Vmware, Inc.

Pascal Renauld

Eric Shannon Knauft

Enning Xiang

Wenguang Wang

Yiqi Xu

Christos Karamanolis

William John Earl

Xin Liang Li

Jin Zhang

Vamsi Gunturu

Mansi Shah

Pratik Shridhar Desai

Yizhou Luo

Sandeep Rangaswamy

Christian Dickmann

Haoran Zheng

Radhika Vullikanti

Ruocheng Li

Bryan Fink

Abhishek Gupta

Blake H Lewis

Junlong Gao

Jorge Guerra Delgado

Srinath Premachandran

Satish Pudi

Radu Berinde

Vamsidhar Gunturu

Luke Lu

Ojan Thornycroft

Zhihao Yao

Kevin Rayfeng Li

Gregory A Smith

Matt Amdur

Venkata Ramanan

Peng Mi

Quanxing Liu

The disclosure herein describes performing resynchronization ('resync') jobs in a distributed storage system based on a parallelism policy. A resync job is obtained from a queue and input/output (I/O) resources that will be used during execution of the resync job are identified. Available bandwidth slots of each I/O resource of the identified I/O resources are determined. The parallelism policy is applied to the identified I/O resources and the available bandwidth slots. Based on the application of the parallelism policy, a bottleneck resource of the I/O resources is determined and a parallel I/O value is calculated based on the available bandwidth slots of the bottleneck resource, wherein the parallel I/O value indicates a quantity of I/O tasks that can be performed in parallel. The resync job is executed using the I/O resources, the execution of the resync job including performance of I/O tasks in parallel based on the parallel I/O value.

Performing resynchronization jobs in a distributed storage system based on a parallelism policy

The disclosure herein describes enhancing data durability of a base component using a delta component. A delta component is generated based on the base component becoming unavailable. The delta component is configured to include unwritten storage space with an address space matching the base component and a tracking bitmap associated with data blocks of the address space of the delta component. Write operations targeted for the base component are routed to the delta component. Based on the routed write operations, bits associated with data blocks affected by the write operations are changed in the tracking bitmap. Based on the base component becoming available, data blocks affected by routed write operations are identified based on the tracking bitmap and the identified data blocks are synchronized from the delta component to the base component. The delta component is then removed.

Synchronizing a stale component of a distributed object using a delta component during maintenance

System and method for managing distributed storage objects for host unavailability in a distributed storage system uses at least one of a crash indicator in a specific on-disk block and a paused object indicator for a distributed storage object to determine whether to perform data recovery for the distributed storage object. When the crash indicator is set or the paused object indicator implies that the distributed storage object is a paused object, the distributed storage object is left as a paused object without perform the data recovery for the distributed storage object. When the crash indicator is unset and the paused object indicator implies that the distributed storage object is not a paused object, the data recovery for the distributed storage object is performed.

Limiting blast radius for critical on-disk data corruptions in a distributed storage system

The present disclosure is related to methods, systems, and machine-readable media for log-structured file system management operations. An aggregate amount of over-provisioned computing resources assigned to a plurality of log-structured file system (LFS) objects that are allocated for performance of memory management operations associated with a virtual storage area network (vSAN) can be determined. A subset of LFS objects that are candidates for performance of a particular memory management operation based on an amount of over-provisioned computing resources consumed by one or more LFS objects among the plurality of LFS objects exceeding a resource consumption threshold associated with the determined amount of over-provisioned computing resources assigned to the plurality of LFS objects can be selected. The particular memory management operation using one or more of the candidate LFS objects can be performed.

Dynamic over-provisioning of space in a log-structured file system

Embodiments of the disclosure provide techniques for partitioning a resource object into multiple resource components of a cluster of host computer nodes in a distributed resources system. The distributed resources system translates high-level policy requirements into a resource configuration that the system accommodates. The system determines an allocation based on the policy requirements and identifies resource configurations that are available. Upon selecting a resource configuration, the distributed resources system assigns the allocation and associated values to the selected configuration and publishes the new configuration to other host computer nodes in the cluster.

Translating high level requirements policies to distributed storage configurations

A method for compressing is provided. The method including receiving a block of data to store on at least one physical disk; determining whether to store the data in a data log as uncompressed or compressed data based on a determined size of resulting compressed data. When the method determines to store the data as compressed, compressing the data and storing the compressed data in at least one sector in the data log. Otherwise, the method stores the data, uncompressed, in a plurality of sectors in the data log. The method generates a one or more state bits indicating (i) whether the data is stored as uncompressed or compressed, and (ii) if the data is stored as compressed, a size of the compressed data. The method then stores the one or more state bits in an entry of a logical map table associated with an LBA that corresponds to the data block.

Data compression in a multi-layer distributed datastore

Hybrid synchronization using a shadow component includes detecting a first component of a plurality of mirrored components of a distributed data object becoming unavailable. The mirrored components include a delta component (a special shadow component) and a regular mirror (shadow) component. The delta component indicates a shorter history of changes to data blocks of a log-structured file system (LFS) than is indicated by the regular mirror component. During the unavailability of the first component, at least one write I/O is committed by the delta component. The commit is tracked by the delta component in a first tracking bitmap associated with the delta component. Based at least on detecting the first component becoming available, the first component is synchronized with data from the delta component, based at least on changed data blocks indicated in the first tracking bitmap.

Hybrid synchronization using a shadow component

The disclosure herein describes tracking changes to a stale component using a synchronization bitmap. A first component of a plurality of mirrored components of the distributed data object becomes available from an unavailable state, and a stale log sequence number (LSN) and a last committed LSN are identified. A synchronization bitmap of the first component associated with a range of LSNs (e.g., from the stale LSN to the last committed LSN) is created and configured to track changes to data blocks of the first component. A second component is identified based on the second component including a tracking bitmap associated with an LSN that matches the stale LSN of the first component. The first component is synchronized with data from the second component based on, wherein the synchronizing includes updating the synchronization bitmap to track changes made to data blocks of the first component.

Synchronizing changes to stale components of a distributed object using synchronization bitmaps

A method for encrypting data in one or more data blocks is provided. The method receives a first data block to be written to a physical storage that includes one or more physical disks. The method applies a first random tweak to data indicative of the first data block to generate a first encrypted data block, and writes the first encrypted data block and the first random tweak to a first physical block of the physical storage. The method receives a second data block to be written to the physical storage. The method then applies a second random tweak, different than the first random tweak, to data indicative of the second data block to generate a second encrypted data block, and writes the second encrypted data block and the second random tweak to a second physical block of the physical storage.

Enhanced data encryption in distributed datastores using random tweaks stored in data blocks

Techniques for efficiently storing client data blocks on a distributed-computing system are provided. The system includes a fast performance tier and a large capacity tier. The capacity tier stores the client data blocks in erasure encoded data stripes. The performance tier stores logical map data including an address map indicating a correspondence between logical addresses associated with a first layer of the system and physical addresses associated with a second layer. A method includes receiving a request to include additional client data blocks in the client blocks. The request indicates logical addresses for additional blocks. Corresponding physical addresses for additional block are determined. Each additional block is stored at the physical address. Additional logical map data is stored in the performance tier. Storing the additional logical map data includes updating the address map to indicate the correspondence between the logical addresses and the physical addresses for the additional blocks.

Efficient erasure-coded storage in distributed data systems

Techniques for issuing efficient writes to an erasure coded storage object in a distributed storage system via adaptive logging are provided. In one set of embodiments, a node of the system can receive a write request for updating one or more logical data blocks of the storage object and determine whether a size of the one or more logical data blocks meets or exceeds a threshold size. Upon determining that the size of the one or more logical data blocks meets or exceeds the threshold size, the node can allocate a segment in a capacity object of the storage object, write the one or more logical data blocks via a full stripe write to the segment, and write metadata for the one or more logical data blocks to a log record in a log of a metadata object of the storage object. The metadata written to the log record can include mappings between logical block addresses (LBAs) of the one or more logical data blocks and physical block addresses (PBAs) where the one or more logical data blocks reside in the segment.

Issuing efficient writes to erasure coded objects in a distributed storage system via adaptive logging

Techniques for enabling deduplication-aware load balancing in a distributed storage system are provided. In one set of embodiments, a node of the distributed storage system can receive an I/O (Input/Output) request pertaining to a data block of a storage object stored on a local storage component of the node. The node can further determine whether the I/O request requires insertion of a new entry into a deduplication hash table associated with the local storage component or deletion of an existing entry from the deduplication hash table. If the I/O request requires insertion of a new hash table entry, the node can add an identifier of the data block into a probabilistic data structure associated with the local storage component, where the probabilistic data structure is configured to maintain information regarding distinct data blocks that are likely present in the local storage component. Alternatively, if the I/O request requires deletion of an existing hash table entry, the node can remove the identifier of the data block from the probabilistic data structure.

Deduplication-aware load balancing in distributed storage systems

An efficient scheduling of IOs in a computing system using dynamic bandwidth regulation includes building up a shared regulator to limit the total IOPS scheduling among all IO classes at any given time. Reserved regulators may be used to place limits on the IOPS scheduled for a particular IO class at any given time. An outstanding IO window may also limit the overall number of outstanding IOs, and/or the bytes of outstanding IOs at any particular time. A first stage of IO scheduling may involve enforcing the reserved regulators to limit the IOPS scheduled for particular IO classes. A second stage of IO scheduling may involve enforcing the shared regulator to limit the total IOPS scheduled for all IO classes.

DRR-based two stages IO scheduling algorithm for storage system with dynamic bandwidth regulation

Methods, systems, and apparatus, including computer programs encoded on computer storage media, for resynchronizing data in a storage system. One of the methods includes determining that a particular primary disk of a capacity object of a storage system has failed, wherein the capacity comprises a plurality of segments, and wherein the each segment comprises: a plurality of primary columns each corresponding to a respective primary disk of the capacity object, and a plurality of parity columns each corresponding to a respective parity disk of the capacity object; and resynchronizing, for each segment of one or more segments of the capacity object, the primary column of the segment corresponding to the particular primary disk using i) the primary columns of the segment corresponding to each other primary disk of the capacity object, ii) one or more parity columns of the segment, and iii) the column summaries of the segment.

Growing community of inventors

Palo Alto, CA, United States of America

Pascal Renauld