Abstract: Systems and methods are provided for an asynchronous data replication system in which the remote replication reduces bandwidth requirements by copying deduplicated differences in business data from a local storage site to a remote, backup storage site, the system comprising: a local performance storage pool for storing data; a local deduplicating storage pool for storing deduplicated data, said local deduplicating storage pool further storing metadata about data objects in the system and which has metadata analysis logic for identifying and specifying differences in a data object over time; a remote performance storage pool for storing a copy of said data, available for immediate use as a backup copy of said data to provide business continuity to said data; a remote deduplicating storage pool for storing deduplicated data; and a controller for synchronizing the remote performance storage pool to have the second version of the data object using deduplicated data.

Abstract: Various embodiments of systems and methods for synchronizing a primary repository and a secondary repository in a business infrastructure are disclosed. Content of artifacts present in the primary repository and the secondary repository are indexed. Based upon a search criteria received on the secondary repository, an integrated search service provided on the secondary repository is invoked to perform a combinatorial query on the indexed content and retrieve an identifier of each artifact corresponding to the search criteria. Based upon relevancy parameters, the retrieved identifiers are ranked and rendered as a categorized table including a first column and a second column for identifiers present in the primary repository and the second repository respectively. Based upon a user action received on the ranked identifiers in the first column of the categorized table, the associated artifacts are determined and downloaded from the primary repository to the secondary repository.

Abstract: A method, system, and program product for data replication comprising, building a virtual image of a point in time in a Data Protection Appliance (DPA), taking a storage array based snapshot of a current image, creating a hot pull session and exposing a LUN containing the point in time image at the DPA.

Abstract: Embodiments of the present invention provide a method, system and computer program product for consistent replication of transactional updates. In an embodiment of the invention, a method for consistent replication of data in a transaction processing system is provided. The method includes recording entries in a replication log of different data updates and corresponding transactions and additionally recording entries in the replication log indicating whether or not the transactions have been backed out. The method also includes replicating only those data updates referenced in the log which do not correspond to transactions indicated in the log to have been backed out. For instance the additionally recorded entries in the replication log indicate when a transaction has been backed out. Alternatively, the additionally recorded entries in the replication log indicate when a transaction has been committed.

Abstract: Replicating modifications of a directory that include receiving in a source directory service for a source directory requests for modifications in the source directory and transmitting, from the source directory service to a duplicate directory service for a duplicate directory asynchronously in parallel over a plurality of data communications connections, requests for the same modifications in the duplicate directory. In typical embodiments, a duplicate directory may replicate a subtree of a source directory, receiving requests for modifications may include receiving in the source directory service requests for modifications in the subtree in the source directory, and transmitting requests may include transmitting requests for the same modifications in the subtree in the duplicate directory.

Abstract: Method, system, and programs for resolving data inconsistency in a distributed system having a plurality of replica instances. In response to a read request of at least one data element, a key-specific probability value for the at least one data element of the read request is retrieved. Based on the retrieved key-specific probability value and a system load of the distributed system, a self-healing probability of the at least one data element of the read request is computed. A self-healing transaction that includes the at least one data element of the read request is then generated based on the self-healing probability of the at least one data element. The generated self-healing transaction is further applied to at least some of the plurality of replica instances in the distributed system to resolve data inconsistency.

Abstract: A database server receives a request from a client application for performing a data transaction on persistent data storage. The request is sent to a set of replication servers. An acknowledgement for the request is received from each replication server, including a start sequence number and an end sequence number for data that is stored in local cache of the replication server, and a latest committed sequence number for data that was written to the persistent data storage by the replication server. A maximum value of latest committed sequence numbers received from the set of replication servers is determined. For each replication server, it is examined whether there is a gap between the start sequence number for data stored in local cache and the maximum value of the latest committed sequence numbers. Based on the examining, it is determined whether there is an occurrence of loss of data.

Abstract: An on-disk structure of a file system has the capability to efficiently manage and organize data containers, such as snapshots, stored on a storage system. A multi-bit, monotonically increasing, snapshot identifier (“snapid”) is provided that represents a snapshot and that increases every time a snapshot is generated for a volume of the storage system. The snapid facilitates organization of snapshot metadata within, e.g., a data structure used to organize metadata associated with snapshot data. In the illustrative embodiment, the data structure is a balanced tree structure configured to index the copy-out snapshot data blocks. The snapid is also used to determine which blocks belong to which snapshots. To that end, every block that is used in a snapshot has an associated “valid-to” snapid denoting the newest snapshot for which the block is valid. The oldest snapshot for which the block is valid is one greater than the valid-to field of the next older block at the same file block number.

Abstract: Methods and systems are disclosed to perform dynamic disk data redistribution in a database environment. The methods and systems disclosed herein provide for the transparent redistribution of data across multiple physical disks of a database while maintaining accessibility to the database. Methods and systems are also disclosed for handling requests to perform database operations on data that is in the process of being redistributed.

Abstract: A method, a system, an apparatus, and a computer readable medium for transmission of data across a network are disclosed.

Abstract: One implementation of a method for providing backup storage services for encrypted data includes receiving signatures of convergently encrypted portions of data from client computers, determining whether the encrypted portions are already present on a backup storage, and obtaining only the needed portions. Users unassociated with a particular user account are denied access to information backed up in that account. The backup storage also stores password protected key files holding signatures of the unencrypted portions of data. One implementation of a system includes a memory, a single-instance storage circuit, a user account management circuit, and a signature index. The memory holds a user-account database and backup copies of convergently encrypted portions of data. The single-instance storage circuit uses the signature index to prevent duplicative backup copies. The user account management circuit responds to download requests after authenticating the user information associated with the requested data.

Abstract: Provided herein are systems and methodologies for highly efficient restoration in a network-based backup system. As described herein, differential-based analysis can be utilized such that a new complete differential is calculated based on signatures and/or other information relating to a given item to be restored prior to retrieving backup data. Based on the differential, only blocks determined to be unique between the current version of the item and the desired version are transmitted, which can then be merged with non-unique locally present blocks to obtain the fully restored version of the item. Further, a hybrid architecture can be employed, wherein signatures and/or data are stored at a global location within a network as well as one or more local peers. Accordingly, a backup client can obtain information necessary for restoration from either the global location or a nearby peer, thus further reducing latency and bandwidth consumption.

Abstract: The technique introduced here involves using a block address and a corresponding generation number as a “fingerprint” to uniquely identify a sequence of data within a given storage domain. Each block address has an associated generation number which indicates the number of times that data at that block address has been modified. This technique can be employed, for example, to determine whether a given storage server already has the data, and to avoid sending the data to that storage server over a network if it already has the data. It can also be employed to maintain cache coherency among multiple storage nodes.

Abstract: A method, device, and program product are provided for inter-site database replication. A replicating server tracks other servers that it needs to replicate with and the sites where the other servers are located. In response to an indication to replicate, the replicating server determines whether each of the other servers is at the same location as the replicating server or a different location from the replicating server. The replicating server replicates with each of the other servers at the same site as the replicating server, and the replicating server replicates with only one of the servers at each different site. The replicated server at each different site replicates with the other servers at the site of the replicated server.

Abstract: Methods and systems for managing for migrating feedback data from one digital asset to another digital asset are disclosed. Typically, the one digital asset is available for distribution from a network-based media distribution system, but then subsequently is removed from distribution for any of a number of reasons. However, since the one digital asset has been in use at the network-based media distribution system, it has accumulated feedback data. Hence, if the another digital asset serves (e.g., due to equivalency) to replace the one digital asset, then the accumulated feedback data from the one digital asset can be transferred to the another digital asset. As a result, the another digital asset can benefit from the feedback data that was previously associated with the one digital assert.

Abstract: An improved scalable object storage system includes methods and systems allowing multiple clusters to work together. In one embodiment, there is a multi-cluster synchronization system between two or more clusters. The multi-cluster synchronization system uses variable compression to optimize the transfer of information between the clusters. Compression is used not only to minimize the total number of bytes sent between the two clusters, but to dynamically vary the size of the objects sent across the wire to optimize for higher throughput after considering packet loss, TCP windows, and block sizes. This includes both the packaging of multiple small files together into one larger compressed file, saving on TCP and header overhead, but also the chunking of large files into multiple smaller files that are less likely to have difficulties due to intermittent network congestion or errors.

Abstract: The claimed subject matter relates to systems and/or methodologies that facilitate intelligent distribution of backup information across storage locations in network-based backup architectures. A virtual layering of backup information across storage locations in the backup architecture can be implemented. Statistical models are utilized to dynamically re-allocate backup information among storage locations and/or layers to ensure availability of data, minimum latency upon restore, and minimum bandwidth utilization upon restore. In addition, heuristics or machine learning techniques can be applied to proactively detect failures or other changes in storage locations such that backup information can be reallocated accordingly prior to a failure.

Abstract: The claimed subject matter provides a system and/or a method that facilitates distributed backup and versioning techniques for network-based backup architectures. When data is committed to a backup system, a backup version can be created. The generated backup version can be a complete rendition of data or the generated backup version can be a differential between the data and backup information previously committed to the system. The generated backup version can be replicated across a distributed network of peers. Replication can be based upon a desire level of redundancy of versions, location of peers, network topology of peers, storage capacity of peers, health of peers, type of peer, and the like. In addition, versions can be retained based upon a configurable number of desired versions.

Abstract: A transference control apparatus includes a communication controller, and a processor that executes a procedure, the procedure including processing a control that causes first processing, in which an update log representing the update content of updating performed for a first database is transmitted, by the communication controller, to a second database serving as a copy destination for the first database, to be performed with priority over second processing, in which transmission target data contained in the first database is transmitted to the second database, and restraining, in the case that updating of the transmission target data is allowed in the first database, the control that causes the first processing to be performed with priority.

Abstract: Described in detail herein are systems and methods for deduplicating data using byte-level or quasi byte-level techniques. In some embodiments, a file is divided into multiple blocks. A block includes multiple bytes. Multiple rolling hashes of the file are generated. For each byte in the file, a searchable data structure is accessed to determine if the data structure already includes an entry matching a hash of a minimum sequence length. If so, this indicates that the corresponding bytes are already stored. If one or more bytes in the file are already stored, then the one or more bytes in the file are replaced with a reference to the already stored bytes. The systems and methods described herein may be used for file systems, databases, storing backup data, or any other use case where it may be useful to reduce the amount of data being stored.

Abstract: Techniques for replicating data chunks in a storage system are described herein. In one embodiment, in response to a request for replicating data chunks of a source storage system having a first average chunk size to a target storage system having a second average chunk size, a new chunk size is determined based on metadata of the data chunks in view of an average chunk size of the target storage system. The data chunks are resized based on the new chunk size to generate resized data chunks. The resized data chunks are transmitted from the source storage system to the target storage system for replication.

Abstract: According to one embodiment, an apparatus may receive a token that indicates a change that occurs during a session. The session may facilitate access to a resource. The token may indicate a risk token should be computed. The apparatus may determine, from the token, a first set of attributes. The first set of attributes may include attributes required to compute the risk token. The apparatus may determine that a cache contains a set of cached attributes. The apparatus may examine an attribute in the set of cached attributes, and determine the attribute in the set of cached attributes is not in the first set of attributes. The apparatus may then remove the attribute in the set of cached attributes from the cache.

Abstract: A client caches objects from a data set on a server. The client registers to receive notification of events that occur to cached objects. As events occur to cached objects, the client determines if a specified time interval has elapsed since the cached objects were last updated. If a specified time interval has elapsed, then the client updates the cached objects. If a specified time interval has not elapsed, then the client waits for the specified time to elapse, during which time additional updates to the objects on the server can occur.

Abstract: Embodiments deploy delayering techniques, and the relationships between successive versions of a rich-media file become apparent. With this, modified rich-media files suddenly present far smaller storage overhead as compared to traditional application-unaware snapshot and versioning implementations. Optimized file data is stored in suitcases. As a file is versioned, each new version of the file is placed in the same suitcase as the previous version, allowing embodiments to employ correlation techniques to enhance optimization savings.

Abstract: A method and apparatus are provided for enhancing the impact of data de-duplication by preferential selection of the master copy to be retained based on current loads and performance metrics of the storage media devices. The computer system is configured to take as input the identified duplicate copies of data and evaluating their locations in storage devices to determine the cumulative affects of retaining one of the identified duplicate copies as a master copy and optionally allocating a new location if needed. Once a master copy has been designated, the remaining identified duplicate copies are removed from storage.

Abstract: In each of a number of passes to deduplicate a data object, a transaction is started. Where an offset into the object has previously been set, the offset is retrieved; otherwise, the offset is set to reference a beginning of the object. A portion of the object beginning at the offset is deduplicated until an end-of-transaction criterion has been satisfied. The transaction is ended to commit deduplication; where the object has not yet been completely deduplicated, the offset is moved just past where deduplication has already occurred. The object is locked during each pass; other processes cannot access the object during each pass, but can access the object between passes. Each pass is relatively short, so the length of time in which the object is inaccessible is relatively short. By comparison, deduplicating an object within a single pass prevents other processes from accessing the object for a longer time.

Abstract: Systems and methods for checkpointing a computation distributed over multiple peer servers. On each server, sequentially storing checkpoints collectively representing a current state of the computation on that server as of a most recent checkpoint, each checkpoint having a checkpoint timestamp. When restarting a first server, rebuilding a most recent state of the first server from the checkpoints written by the first server through a most recent checkpoint having a most recent checkpoint timestamp, and requesting from each of the other peer servers updates from the most recent checkpoint timestamp time of the first server. On each server, in response to a first request for updates as of a particular time, deriving the requested updates from the state data in the server uncommitted to a checkpoint and the state data in checkpoints of the server that have a timestamp no earlier than the particular time of the first request, and providing the requested updates to the first server.

Abstract: A method for managing multiple generations of an object within a distributed storage system is implemented at a computing device. The computing device receives metadata and content of a first generation of an object from a first client connected to the distributed storage system and stores the first generation's metadata and content within a first storage sub-system. The computing device receives metadata and content of a second generation of the object from a second client connected to the distributed storage system and stores the second generation's metadata and content within a second storage sub-system. The computing device independently replicates the first generation's metadata and content from the first storage sub-system to the second storage sub-system and replicates the second generation's metadata and content from the second storage sub-system to the first storage sub-system such that both storage sub-systems include a replica of the object's first and second generations.

Abstract: A harvester is disclosed for harvesting metadata of managed objects (files and directories) across file systems which are generally not interoperable in an enterprise environment. Harvested metadata may include 1) file system attributes such as size, owner, recency; 2) content-specific attributes such as the presence or absence of various keywords (or combinations of keywords) within documents as well as concepts comprised of natural language entities; 3) synthetic attributes such as mathematical checksums or hashes of file contents; and 4) high-level semantic attributes that serve to classify and categorize files and documents. The classification itself can trigger an action in compliance with a policy rule. Harvested metadata are stored in a metadata repository to facilitate the automated or semi-automated application of policies.

Abstract: In an information processing system, a client terminal includes a client process executing section issuing a request; a client execution control section generating a distribution rule based on an instruction from the server apparatus and operation environment data of the client terminal; a distributing section determining whether or not the request should be processed by the client terminal or by the server apparatus, based on the distribution rule, distributing the request to the server apparatus when it is determined that the request should be processed by the server apparatus, and returning a received processing result to the client process executing section; and a client side server process executing section executing the request when it is determined when the request should be processed by the client terminal, and notifying a processing result to the distributing section.

Abstract: The invention is a system and method for quickly cloning a backup (file system) volume. Present volume cloning methods are based on creating copies of individual save sets written to a backup volume. However, this requires the steps of reading data blocks to cull the backed up data, stored as data chunks, from the chunk headers that separate and describe each data chunk, then writing the data chunks to new media. As a result, cloning a backup volume is time consuming and tedious. The invention applies the features and benefits of advanced file system devices (“AFTDs”) to volume cloning. AFTDs store data as streams, rather than chunks, since AFTDs store data on disk media rather than on tape with tape blocks. As such, data stored on AFTDs are not separated by chunk headers. Instead, an AFTD contains a single volume label file that identifies the AFTD volume. The invention is a system and method for cloning backup volumes stored on AFTDs.

Abstract: Systems, methods, and software for generic extraction of business object data are provided. One computerized method includes retrieving multiple identifiers of sub-nodes. The sub-nodes are associated with a root node of a business object. The method further includes generating a tree data structure. The tree includes multiple tree nodes and multiple tree edges. The multiple nodes correspond to the root node of the business object and the retrieved multiple identifiers. The method also includes traversing the tree to retrieve data associated with the root business object and the retrieved multiple identifiers. The method further includes storing the retrieved data.

Abstract: A system and method for simulating a state of a distributed storage system is provided. A current state of a distributed storage system and replication policies for the objects in the distributed storage system is obtained. Proposed modifications to the current state of the distributed storage system are received. The state of the distributed storage system is simulated over time based on the current state of the distributed storage system, the replication policies for the objects in the distributed storage system, and the proposed modifications to the current state of the distributed storage system. Then reports relating to the time evolution of the current state of the distributed storage system are generated based on the simulation.

Abstract: A reasonably-sized testing database instance can be efficiently replicated and maintained for a very large production database while retaining the characteristics and cross-sectional data. The performance characteristics are maintained in order to provide for proper testing of the production database for various application programs. Statistics on the type of data distribution for the customer data are obtained, allowing for parameters to be determined which can be used to store data only near the endpoints of the distribution (and/or at other key locations). In this way, a substantial amount of data skew is retained in a much smaller instance of the production database, allowing for easier performance testing, upgrade testing, etc.

Abstract: A method, system, and program product are provided for minimizing replication when a new server is replicated with. A replication application for a client computer having a replica of a database thereon is executed by a processor in a computer to receive for a replication service comprising a plurality of servers containing replicas of the database, an indication of each healthy server in the replication service and a determined period of time within which replication is assured to have been completed for each healthy server. The replication application updates a replication history after successful completion of each replication by adding an entry comprising an identification for the server replicated with and a timestamp for the time of replication and by adding an entry for each other healthy server in the replication service containing an identification for the respective server and an imputed timestamp adjusted for the determined period of time.

Abstract: An on-disk structure of a file system has the capability to maintain snapshot and file system metadata on a storage system. The on-disk structure arranges file system data sequentially on disk within regions, wherein each region comprises a predetermined amount of disk space represented by blocks. The snapshot and file system metadata is maintained within level 1 (L1) indirect blocks of the on-disk structure. Each L1 indirect block describes (i.e., represents) a corresponding region of the on-disk structure of the file system; in the case of an active region, e.g., an L1 indirect block represents an active file data portion of a large file. The L1 indirect block that references an active region also performs file block number (fbn) to disk block number (dbn) mapping for the region.

Abstract: A method and apparatus perform file copying, including selecting a plurality of files, which a first terminal desires to copy, and temporarily storing the selected files in a second terminal; identifies whether there is a file collision between the first terminal and the second terminal before performing a copying of the temporarily stored files; generates a queue, which is a copying process order of the temporarily stored files, according to the existence of the file collision; copies the temporarily stored files using the queue; and displays a copying process speed of each file when a copying of queued files is performed.

Abstract: An information retrieval system having: a client adapted for accessing a plurality of file sets stored on one of a plurality of file servers; a plurality of file servers configured to operate with a federated file system namespace; and a memory for storing re-direction information accessible by the client for identifying a request issued by the client for a file set at a first location in the namespace where the file set is located at a second, different location on one of the file servers and wherein the client in examining the re-direction information in the memory, re-directs the request to the second location in accordance with the re-direction information.

Abstract: The present invention provides for an enhanced method and system for assuring integrity of deduplicated data objects stored within a storage system. A digital signature of the data object is generated to determine if the data object reassembled from a deduplicated state is identical to its pre-deduplication state. In one embodiment, generating the object signature of a data object before deduplication comprises generating an object signature from intermediate hash values computed from a hash function operating on each data chunk within the data object, the hash function also used to determine duplicate data chunks. In an alternative embodiment, generating the object signature of a data object before deduplication comprises generating an object signature on a portion of each data chunk of the data object.

Abstract: A set of virtualized computing services may include multiple types of virtualized data store differentiated by characteristics such as latency, throughput, durability and cost. A sequence of captures of a data set from one data store to another may be scheduled to achieve a variety of virtualized computing service user and provider goals such as lowering a probability of data loss, lowering costs, and computing resource load leveling. Data set captures may be scheduled according to policies specifying fixed and flexible schedules and conditions including flexible scheduling windows, target capture frequencies, probability of loss targets and/or cost targets. Capture lifetimes may also be managed with capture retention policies, which may specify fixed and flexible lifetimes and conditions including cost targets. Such data set capture policies may be specified with a Web-based administrative interface to a control plane of the virtualized computing services.

Abstract: A computer is programmed to parse a network that interconnects databases in a distributed database system, into a number of paths, wherein each path starts in a source database that is a source of changes and ends in a destination database that uses the changes to replicate the source database. The computer identifies databases that occur within each path in the network, and queries the identified databases to obtain statistics for each path. Statistics along each path are analyzed automatically, to identify a component in each path as a candidate for further investigation. The identified component may be marked as a bottleneck e.g. if there are no other paths connected to it. Alternatively the identified component may be a source of changes for another path that has a bottleneck or that has another candidate for further investigation. The identified bottleneck is displayed to the user with recommendations for improvement.

Abstract: Methods are disclosed for, within a note-taking application, selectively synchronizing different notebook components with a centralized server. In one particular embodiment, different notebook components are assigned different synchronization rates based on usage characteristics. The different notebook components are synchronized more or less frequently relative to each other based on the assigned synchronization rates.

Abstract: A system and method for providing a substantially constant-time copy operation for file system objects managed by a network storage server begins by generating a new file based on metadata in a source file. The system then generates a snapshot of the source file or the logical volume in which the source file resides. The system then copies each of the indirect blocks of the source file to a corresponding location in the destination file. During the copy process, a modified set of file system operations can be executed on the destination metadata container. In response to a read request directed to a target block in the destination file, the system determines if target block has been copied. If the block has not been copied, the system provides a data segment from the corresponding data block in the snapshot. Write requests are processed by generating a new data block and adding a reference to the new data block to the target location in the destination metadata container.

Abstract: The system provides a technique for finding relevant content and content sources based on the aggregation and analysis of individual indications of relevance. The system identifies and provides selections of relevant content. It may comprise a selection acquisition subsystem, a selection network repository subsystem, a scoring engine, and a recommendation engine, and is used to generate sources of content comprising sets of prioritized links directed to a topic or community of interest.

Abstract: A method for replicating I/O performance in data replication environments, such as PPRC environments, is described. In selected embodiments, such a method includes monitoring I/O workload at a primary storage device over a period of time, such as a period of hours, days, or months. The method then generates learning data at the primary storage device describing the I/O workload over the selected time period. The learning data is replicated from the primary storage device to a secondary storage device. The method uses the learning data to optimize the secondary storage device to handle the I/O workload of the primary storage device. This will enable the secondary storage device to provide substantially the same I/O performance as the primary storage device in the event a failover occurs. A corresponding system and computer program product are also disclosed and claimed herein.

Abstract: Techniques are described for sharing multidimensional data between software systems, such as enterprise software systems. Specifically, the techniques provide mechanisms for defining inter-application “links” for automatically moving data among different databases associated with the enterprise software systems. For example, a system may include a first database storing multidimensional data for a first software application, and a second database storing multidimensional data for a second software application. The system further includes an interface by which an administrator defines a link that specifies a source area of the first database and a target area of the second database. A link control module automatically moves multidimensional data from the source area to the target area in accordance with the link. The link control module may automatically move the multidimensional data directly between live portions of the respective databases.

Abstract: A method for replicating I/O performance in data replication environments, such as PPRC environments, is described. In selected embodiments, such a method includes monitoring I/O workload at a primary storage device over a period of time, such as a period of hours, days, or months. The method then generates learning data at the primary storage device describing the I/O workload over the selected time period. The learning data is replicated from the primary storage device to a secondary storage device. The method uses the learning data to optimize the secondary storage device to handle the I/O workload of the primary storage device. This will enable the secondary storage device to provide substantially the same I/O performance as the primary storage device in the event a failover occurs.

Abstract: Example methods, apparatus and articles of manufacture to transfer files are disclosed. A disclosed example method includes calculating ratios for nodes within a first file system, wherein the ratios are based on a ratio of a number of files at a node to a total file size of the files at the node and distributing the nodes among sub-traversal paths based on the ratios to minimize deviation of the ratios of the sub-traversal paths.

Abstract: A method and system are disclosed for storage optimization. Data parts and metadata within a source data unit are identified and the data parts are compared with data which is already stored in the physical storage space. In case identical data parts are found within the physical storage, the data parts from the source data unit are linked to the identified data, while the data parts can be discarded, thereby reducing the required storage capacity. The metadata parts can be separately stored in a designated storage area.

Abstract: Embodiments of erasure coded storage aggregation are disclosed. The erasure coded storage aggregation includes storing a data file as erasure coded fragments in a plurality of nodes of one or more data centers. The erasure coded storage aggregation further includes monitoring an access frequency of the data file. Based on the comparison between the access frequency and a predetermined threshold, the data file is either reconstructed from the erasure coded fragments and stored in a storage node or retained as erasure coded fragments in the plurality of nodes of the one or more data centers.