Emc IP Holding Company LLC

Itay Azaria

Kfir Wolfson

Jehuda Shemer

Saar Cohen

Amihai Savir

Avitan Gefen

Lev Ayzenberg

One example method includes intercepting an IO issued by an application, writing the IO and IO metadata to a splitter journal in NVM, forwarding the IO to storage, and asynchronous with operations occurring along an IO path between the application and storage, evacuating the splitter journal by sending the IO and IO metadata from the splitter journal to a replication site. In this example, sending the IO and IO metadata from the journal to the replication site does not increase a latency associated with the operations on the IO path.

Achieving near-zero added latency for any point in time OS kernel-based application replication

One example method includes intercepting an IO issued by an application of a VM, the IO including IO data and IO metadata, storing the IO data in an IO buffer, writing the IO metadata and a pointer, but not the IO data, to a splitter journal in memory, wherein the pointer points to the IO data in the IO buffer, forwarding the IO to storage, and asynchronous with operations occurring along an IO path between the application and storage, evacuating the splitter journal by sending the IO data and the IO metadata from the splitter journal to a replication site.

Achieving near-zero added latency for modern any point in time VM replication

One example method includes monitoring performance of an element of a runtime environment, where the monitoring includes collecting performance information concerning the element, analyzing the collected information, detecting, based on the analysis of the collected information, an anomaly in the performance of the element and, in response to detection of the anomaly, automatically marking a snapshot of the runtime environment element, and the marking of the snapshot overrides a retention policy applicable to the snapshot.

Anomaly-based retention policies for snapshot-based data protection systems

Systems, apparatus, and methods for any point in time replication to the cloud. Data is replicated by replicating data to a remote storage or a data bucket in the cloud. At the same time, a metadata stream is generated and stored. The metadata stream establishes a relationship between the data and offsets of the data in the production volume. This allows continuous replication without having to maintain a replica volume. The replica volume can be generated during a rehydration operation that uses the metadata stream to construct the production volume from the cloud data.

Any point in time replication to the cloud

Systems and methods for providing any point in time functionality. With a storage system such as a VSAN, any point in time protection is achieved by combining a metadata stream with snapshots of the storage system. This allows snapshots to be generated in hindsight such that any point in time functionality is provided.

Snapshots for any point in time replication

Prioritization and quality of service control operations for any PiT replication operations. When replicating data from a production site to a replica site, the transmission of data is prioritized based on the perspective of a splitter and based on the perspective of an appliance, which may coordinate with multiple splitters. Quality is controlled by prioritizing which data is sent. The transmission may also perform compression when transmitting the data and may handle situations where available resources are scarce or unavailable.

Prioritization and quality-of-service control for near-zero-added-latency any point in time virtual replication

Described is a system that provides deduplication management at a virtualization layer allowing virtual machines to store deduplicated data amongst various back-end data storage architectures. For example, the system may be storage-agnostic by operating at a hypervisor layer to fully leverage the benefits provided by virtualization. Accordingly, implementing deduplication by a hypervisor results in 'pluggable' software-defined storage management. In addition, such a system may be implemented on any on-demand infrastructure (e.g. Infrastructure as a Service (IaaS)) environment that includes varying types of back-end storage solutions including solutions from various vendors. In addition, by operating within the virtualized layer of an on-demand computing resource environment (e.g. cloud environment), the system may provide deduplication across different resource clusters (e.g. hypervisor clusters) and environments. Moreover, the system may operate in a multitenancy environment to support global deduplication amongst multiple tenants, or tenant-specific deduplication.

Data storage agnostic deduplication for virtual machines

A data protection system configured to replicate data may generate rescue packages that allow the system to recover when communication between a splitter or source of the production data being replicated and an appliance that stores the replicated data is disrupted. The rescue package is stored on a datastore and is then retrieved by the data protection system or another splitter. After processing the rescue package, which may contain IOs that the data protection is unaware of due to the communication disruption, replication may resume normally.

Rescue package for uncontrollable splitters

Embodiments for dynamically allocating journal space for Do streams across multiple applications. A shared Do stream process has a dynamic block allocation component that provides a certain amount of buffering of a data flush for an application, using space that would normally be allocated for, but unused by, other applications, thus preventing the need for one or more of the applications to move to fast-forward mode when possible. Certain machine learning techniques are used in order to predict the required Do stream for each application according to past experience with the application, and this prediction is used to intelligently allocate Do Streams between the different applications.

Efficient utilization of storage resources on data recovery sites using machine learning

Predicting large data flushes by collecting usage data for system assets, analyzing the data using machine learning on each asset and the whole system to determine usage trends, predicting a next large data flush using a time-series model, and determining if a size of the predicted next flush size is too large relative to journal storage space in order to advance fast forward mode. Further, protecting history information by pausing distribution of data from journal volumes to replica volumes, taking storage-level snapshots of the replica and the journal volumes, storing a snapshot timestamp for each of the storage-level snapshots in a snapshot database prior to advancing the fast forward mode or un-pausing distribution.

Growing community of inventors

Dimona, Israel

Itay Azaria