Microsoft Technology Licensing, LLC

Amazon Technologies, Inc.

Other

Ernest S Cohen

Matthew Douglas Hendel

Landy Wang

Rene Antonio Vega

John Te-Jui Sheu

Sharvil A Nanavati

Arkady Michael Degtiarov

Srinivasa Rao Vempati

Sameer Choudhary

Nikhil Shah

Vishwas Narendra

James Zuber

Michael F Brown

Vandana Rungta

Benjamin Scott Dow

Abhishek Kannan

Arvinth Ravi

Phillip H Pruett, Iv

Bryan T Wilkinson

John Pender

Preetham Kowshik

RaghuKishore Balivada

Harishkumar Katagal

Wilson Li

Mehak Mehta

Addison Joseph Burns

Aspects related to a resource-constrained system are described herein that can provide object storage services after a service interruption is resolved, even if all of the transactions that were pending and incomplete prior to the service interruption have not yet been recovered and/or executed. For example, file systems implemented by computing systems of the resource-constrained system may treat each file or directory as a separate object. Thus, a transaction directed to one file may not affect the file's directory or other files in the directory. As a result, the resource-constrained system can achieve read-after-write consistency without first recovering and executing the pending, incomplete transactions. Instead, read-after-write consistency for an object can be achieved simply by completing any pending, incomplete transaction directed to that object. Accordingly, the resource-constrained system can provide object storage services to user devices immediately after the service interruption is resolved, thereby resulting in fast crash recovery times.

Object transaction system for a resource-constrained system

Systems and methods are provided for efficiently maintaining a transaction sequence witness service to ensure that requests to read data provide consistent results across all storage nodes in a distributed system. Each storage node that stores data in response to a particular update event may first update the transaction sequence witness service with the sequence number for the update event. Thus, any other storage node that services a subsequent request for the data will be able to see, via the transaction sequence witness service, the sequence number that must be available on the storage node to respond to the request.

Data event management for monotonic read consistency in a distributed storage system

Systems and methods for implementing lockless, leaderless transaction protocols for scalable applications such as distributed data stores are disclosed. Applications may be implemented using multiple nodes individually implementing all or a subset of lockless, leaderless protocols. A transaction, such as a write to a data object, may be received at a node, the operation including a user-provided key. A composite key may be generated using a monotonic sequence number, the composite key globally unique across the application. A new data object is created based on data provided and once the object is created, the object is immutable. Other operations are implemented such that no data object may be written, modified, or deleted by more than one operation, thus eliminating contention and allowing for lock-free operation. Leaderless operation, as well as replication using erasure coding and redundant data paths, further minimize latency of the protocol.

Lockless, leaderless protocol for scalable applications

Multi-action background process for a resource-constrained system

Systems and methods are provided for managing the order of data written to a transaction log in a distributed storage system. In a system with multiple nodes, if sequencers are naively generated without taking into consideration inconsistencies among the different nodes generating the sequencers, then the sequencers may not increase for each data transaction. To alleviate this problem, the node committing the transaction to the transaction log may perform a consistent read and verifies that the sequencer advances. If the sequencer does not advance, the node can perform a context-dependent operation such as adjusting the sequencer, acknowledging the write without committing it to the transaction log, or rejecting the write altogether.

Transaction management for monotonic write consistency in a distributed storage system

Various mechanisms are disclosed for improving the operational efficiency of a virtual translation look-aside buffer (TLB) in a virtual machine environment. For example, one mechanism fills in entries in a shadow page table (SPT) and additionally, speculatively fills in other entries in the SPT based on various heuristics. Another mechanism allows virtual TLBs (translation look-aside buffers) to cache partial walks in a guest page table tree. Still another mechanism allows for dynamic resizing of the virtual TLB to optimize for run-time characteristics of active workloads. Still another mechanism allows virtual machine monitors (VMMs) to support legacy and enlightened modes of virtual TLB operation. Finally, another mechanism allows the VMM to remove only the stale entries in SPTs when linking or switching address spaces. All these mechanisms, together or in part, increase the operational efficiency of the virtual TLB.

Operational efficiency of virtual TLBs

Various operations are disclosed for improving the operational efficiency of a virtual translation look-aside buffer (TLB) in a virtual machine environment. For example, operations are disclosed that allow for determination of whether present entries in shadow page tables (SPTs) are stale by comparing shadowed guest page table (GPT) entries against snapshots taken when the entries were cached. Other operations are disclosed that allow a virtual machine monitor (VMM) to access shadow page table trees (SPTTs) by walking trees in software or in hardware. Still other operations are disclosed allowing the VMM to use a hash table to relate GVA ranges to SPTs that map them, thus significantly reducing the cost of having to walk each SPTT in order to invalidate desired GVA(s). And, finally, operations are disclosed allowing the VMM to determine global GVA ranges by checking a bitmap, when invalidating global GVAs.

Reduction of operational costs of virtual TLBs

Model-based testing is performed by repeatedly constructing a test strategy in which each test stimulus will lead to increased test coverage regardless of the nondeterministic choices made by the system under test, and following said strategy until coverage is increased. As soon as no such strategy exists, testing stops.

Pessimistic model-based testing

Various operations are provided that improve the scalability of virtual TLBs in multi-processor virtual machines, and they include: implicitly locking SPTs using per-processor generation counters; waiting for pending fills on other virtual processors to complete before servicing a GVA invalidation using the counters; write-protecting or unmaping guest pages in a deferred two-stage process or reclaiming SPTs in a deferred two-stage process; periodically coalescing two SPTs that shadow the same GPT with the same attributes; sharing SPTs between two SASes only at a specified level in a SPTT; flushing the entire virtual TLB using a generation counter; allocating a SPT to GPT from a NUMA node on which the GPT resides; having an instance for each NUMA node on which a virtual machine runs; and, correctly handling the serializing instructions executed by a guest in a virtual machine with more than one virtual processor sharing the virtual TLB.

Scalability of virtual TLBs for multi-processor virtual machines

A method of replaying a recorded sequence of execution steps of a computer program or system to a selected target step while taking a minimal number of execution breaks includes calculating, for each of the steps, a predecessor step that can be reached in a minimal number of execution breaks. The total calculation time for the entire execution is linear in the number of steps, and allows subsequent calculation of an optimal path to any selected step to be obtained in time linear in the length of the optimal path.

Optimal program execution replay and breakpoints

Tagged translation lookaside buffer consistency is enabled in the presence of a hypervisor of a virtual machine computing environment, in which multiple processes of multiple logical processors of guests are hosted by a virtual machine monitor or hypervisor component. The virtual machine monitor or hypervisor component maintains tagged TLB data associated with the plurality of processes on behalf of each of the plurality of logical processors, thereby ensuring consistency of the tagged TLB data across all of the plurality of processes.

Tagged translation lookaside buffers in a hypervisor computing environment

In a computer system that employs virtual memory, multiple versions of a given page are stored: a directory version, a table version, and a data version. The data version contains the data that a software object believes to be stored in the page. The directory and table versions of the page contains versions of the page's contents that have been modified in some manner to comply with a restriction on the address translation map employed by the virtual address system. When a page is being used by the virtual address system as a directory or table, then the directory or table versions, respectively, of that page are used. When a page is the target of a read request, the data version of the page is used.

Shadow page tables for address translation control

Address translation control (ATC) limits the mappings between virtual and physical addresses in order to implement a memory access policy. Each processor in a multi-processor system maintains a translation lookaside buffer (TLB) that caches mappings to speed translation of virtual addresses. Each processor also maintains a counter. Each time a processor's TLB is flushed, the processor's counter is incremented. When a link to a page is removed from an address translation map, the counter values for all of the processors are recorded. When that page is accessed by a processor, the recorded counter values are compared with the processor's current counter value to determine whether the processor's TLB has been flushed since the link to the page was removed from the map. An expensive TLB flush operation is delayed until needed, but still occurs early enough to prevent an invalid TLB entry from being used to violate the access policy.

Growing community of inventors

Wyncote, PA, United States of America

Ernest S Cohen