, the currently unused storage capacity is provisioned as primary storage, but is not currently being used to store primary storage data. The opportunistic storage service advertises at least a portion of the currently unused storage capacity as opportunistic storage capacity and provisions the opportunistic storage capacity subject to revocation if additional storage capacity of the physical storage components is needed to store primary storage data to fulfill a primary storage commitment.
-
公开(公告)号:US10387340B1
公开(公告)日:2019-08-20
申请号:US15447546
申请日:2017-03-02
Applicant: Amazon Technologies, Inc.
Inventor: Leonid Baryudin , Wenzhou Chen
Abstract: The following description is directed to managing a nonvolatile medium. The nonvolatile medium can be organized as a plurality of storage units. In one example, a method can include measuring read latencies for the individual storage units of the nonvolatile medium. A probability distribution of future read latencies for the nonvolatile medium can be estimated based on the measured read latencies for the individual storage units of the nonvolatile medium. Information can be moved from a particular storage unit of the nonvolatile medium to a different storage unit of the nonvolatile medium based on the estimated probability distribution of future read latencies for the nonvolatile medium.
-
公开(公告)号:US10768835B1
公开(公告)日:2020-09-08
申请号:US16020755
申请日:2018-06-27
Applicant: Amazon Technologies, Inc.
Inventor: Christopher Nathan Watson , Leonid Baryudin , Tyler Huston Doornenbal , Truong Nguyen , Phillip Peterson , Wenzhou Chen , Christopher J. Douglass
Abstract: An opportunistic storage service, or system, identifies currently unused storage capacity on a plurality of physical storage components of computing devices dispersed throughout a provider network. In some embodiments, the currently unused storage capacity is provisioned as primary storage, but is not currently being used to store primary storage data. The opportunistic storage service advertises at least a portion of the currently unused storage capacity as opportunistic storage capacity and provisions the opportunistic storage capacity subject to revocation if additional storage capacity of the physical storage components is needed to store primary storage data to fulfill a primary storage commitment.
-
公开(公告)号:US10540102B2
公开(公告)日:2020-01-21
申请号:US15282295
申请日:2016-09-30
Applicant: Amazon Technologies, Inc.
Inventor: Phillip Peterson , Leonid Baryudin , Daniel Sladic , Sujan Biswas
Abstract: An indirection mapping data structure can maintain a mapping between logical block addresses used by a host computer and physical data storage locations on a solid state drive. Changes to the indirection mapping data structure can be stored in journals. When a journal is full, the journal can be stored to a predetermined location on the cluster block determined based on the number of entries stored by the journal, leading to a number of journals scattered throughout the cluster block at predetermined locations. Each physical chunk of media, whether written with data or marked as defective is journaled. Such a journaling scheme, where the journal locations are predetermined and each physical chunk of media is journaled is referred to as physical media-aware spatially coupled journaling. During replay the spatially coupled journals can be retrieved from the predefined locations within cluster blocks and used to rebuild the indirection mapping data structure.
-
公开(公告)号:US20180095680A1
公开(公告)日:2018-04-05
申请号:US15282295
申请日:2016-09-30
Applicant: Amazon Technologies, Inc.
Inventor: Phillip Peterson , Leonid Baryudin , Daniel Sladic , Sujan Biswas
Abstract: An indirection mapping data structure can maintain a mapping between logical block addresses used by a host computer and physical data storage locations on a solid state drive. Changes to the indirection mapping data structure can be stored in journals. When a journal is full, the journal can be stored to a predetermined location on the cluster block determined based on the number of entries stored by the journal, leading to a number of journals scattered throughout the cluster block at predetermined locations. Each physical chunk of media, whether written with data or marked as defective is journaled. Such a journaling scheme, where the journal locations are predetermined and each physical chunk of media is journaled is referred to as physical media-aware spatially coupled journaling. During replay the spatially coupled journals can be retrieved from the predefined locations within cluster blocks and used to rebuild the indirection mapping data structure.
-
-
-
-
-
-
-
-