Decentralized memory techniques and related systems and methods

A shared memory system in distributed networks using logical tags and linear coding addresses the challenge of maintaining operation consistency and consensus, ensuring efficient and reliable object management in dynamic environments.

WO2026241656A1PCT designated stage Publication Date: 2026-11-19SPICE SOLUTIONS INC
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
PCT/US2026/028406
Authority / Receiving Office
WO · WO
Patent Type
Applications
Current Assignee / Owner
Priority Date
2025-05-15
Filing Date
2026-05-15
Publication Date
2026-11-19

AI Technical Summary

Technical Problem

Distributed networks face challenges in maintaining a consistent order of operations across multiple devices, particularly in achieving consensus in fully asynchronous environments, which can be computationally expensive and may not guarantee consensus is achieved.

Method used

Implementing a shared memory system in distributed networks where each object is stored by a subset of network nodes, using logical tags to order operations, and employing linear coding techniques to disseminate values, allowing nodes to join or depart dynamically while ensuring efficient read and write operations.

Benefits of technology

The solution enables efficient and reliable management of multiple objects in distributed networks by ensuring consistent ordering of operations and high fault tolerance, even in dynamic network conditions.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure US2026028406_19112026_PF_FP_ABST
    Figure US2026028406_19112026_PF_FP_ABST
Patent Text Reader

Abstract

Techniques are described for implementing a shared memory in a distributed network capable of concurrently managing multiple objects. In particular, each object is stored by a subset of the network nodes, with each object being in general stored by a different subset of the network nodes. The techniques described herein allow a network node to join the network, determine which of the objects the network node will be responsible for, and retrieve and store those objects with minimal communication needed with other network nodes. Furthermore, the techniques allow a given object to be read or written to by any network node in the network while dynamically allowing network nodes to join or depart the network.
Need to check novelty before this filing date? Find Prior Art

Description

Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026DECENTRALIZED MEMORY TECHNIQUES AND RELATED SYSTEMS AND METHODSBACKGROUND

[0001] Distributed networks employ various processes to disseminate information across the network. Distributed networks have no central registry and instead rely on network nodes to periodically exchange information with neighboring network nodes, thereby disseminating information. For instance, in some approaches each network node in a distributed network periodically selects a subset of neighboring network nodes and exchanges information with them. In this way, information spreading in a distributed system can be considered conceptually similar to the spread of a disease or the spread of gossip in a social network. This approach allows information to propagate rapidly and reliably through a distributed network, and is therefore desirable for applications such as distributed databases, blockchains, and peer-to-peer networks.SUMMARY

[0002] According to some aspects, the techniques described herein relate to a method including: reading a value of a first object from a distributed memory by: identifying one or more network nodes responsible for the first object; receiving a tag-value pair from each of the one or more network nodes responsible for the first object, thereby receiving a plurality of tag-value pairs, wherein each tag-value pair includes a tag number and a plurality of coded elements of the first object; identifying one or more of the plurality of tag- value pairs that include a most recent tag number among the plurality of tag-value pairs; and decoding the value of the first object from the plurality of coded elements of the first object in the one or more of the plurality of tag-value pairs that include the most recent tag number.

[0003] According to some aspects, the techniques described herein relate to a method including: writing a value of a first object to a distributed memory by: identifying one or- 1 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026more network nodes responsible for the first object; receiving a tag number from each of the one or more network nodes responsible for the first object, thereby receiving a plurality of tag numbers; and sending, to each of the one or more network nodes responsible for the first object: a plurality of coded elements of the first object; and a new tag number generated based on a most recent tag number among the plurality of tag numbers.

[0004] The foregoing apparatus and method embodiments may be implemented with any suitable combination of aspects, features, and acts described above or in further detail below. These and other aspects, embodiments, and features of the present teachings can be more fully understood from the following description in conjunction with the accompanying drawings.BRIEF DESCRIPTION OF DRAWINGS

[0005] Various aspects and embodiments will be described with reference to the following figures. It should be appreciated that the figures are not necessarily drawn to scale. In the drawings, each identical or nearly identical component that is illustrated in various figures is represented by a like numeral. For purposes of clarity, not every component may be labeled in every drawing.

[0006] FIG. 1 is a flowchart of a method of reading a value of a first object from a distributed memory, according to some embodiments;

[0007] FIG. 2 depicts an illustrative approach to measuring a pseudo-distance between nodes and object, according to some embodiments;

[0008] FIG. 3 depicts a part of an illustrative network, according to some embodiments;

[0009] FIG. 4 depicts an illustrative process of updating a local estimate of a state machine, according to some embodiments;

[0010] FIG. 5 is a flowchart of a method of writing a value of a first object to a distributed memory, according to some embodiments;

[0011] FIG. 6 depicts a process of joining a network, according to some embodiments;- 2 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026

[0012] FIGs. 7A-7E depict stages in the process of joining a network, according to some embodiments; and

[0013] FIG. 8 illustrates an example of a computing system environment on which aspects of the disclosure may be implemented.DETAILED DESCRIPTION

[0014] As described above, network nodes in distributed networks can share information with one another to disseminate information through the network. Maintaining a consistent order of operations when operations can be executed concurrently across multiple devices can be a challenge, however. In some approaches, the order of operations in the system is agreed upon by the network nodes according to a consensus protocol. For example, some blockchain systems require participants to validate new information by performing work or by staking assets, and only when the necessary validation is performed can the whole network agree to commit the new information to the system. Achieving consensus in this manner in a fully asynchronous environment introduces technical challenges. For example, it may not be possible to guarantee that consensus is achieved, and consensus mechanisms may be computationally expensive.

[0015] The inventors have recognized and appreciated techniques for implementing a shared memory in a distributed network capable of concurrently managing multiple objects. In particular, each object is stored by a subset of the network nodes, with each object being in general stored by a different subset of the network nodes (also referred to herein simply as “nodes”). The techniques described herein allow a node to join the network, determine which of the objects the node will be responsible for, and retrieve and store those objects with minimal communication needed with other nodes. Furthermore, the techniques allow a given object to be read or written to by any node in the network while dynamically allowing nodes to join or depart the network. A network operated according to these techniques may also be referred to herein as implementing a “distributed memory.”- 3 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026

[0016] According to some embodiments, nodes in the network may each have access to a state machine that keeps track of which nodes have joined and left the network. A first node in the network may receive information from the state machine and from other nodes in the network indicating participant changes to the network (e.g., joins, departures). The first node may store updates to the last known state of the state machine when the received changes represent changes to first node’s knowledge of the network participants. In some embodiments, network participant changes may be provided in response to read or write requests so that changes to the network participants are propagated between nodes as a byproduct of the read or write request, rather than through additional communication (e.g., communications that only share network updates). Furthermore, in some embodiments, changes to network participants may be sent only when they relate to nodes responsible for a particular object being read, or being written to. For example, when a read request for an object is sent from a first node to a second node, the second node may respond to the read request in part with a list of network changes with respect to the nodes that are responsible for that object. Participant changes relating to other nodes that are not responsible for the object is unnecessary to complete the read request, and so it may be less efficient to share information on such changes.

[0017] According to some embodiments, nodes in the network may generate and store logical tags to order operations. Tags may comprise some indication of their order, so that the most recent tag may comprise, or otherwise be associated with, the most recently written values for a particular object within the network. When a node writes a new value of the object to the network, it determines the most recent tag in the network, then disseminates the new value of the object with a new, more recent, tag value. In some embodiments, for example, the tags may each comprise a number that is incremented or otherwise increased with each successive write operation. When a node reads an object from the network, it may similarly identify the most recent tag in the network and then read the object by accessing data describing the object that was provided with the most recent tag. In both read and write operations, a request for a tag from another node is effectively a request for the most recent- 4 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026tag that those nodes have received, which in some cases may not be the most recent tag in the network. In some embodiments, the current most recent tag may be identified by receiving a number of tags above a threshold number from other nodes that contain the most recent tag among all the tags received. For example, receiving one instance of the most recent tag may not provide sufficient reliability that this tag is in fact the most recent tag in the network. As will be described below, by establishing a quorum system, a threshold number of the most recent tags received may be identified that provides high fault tolerance.

[0018] According to some embodiments, values representing an object are disseminated through the network through linear coding of the values. In such cases, each coded element being provided (e.g., in response to a read request) represents a linear equation, and once a sufficiently large set of such linear equations has been received by a node, a node can solve the set of equations to decode the message. In general, nodes may transmit linear combinations of coded elements they have received to other nodes in the network. In some embodiments, the messages are disseminated through the distributed network according to linear network coding (LNC) or random linear network coding (RLNC) techniques. These coding techniques are described further below.

[0019] According to some embodiments, the nodes in the network may send data to one another with an identifier that identifies the sending node of the data or can otherwise be verified as having come from the node that transmitted the data. For instance, data representing an object may be sent by nodes with a signature that can be checked by a receiving node as having been validly sent by the sending node. Signatures may also be generated based on additional data, such as data that indicates how recent a supplied tag is (e.g., a number). In some embodiments, a tag may comprise data that indicates how recent the tag is, in addition to an identifier of a network node generating the tag. The tag containing these two elements may be combined with data representing an object and a signature generated for the tag and the data, thereby allowing a recipient to verify the identity of the sender, confirm how recent the tag is, and confirm the validity of the data representing the object.- 5 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026

[0020] Following below are more detailed descriptions of various concepts related to, and embodiments of, techniques for implementing a shared memory in a distributed network capable of concurrently managing multiple objects. It should be appreciated that various aspects described herein may be implemented in any of numerous ways. Examples of specific implementations are provided herein for illustrative purposes only. In addition, the various aspects described in the embodiments below may be used alone or in any combination, and are not limited to the combinations explicitly described herein.

[0021] FIG. 1 is a flowchart of a method of reading a value of a first object from a distributed memory, according to some embodiments. Method 100 may be performed by any suitable computer-executable code, processor, controller and / or computing system, including but not limited to the system(s) illustrated in FIG. 8, which is, or is a component of, a network node of a distributed network. References herein to the network node performing acts of method 100 should be considered to implicitly also refer to cases where certain components of a network node are performing those acts of method 100. In the remaining description below, network nodes will also be referred to as “nodes” for clarity.

[0022] In act 102, the network node performing method 100 identifies one or more network nodes that are responsible for the first object. As described above, according to the techniques described herein a subset of nodes in the network may be responsible for maintaining a given object. In act 102 the network node performing method 100 determines which of the nodes in the network is the subset responsible for maintaining the first object, which is to be read from the network via the method 100.

[0023] According to some embodiments, the network node performing method 100 may store a list of one or more object identifiers. This list of object identifiers may be accessed in act 102 when determining which of the nodes in the network is responsible for maintaining that object.

[0024] According to some embodiments, determining which nodes are responsible for storing a first object may comprise calculating the value of a function 3) based on an object identifier, and calculating the value of© based on one or more node identifiers. For- 6 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026instance, ©( / ) where I is an identifier of a node no de ID or an identifier of an object objectID. In this approach, the nodes that are responsible for the object are those nodes for which T)(nodeID is close to the value of ^objectID . For instance, ifl)(nodeID') is within some threshold range of T) (objectID'), the node associated with the node identifier nodelD may be considered to be responsible for the object associated with objectID.Otherwise, if ^ nodelD is not within the threshold range of ^ objectID , the node is not responsible for that object. Using this approach may allow a node to efficiently determine which identified nodes are responsible for an object of interest without communication with any of those nodes being necessary. A measure of closeness between ©(A) and ©(B) is referred to herein as a “distance” between ©(A) and ©(B) for any A or B. The distance may be a numerical difference or may be defined with some other metric, examples of which are described below. The term “distance” as used herein refers to a pseudo-distance, and not a physical distance (Euclidean distance), and may for instance refer to the value of a calculated metric, may refer to the difference between multiple values of a calculated metric, or may refer to the difference between the value of a first calculated metric and the value of a second calculated metric. The terms distance and pseudo-distance may both be used herein and should be considered interchangeable for the purposes of this disclosure.

[0025] In some embodiments, a distance between two values produced by the function © may be defined by identifying the domain of values produced by © as positions around a ring, and measuring a distance around the ring. By utilizing a ring, any two outputs from the function © have a measurable distance, which may be measured in any suitable way, such as the distance from one point to another around the ring in one direction only, or the absolute distance from one point to another in either direction.

[0026] As an example, the values produced from a given function © for each of nine different nodes are arranged around a ring 200 as shown in FIG. 2, according to some embodiments. That is, an identifier for each of the nodes 201, 202, 203, 204, 205, 206, 207, 208, 209 and 210 is passed to the function © and the values produced from © are each mapped to a position around the ring. In the example of FIG. 2, the distance between two- 7 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026points around the ring is measured clockwise as shown by the arrows. The values produced by 2) for each of two objects Eire also identified in FIG. 2.

[0027] In the example of FIG. 2, a range of distances from a value produced by D for a given object is highlighted with a grey band, representing the range of distances from the point around the ring associated with the object within which nodes will be responsible for that object. For example, in the example of FIG. 2, nodes 202, 201 and 204 are responsible for object 1, whereas nodes 203, 207 and 205 are responsible for object 2. It will be appreciated that there are other ways of measuring distance around the ring, and that FIG. 2 is merely one example. For instance, the absolute distance around the ring in either direction might instead be used to determine which nodes are responsible for a given object (e.g., in which case nodes 209, 202 and 201 might be responsible for object 1, and nodes 208, 203 and 207 might be responsible for object 2).

[0028] According to some embodiments, act 102 may comprise providing a list of node identifiers to a function 3), providing an object identifier associated with the first object being read to the function 3), and calculating a distance between the value produced from the object identifier to each of the values produced from the node identifiers. The node identifiers that produced values with a distance below a threshold may be identified, and consequently the node performing act 102 may generate a list of node identifiers corresponding to the nodes that are responsible for the first object. As such, this process may determine, for a given list of nodes, a subset of those nodes that are responsible for a given object. In general, different subsets of the same list of nodes will be responsible for different objects.

[0029] In some embodiments, the function 3) may be a hash function (e.g., a collision-resistant hash function). For example, node identifiers and object identifiers may each be mapped by the hash function to a 256-bit space, thereby supporting N= 2256elements. Given two identifiers a, b G 0 U I with hashes ida= 7f (a) and idb=(b), where I is a set of nodes and 0 is a set of objects, the distance between the identifiers a, b may be measured relative to their positions around a ring. For instance, the distance may be- 8 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026measured clockwise, or counter-clockwise, or by the shortest distance in either direction. In the case of clockwise measurement, the distance may be measured as:Hdrl id if id / , > idaI bJ —+if

[0030] Alternatively, this can be expressed compactly using modulo arithmetic:d(ida, idb') = idb— ida) mod N

[0031] This formulation of the function D ensures that the distance is always nonnegative and measured clockwise. This implementation of the function 2) may be defined as:®suc GU) = {s° . S"}where slG I and d (H(O),M(S1')) < d(M(p'),H(sr)'), for 0 < i < n and Vs' G I \ ‘DCo.I).

[0032] The same mapping can be used as Dpre(f, 7) to find the closest nodes in the “neighborhood” of a node f G I (although the ‘neighborhood’ relates merely to distance between the respective hash values, and may have no relationship to physical distance).

[0033] Returning to FIG. 1, in act 102 the node performing method 100 (referred to below as the “first network node” for convenience) may access a list of node identifiers that represent known participants in the network.

[0034] As described above, nodes in the network may each have access to a shared state machine that keeps track of which nodes have joined and left the network. As used herein, a “state machine” refers to a service that can be accessed by a network node to obtain a response indicating the known participants in the network (e.g., as a complete list, as a list of changes relative to some initial configuration, etc.). In some cases, a state machine may be accessed by a network node to record changes to the participants in the network (e.g., when a node joins or leaves the network, it may send a request to the state machine to be added to, or removed from, the network). According to the techniques described herein, nodes in the network may each store local estimates of the participants in the network, - 9 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026which may be accessed during read and / or write operations (and which functions as the aforementioned list of node identifiers that represent known participants in the network). In this manner, read and write operations may be based on the local estimate and thereby performed efficiently without communication with the state machine. In addition, some or all of a node’s local estimate may be communicated to another node, which may in turn update its local estimate. In some embodiments, the state machine may be implemented as a blockchain.

[0035] FIG. 3 depicts a part of an illustrative network comprising nodes 201, 202, 203 and 204, where each of the nodes stores a respective local estimate of the state machine 201a, 202a, 203a, and 204a, respectively. It is desirable that, before initiating or before completing a read operation on the network for a given object, the reading node’s local estimate of the state machine is current. Techniques for performing this process are described below in relation to act 108 of method 100.

[0036] According to some embodiments, a local estimate of a state machine may comprise, or otherwise provide, a list of nodes in the network. The local estimate may, for instance, store a list of node or may store a list of additions and removals that may be compiled to may determine which nodes are presently understood to be in the network. Irrespective of how the information is represented, the local estimate stored by a node that represents all of the known network participants is also referred to herein as S. In the example of FIG. 3, for instance, the local estimates of nodes 201 and 204 comprise indications that node 209 left the network, and node 204 joined the network. However, the local estimate of node 202 has not yet been updated to indicate that node 204 joined the network, and the local estimate of node 203 has not yet been updated to indicate that node 209 left the network

[0037] In some embodiments, act 102 may comprise the first network node accessing its local estimate to generate a list of participating nodes S (and / or accessing a stored list of participating nodes S), and determining which of the nodes from S are responsible for the- 10 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026first object being read in method 100. For instance, act 102 may comprise determining C <-Dsuc(o,S) where o is an identifier of the first object.

[0038] In act 104, the first network node sends a query to each of the nodes identified in act 102, requesting that they provide data for the first object. In some embodiments, the query may comprise an identifier of the first object and a local estimate of participating nodes C responsible for the first object, which was generated by and / or stored by the first network node (where C is a subset of S). Alternatively, the local estimate of all participating nodes S may be provided in the request, though this may be less efficient in at least some cases.

[0039] According to the techniques described herein, a node receiving a request from a sending node (e.g., a read request, a write request, etc.) may receive a local estimate of participating nodes C' responsible for the object in the request, and may then determine differences between that local estimate and its own local estimate of participating nodes C responsible for the object. Any identified differences may be provided in a response to the request.

[0040] FIG. 4 depicts an example of such a request process, according to some embodiments, which may apply to any request described herein, and is not limited to the read request described in relation to FIG. 1. In the example of FIG. 4, the network node 410 issues a request and the network node 420 responds to this request. In the context of FIG. 1, for instance, node 410 may represent the first network node and sending requests in act 104, and node 420 may represent any of the nodes receiving the read requests sent in act 104.

[0041] In step 421, a request is sent from node 410 to node 420 that includes a list of nodes, such as list of participating nodes C responsible for an object or a full list of participating nodes S. In step 422, the node 420 compares the received list to its own version of the same list (e.g., compares received C to its determined C, or compares received S' to its determined S). The request sent in step 421 may also include an identifier of an object (e.g., an object being read or being written to).- 11 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026

[0042] In some embodiments, node 420 may receive a list of participating nodes C responsible for the object in the request from node 410 (representing node 410’s list of participating nodes), and in step 422 node 420 may determine C from its own list of participating nodes S. For example node 420 may determine C «- T)suc(o,S) as described above, using an object identifier o received in the request in step 421. If the local estimates of the state machine stored by node 410 and node 420 are different, then C and C may also be different if those differences relate to nodes responsible for the object. In the case of FIG. 3, if node 202 sends a request to node 201 relating to an object, the request may comprise node 202 ’s local estimate of participating nodes C responsible for the object, and node 204 may compare this list to its own local estimate of participating nodes C responsible for the object.

[0043] In some embodiments, in step 422, node 420 identifies a node in C that is not in C, and in response generates data indicating that this node should be added by node 410 to its local estimate of the state machine. For example, node 420 may generate an entry for node 410’s local estimate of the state machine adding the identified node. Such an entry may be generated by node 420 for each node identified in C that is not in C' . In the case of the above example relating to FIG. 3, for instance, node 201 may generate an entry “add 204” to be sent back to the node 202 when responding to its request in step 423.

[0044] In some embodiments, node 420 identifies a node in C' for which node 420’ s local estimate of the state machine includes an entry indicating that node was removed from the network, and in response generates data indicating that this node should be removed by node 410 from its local estimate of the state machine. For example, node 420 may generate an entry for node 410’s local estimate of the state machine removing the identified node. In the case of the above example relating to FIG. 3, for instance, if node 201 receives a request from node 203, node 201 may generate an entry “remove 209” to be sent back to the node 203 when responding to its request in step 423.

[0045] Returning to FIG. 1, in act 106 the first network node receives a tag and coded elements of the first object from one or more of the nodes to which queries were sent in act- 12 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026104. The combination of a tag and associated coded elements may also be referred to herein as a “tag-value pair” for convenience. This term encompasses the various different implementations of tags described herein, including the various different elements that may be provided within a tag, in addition to any suitable set of values that describe an object.

[0046] In some embodiments, act 106 comprises receiving one or more changes to the local estimate of the state machine of the first network node (e.g., one or more entries to add or remove certain nodes to or from the local estimate, as determined by the node receiving the query as shown in FIG. 4).

[0047] As described above, a tag may comprise data that indicates how recent the tag is, in addition to an identifier of a network node generating the tag. Such a tag may be received in act 106. In some embodiments, the data that indicates how recent the tag is consists of, or comprises, a number. For instance, the number may represent a ‘version’ of the associated data that is incremented each time the data is written to.

[0048] In the example of FIG. 1, data representing an object is encoded into a plurality of coded elements that are configured to be decoded using linear network coding (LNC) or random linear network coding (RLNC). In both LNC and RLNC, coded elements may be generated from linear combinations of other previously received coded elements associated with the same underlying object. Such linear combinations may be generated in part by multiplying the previously received coded elements with coefficients chosen from a finite field, to generate new coded elements (a process that may be referred to herein as “recoding”). As such, in some embodiments the first network node may be configured to generate new coded elements for a given object by generating a linear combination of previously received coded elements for that object, and transmit such new coded elements to other nodes in the network.

[0049] According to some embodiments, nodes in the distributed network that includes the network node performing method 100 may utilize codes defined over a finite field Fq to encode the contents of an object’s value v G V. In particular, a network node in the distributed network may encode the value using RLNC into n coded elements.- 13 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026Subsequently, the encoder or any other node in the network may linearly combine any subset of coded elements to derive new linear combinations (i.e., new coded elements). Any k < n of the generated coded elements is sufficient to decode the value v. According to some embodiments, to encode the value of an object in this scheme, a network node: (i) divides v into a vector of k elements (vltv2, ... , vfc) of Fq; (ii) selects a matrix A of coefficients at random from the finite field FQsuch that A may be composed of k rows and k columns; and (iii) multiplies A with vlfv2, ... , Vfc) to generate a vector of c = ( , C2, ••• , cfc) of k elements. This ( multiplication is formally illustrated below:al,lal,2a2,la2,2 (Eqn. 1)an,l «n,2

[0050] A node in the network may communicate the coded elements clfc2, ckto other nodes in the network to convey information about the value of the object.

[0051] In some embodiments, a response received in act 106 comprises {id, F9, k, c), where id G is an identifier of the message, Fqis the finite field over which the coefficients of A are selected; k G M is the number of elements of the vector v, and c G (FQ)fe. The matrix A, being randomly generated, can be shown to be invertible with sufficiently high probability. In the event that a random matrix is generated that is not invertible, it can be discarded by the node, and the node can wait to receive more coded elements until an invertible matrix can be generated by the node.

[0052] In some embodiments, act 106 comprises receiving a tag version number, an identifier of the sending node, a plurality of coded elements, and a signature generated based on the tag version number, identifier, and coded elements. The signature may be generated by a signature function that also validates signatures that it has generated, so that verifying the signature validates that the tag and coded elements have been generated by the original generator of the coded element. For instance, act 106 may comprise verifying a received signature by passing the received data, the received signature, and an identifier of the- 14 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026network node from which the data was received to a preconfigured signature verification function that generates an indication of whether the signature was properly generated by the network node having the supplied identifier in response (e.g., true or false).

[0053] In act 108 of method 100, the first network node applies any network changes received in act 106 to its local estimate of the state machine (this act is also represented by step 424 in FIG. 4). Updating the local estimate of the state machine in act 106 may, for instance, comprise appending entries received from other nodes to the local estimate, or rewriting the local estimate based on data received from the other nodes.

[0054] In the example of FIG. 1, when one or more network updates are received in act 106 and applied in act 108 to the local estimate of the state machine of the first network node, the method returns to act 102. Performing acts 102, 104, 106 then 108 in this manner may be repeated until no network updates are received in act 106, then proceeding to act 110. In some embodiments, method 100 proceeds to act 110 only when the acts 102, 104, 106 and 108 are repeated for a preselected number of times without receiving any network updates (e.g., to increase the likelihood that the network has stabilized before proceeding). Performing the acts 102, 104, 106 and 108 again may allow the first network node to cope with the dynamic nature of the network.

[0055] The process depicted by FIG. 4 is another way to view this same sequence of events from the perspective of both nodes, where steps 421, 422, 423 and 424 may be repeated until no network changes have been received (or are not received over a certain amount of time). Step 425 may represent a second instance of step 421 being performed (which may be followed by another instance of step 422, etc.).

[0056] In act 110, the first network node identifies the most recent tag among the tags received in act 106. In the case of a tag that comprises a version number, for instance, act 110 may comprise identifying the highest version number among the tags. There will always be at least one most recent tag received, although if the number (the count) of such tags received is one (or some other small number) there may not be sufficient reliability to conclude that the values provided with that tag are in fact the most recent in the network. As - 15 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026such, in act 110 the first network node determines the number of tags received that are all the most recent tag, and determines whether this number is above a quorum threshold. In some embodiments, the first network node may wait until responses have been received from all the network nodes to which queries were sent in act 104. In other embodiments, the first network node may wait for a predetermined time before identifying the most recent tag among the tags received.

[0057] According to some embodiments, the techniques described herein may employ a b-dissemination Byzantine quorum system. A b-dissemination quorum system guarantees that the intersection of any two quorums contains at least one correct element, even in the presence of b faults. More precisely, Q £ 2Cis a b-dissemination quorum system if V Qi, Q2 GQ and | A Q21 > b + 1. If the use of erasure codes is assumed, where each node can store at most a single coded element, then the intersection of two quorums contain at least k correct elements (where k is the number of coded elements needed to decode the values of an object). Consequently, V QltQ26 Q and |Qi A Q2| > b + k. Furthermore, it may also be assumed that at any given time in the execution, there exists a single quorum in Q that does not contain any failed nodes, i.e. 3Q G Q, such that V / G Q,f is correct.

[0058] Given that | Q±A Q21 > b + k and that b of the nodes inmay not be included in the quorum Q2, it may be guaranteed that the two quorums overlap on b 4- k elements if | Qi, 21 2b + k. Moreover, any quorum should have size at most |Q| < |C| — b, otherwise an operation may not receive enough responses to terminate the operation. From the two inequalities it follows:|C| - k\C\ — b > 2b + k implies b < — - —

[0059] From this it follows that the size of each quorum Q G Q is:2ICI + k|Q| < |C| - b implies |Q| < - - -- 16 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026|C|— k

[0060] To achieve high fault tolerance, it may be assumed that b < , and the sizeof each quorum Q G Q, |Q| = [2lc^+k] T0this end, Q E Q, such that, V G Q,f is correct.

[0061] Accordingly, act 110 may comprise determining whether the number (the r2|C|4-fclcount) of responses received is greater than — - — , where |C| is the number of nodes in thelist C of the first network node (which are the nodes responsible for the first object according to the first network node’s local estimate of the state machine). As one example, if there are 20 nodes in C and the number of elements needed to decode the value of the object is 8, then a quorum of nodes has responded when at least 16 of the nodes in C respond. Act 110 may further comprise determining whether a quorum of nodes have each responded with the same most recent tag. That is, determining whether a quorum of nodes has responded may comprise determining whether the number of nodes that responded with r2ici+fcithe same most recent tag is greater than a quorum threshold of — - — .Other quorum thresholds may also be considered, however, such as at least a number of nodes that would allow decoding of the value in the most recent tag.

[0062] More generally, act 110 may comprise determining whether the number of tags received that are all the most recent tag is greater than a threshold, where the threshold is a function of the number of nodes in the list C of the first network node and a function of the number of elements needed to decode the value of the object.

[0063] In act 110, if the first network node determines that a quorum of nodes has responded with the same most recent tag (i.e., the number of nodes that responded with the same most recent tag is above a predetermined threshold value), the method proceeds to act 112 in which a value of the first object is decoded from coded elements received in act 106 from one or more of the nodes that responded with the maximum tag. This completes reading the value of the first object from the network.

[0064] Alternatively, if in act 110 the first network node determines that a quorum of nodes has not responded (i.e., the number of nodes that responded with the same maximum- 17 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026tag is at or below a predetermined threshold value), the method proceeds to act 114 in which the first network node propagates coded elements for the first object into the network. In some embodiments, the coded elements propagated in act 114 may include the most recent coded elements received by the first network node, and / or may include coded elements generated based on values decoded by the first network node from the most recent tag received.

[0065] In some embodiments, act 114 may comprise the first network node sending, to the list of nodes C that are responsible for the first object, the list of nodes C, the identifier of the first object, and a tag and its associated coded elements representing the most recent tag and associated coded elements for the first object that the first network node has previously read from the network. The first network node may subsequently receive changes to the network (e.g., as in the process of FIG. 4), which may be applied to its local estimate of the state machine describing the network as in act 108. These acts of sending data and receiving network changes in response may be repeated a plurality of times, and in some cases, may be repeated until the list of nodes C no longer changes.

[0066] FIG. 5 is a flowchart of a method of writing a value of a first object to a distributed memory, according to some embodiments. Method 500 may be performed by any suitable computer-executable code, processor, controller and / or computing system, including but not limited to the system(s) illustrated in FIG. 8, which is, or is a component of, a network node of a distributed network. References herein to the network node performing acts of method 500 should be considered to implicitly also refer to cases where certain components of a network node are performing those acts of method 500.

[0067] In act 502, the node performing method 500 (referred to below as the “first network node” in the description of FIG. 5 for convenience) identifies one or more network nodes that are responsible for the first object. Act 502 may be performed in any of the ways described above in relation to act 102 in method 100. For example, act 502 may comprise the first network node providing a list of node identifiers to a function 5), providing an object identifier associated with the first object being read to the function D, and calculating- 18 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026a distance between the value produced from the object identifier to each of the values produced from the node identifiers and / or may comprise the first network node accessing its local estimate of the state machine to obtain a list of participating nodes S (e.g., accessing a previously generated list of participating nodes S), and determining which of the nodes from S are responsible for the first object.

[0068] According to some embodiments, the network node performing method 500 may store a list of one or more object identifiers. This list of object identifiers may be accessed when determining which of the nodes in the network is responsible for maintaining that object.

[0069] In act 504, the first network node sends a query to each of the nodes identified in act 502, requesting that they provide data for the first object. Act 504 may be performed in any of the ways described above in relation to act 104 in method 100. For example, act 504 may comprise the first network node sending a query to each of the nodes that comprises an identifier of the first object and the list of participating nodes C responsible for the first object.

[0070] In act 506, the first network node receives a tag for the first object from one or more of the nodes to which queries were sent in act 504. In some cases, act 506 may also comprise receiving coded elements of the first object - however, since the coded elements are unnecessary to perform the remainder of the write operation it may be more efficient for the queried nodes to respond with a tag without associated coded elements. In some embodiments, the tag received in act 506 may be configured in any of the ways described above in relation to act 106. For instance, the tag may comprise data that indicates how recent the tag is, in addition to an identifier of a network node that generated the tag, which may be verified by the first network node.

[0071] According to some embodiments, act 506 comprises receiving one or more changes to the local estimate of the state machine of the first network node (e.g., one or more entries to add or remove certain nodes, as determined by the node receiving the query as shown in FIG. 4).- 19 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026

[0072] In act 508, the first network node applies any network changes received in act 506 to its local estimate of the state machine via any of the techniques described above in relation to act 108 in method 100. In the example of FIG. 5, when one or more network updates are received in act 506 and applied in act 508 to the local estimate of the state machine of the first network node, the method returns to act 502. This may be repeated until no network updates are received in act 506 before proceeding to act 510. In some embodiments, method 500 proceeds to act 510 only when the acts 502, 504, 506 and 508 are repeated for a preselected number of times without receiving any network updates (e.g., to increase the likelihood that the network has stabilized before proceeding). Performing the acts 502, 504, 506 and 508 again may allow the first network node to cope with the dynamic nature of the network.

[0073] In act 510, the first network node identifies the most recent tag among the tags received in act 506 via any of the techniques described above in relation to act 110 of method 100. Subsequently, the first network node generates a new tag that is more recent than the most recent tag received in act 506. For instance, if the tag comprises a version number, act 510 may comprise the first network node incrementing that number (e.g., by 1 or some other fixed number).

[0074] In some embodiments, act 510 comprises the first network node generating a signature based on the tag generated in act 510 and based on values for the first object being written to the network. For example, the new tag may comprise a new tag version number and an identifier of the first network node. The first network node may then generate a signature from the tag-value pair for the first object (e.g., coded elements encoded using LNC or RLNC), e.g., by passing the tag-value pair (the tag and values of the first object) to a preconfigured signing function that generates a signature in response.

[0075] In act 512, the first network node propagates the values being written to the network. In some embodiments, act 512 may comprise the first network node sending to the list of nodes C that are responsible for the first object, the list of nodes C, the identifier of the first object, and a tag and its associated coded elements representing the tag generated in act- 20 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026508 and associated coded elements for the first object. The first network node may subsequently receive changes to the network (e.g., as in the process of FIG. 4), which it may apply to its local estimate of the state machine describing the network, as in act 508. These acts of sending data and receiving network changes in response may be repeated a plurality of times, and in some cases, may be repeated until the list of nodes C no longer changes.

[0076] FIG. 6 depicts a process of joining a network, which is also visually presented in FIGs. 7A-7E, according to some embodiments. In process 600 the node 201 joins a distributed network, which is configured via the techniques described above. In step 621, also shown in FIG. 7A, the node 201 sends a request to other nodes in the network requesting information on the objects that it will be responsible for storing. In some embodiments, an identifier of node 201 is generated by node 201 and provided to the node 204 as part of the request in step 621. In some embodiments, process 600 may comprise an initial step of the node 201 sending a request to one or more nodes requesting a list of nodes in the network to generate an initial configuration of its local estimate of the state machine.

[0077] According to some embodiments, node 201 may send the request in step 621 to a subset of the nodes in the network. In particular, the node 201 may select a subset of nodes by determining a distance between values produced from the function 5) when input the respective node identifiers, including the node identifier of node 201. The subset may be selected as those nodes which have a distance to node 201 that is less than some threshold distance. For instance, the node 201 may identify nodes that are within a particular distance from its position along the ring shown in FIG. 2. In the example of FIG. 2 and FIGs. 7A-7E, the distance along the ring is extended along the ring in both directions so that the closest neighbors to node 201 (in terms of the values © produces from their node identifiers) are nodes 202, 204, 206 and 209. The distance of values around the ring that may be considered for this step in process 600 may be tuned as desired.

[0078] In step 622, the node 204 adds node 201 to a list of joining nodes, which may be a separate list to the list of participating nodes described above. Rather, this list indicates those nodes that are in the process of joining the network. It will be appreciated that in the- 21 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026example of FIGs. 7A-7E, nodes 202, 206 and 209 would also perform the same acts as node 204, and that node 204 is used in FIG. 6 merely as an example.

[0079] In step 623, the node 204 determines the objects that the node 201 will be responsible for. In some embodiments, the node 204 determines the value produced by a distance function 2) for each of a plurality of object identifiers (e.g., that the node 204 stores in a list) and for an identifier of node 201 (e.g., as received from the node 201 in step 621). As described above in relation to act 102 of method 100, a node may be responsible for an object when the distance between the values produced by the distance function © for the node’s identifier and the object’s identifier is less than a threshold distance. Accordingly, in step 623, the node 204 may determine which of the object identifiers in its list of object identifiers meet this criteria for the node 201, and thereby identify the objects that node 201 will be responsible for.

[0080] In step 624, the node 204 sends to node 201 a list of the objects (e.g., the object identifiers) determined in step 623. In addition, the node 204 may send a tag and coded elements for each object identified in the list being sent. In some embodiments, the node 204 sends a plurality of tags and their associated coded elements for each object, representing the most recent collection of tags and coded elements that the node 204 received for that object. This step is also depicted in FIG. 7B.

[0081] According to some embodiments, step 624 also comprises sending one or more changes to the local estimate of the state machine of the node 201. For instance, the request in step 621 may comprise a list of nodes as in step 421 of FIG. 4, and the node 204 may perform step 422 before sending determined changes to the nodes in step 423 along with the objects and object data in step 624.

[0082] In step 625, the node 201 recodes the objects for which it is responsible based on coded elements received from one or more nodes (e.g., in step 624 from node 204). For instance, if the node 201 has collected a sufficient number of replies from the one or more nodes that comprise coded elements for an object, the node 201 can generate new coded- 22 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026elements for that object as described above in relation to LNC and RLNC techniques. This step is also depicted in FIG. 7C.

[0083] In some embodiments, node 201 performs step 625 for a given object only if the number of tags received in step 624 that are all the most recent tag for that object is at least 7~] for some ? < k that is sufficiently large for node 201 to recode the object.

[0084] In step 626, the node 201 adds itself to the state machine. Updating the state machine in step 626 may, for instance, comprise sending an add request to the state machine, or otherwise calling an operation on the state machine that adds node 201. This step is also depicted in FIG. 7C. In some embodiments, step 626 comprises the node 201 performing additional updates to the local estimate of the state machine of the node 201 based on network changes received in step 624.

[0085] In step 627, the node 201 sends a message to the neighboring nodes identified in step 621 (e.g., node 204) that the join process was performed successfully. This step is also depicted in FIG. 7D. In step 628, each of the neighboring nodes (e.g., node 204) adds node 201 to its local estimate of the state machine. This step is also depicted in FIG. 7E, wherein nodes 202, 204, 206 and 209 add node 201 to their local estimates 202a, 204a, 206a and 209a, respectively. Updating the local estimate of the state machine in step 628 may, for instance, comprise appending an add entry to the local estimate, or generating a new list of nodes that includes node 201. In some embodiments, step 628 comprises removing node 201 from the list of joining nodes to which it was added in step 622.

[0086] In some embodiments, before adding the node 201 to its local estimate of the state machine, each of the neighboring nodes (e.g., node 204) sends a request to node 201 for changes to their local estimates, in order to verify that node 201 was properly added to their local estimates. For example, a request such as that shown in step 421 in FIG. 4 may be sent from node 204 to node 201, including a list of participating network nodes indicated by node 204’ s local estimate of the state machine, and receiving changes from node 201 indicating that node 201 should be added to node 204’ s local estimate of the state machine.- 23 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026

[0087] According to some embodiments, a node may depart a network by, for each object the node is responsible for, sending a message to each of the other nodes responsible for that object indicating that the node is departing the network. Any nodes receiving such a message may modify its local estimate of the state machine to remove the departing node. The departing node may also optionally send a message to the state machine to request its removal. In some implementations the receiving node may, if it does not current have a value for the object, read the value of the object from other nodes responsible for the object (e.g., by performing method 100, or a part thereof). Allowing for a graceful departure in this manner may be preferable to nodes simply leaving, which may be indistinguishable from a crash or other failure.

[0088] An illustrative implementation of a computer system 800 that may be used to control a network node, or a portion thereof, to perform any of the techniques described above is shown in FIG. 8. The computer system 800 may include one or more processors 810 and one or more non-transitory computer-readable storage media (e.g., memory 820 and one or more non-volatile storage media 830), in addition to a communication interface 840. The one or more processors 810 may control writing data to and reading data from the memory 820 and the one or more non-volatile storage media 830 in any suitable manner, as the aspects of the disclosure described herein are not limited in this respect. To perform functionality and / or techniques described herein, the one or more processors 810 may execute one or more instructions stored in one or more computer-readable storage media (e.g., the memory 820, storage media, etc.), which may serve as non-transitory computer-readable storage media storing instructions for execution by the one or more processors 810. Communication interface 840 provides an interface for the computer system 800 to transmit data to other suitable computer systems (e.g., from one network node to another network node). For example, communication interface 840 may comprise a local area network (LAN) device and / or a wireless communication device.

[0089] In connection with techniques described herein, code used to, for example, update a state machine, determine a distance value based on an identifier, identify network- 24 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026nodes responsible for a given object, generate a coded element, decode a message from a plurality of coded elements, etc. may be stored on one or more computer-readable storage media of computer system 800. The one or more processors 810 may execute any such code to perform any of the above-described techniques as described herein. Any other software, programs or instructions described herein may also be stored and executed by computer system 800. It will be appreciated that computer code may be applied to any aspects of methods and techniques described herein. For example, computer code may be applied to perform method 100, process 400, method 500, process 600, etc.

[0090] The various methods or processes outlined herein may be coded as software that is executable on one or more processors that employ any one of a variety of operating systems or platforms. Additionally, such software may be written using any of numerous suitable programming languages and / or programming or scripting tools, and also may be compiled as executable machine language code or intermediate code that is executed on a virtual machine or a suitable framework.

[0091] In this respect, various inventive concepts may be embodied as at least one non-transitory computer readable storage medium (e.g., a computer memory, one or more floppy discs, compact discs, optical discs, magnetic tapes, flash memories, circuit configurations in Field Programmable Gate Arrays or other semiconductor devices, etc.) encoded with one or more programs that, when executed on one or more computers or other processors, implement the various embodiments of the present disclosure. The non-transitory computer-readable medium or media may be transportable, such that the program or programs stored thereon may be loaded onto any computer resource to implement various aspects of the present disclosure as described above.

[0092] The terms “program,” “software,” and / or “application” are used herein in a generic sense to refer to any type of computer code or set of computer-executable instructions that can be employed to program a computer or other processor to implement various aspects of embodiments as described above. Additionally, it should be appreciated that according to one aspect, one or more computer programs that when executed perform- 25 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026methods of the present disclosure need not reside on a single computer or processor, but may be distributed in a modular fashion among different computers or processors to implement various aspects of the present disclosure.

[0093] Computer-executable instructions may be in many forms, such as program modules, executed by one or more computers or other devices. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform particular tasks or implement particular abstract data types. Typically, the functionality of the program modules may be combined or distributed as desired in various embodiments.

[0094] Also, data structures may be stored in non-transitory computer-readable storage media in any suitable form. Data structures may have fields that are related through location in the data structure. Such relationships may likewise be achieved by assigning storage for the fields with locations in a non-transitory computer-readable medium that convey relationship between the fields. However, any suitable mechanism may be used to establish relationships among information in fields of a data structure, including through the use of pointers, tags or other mechanisms that establish relationships among data elements.

[0095] Having thus described several aspects of at least one embodiment of this disclosure, it is to be appreciated that various alterations, modifications, and improvements will readily occur to those skilled in the art. For instance, aspects of the techniques described herein may be combined in any of the following ways:

[0096] Aspect 1. A method comprising: reading a value of a first object from a distributed memory by: identifying one or more network nodes responsible for the first object; receiving a tag- value pair from each of the one or more network nodes responsible for the first object, thereby receiving a plurality of tag- value pairs, wherein each tag- value pair comprises a tag number and a plurality of coded elements of the first object; identifying one or more of the plurality of tag- value pairs that comprise a most recent tag number among the plurality of tag-value pairs; and decoding the value of the first object from the plurality of coded elements of the first object in the one or more of the plurality of tag- value pairs that comprise the most recent tag number.- 26 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026

[0097] Aspect 2. The method of aspect 1, wherein identifying the one or more network nodes responsible for the first object comprises: receiving one or more network changes from one or more of a first plurality of network nodes; and updating a list of participating network nodes based on the one or more network changes.

[0098] Aspect 3. The method of any of aspects 1-2, wherein identifying the one or more network nodes responsible for the first object comprises: querying the first plurality of network nodes for the value of the first object; and subsequent to updating the list of participating network nodes, querying a second plurality of network nodes for the value of the first object, the second plurality of network nodes being different from the first plurality of network nodes.

[0099] Aspect 4. The method of any of aspects 1-3, further comprising identifying the first plurality of network nodes by identifying one or more network nodes responsible for the first object among network nodes identified by the list of participating network nodes.

[0100] Aspect 5. The method of any of aspects 1-4, wherein identifying the one or more network nodes responsible for the first object comprises: determining a pseudodistance with respect to the first object for each of the network nodes identified by the list of participating network nodes; and identifying the one or more network nodes responsible for the first object as the network nodes identified by the list of participating network nodes having a pseudo-distance with respect to the first object below a pseudo-distance threshold.

[0101] Aspect 6. The method of any of aspects 1-5, wherein the pseudo-distance with respect to the first object, for a respective network node of each of the network nodes identified by the list of participating network nodes, is generated by a hash function based on the first object and the respective network node.

[0102] Aspect 7. The method of any of aspects 1-6, wherein decoding the value of the first object from the plurality of coded elements of the first object comprises decoding the value of the first object via random linear network coding.- 27 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026

[0103] Aspect 8. The method of any of aspects 1-7, further comprising determining that a count of the one or more of the plurality of tag-value pairs that comprise the most recent tag number is greater than a quorum threshold, before decoding the value of the first object.

[0104] Aspect 9. The method of any of aspects 1-8, further comprising, in response to determining that a count of the one or more of the plurality of tag-value pairs that comprise the most recent tag number is less than a quorum threshold, sending a plurality of coded elements of the first object to each of the one or more network nodes responsible for the first object.

[0105] Aspect 10. The method of any of aspects 1-9, further comprising receiving, in addition to the tag-value pair, an identifier of each of the one or more network nodes responsible for the first object.

[0106] Aspect 11. The method of any of aspects 1-10, further comprising: receiving a signature from each of the one or more network nodes responsible for the first object, in addition to the tag-value pair and the identifier; and verifying that the signature was validly generated.

[0107] Aspect 12. A method comprising: writing a value of a first object to a distributed memory by: identifying one or more network nodes responsible for the first object; receiving a tag number from each of the one or more network nodes responsible for the first object, thereby receiving a plurality of tag numbers; and sending, to each of the one or more network nodes responsible for the first object: a plurality of coded elements of the first object; and a new tag number generated based on a most recent tag number among the plurality of tag numbers.

[0108] Aspect 13. The method of aspect 12, wherein identifying the one or more network nodes responsible for the first object comprises: receiving one or more network changes from one or more of a first plurality of network nodes; and updating a list of participating network nodes based on the one or more network changes.- 28 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026

[0109] Aspect 14. The method of any of aspects 12-13, herein identifying the one or more network nodes responsible for the first object comprises: querying the first plurality of network nodes for the value of the first object; and subsequent to updating the list of participating network nodes, querying a second plurality of network nodes for the value of the first object, the second plurality of network nodes being different from the first plurality of network nodes.

[0110] Aspect 15. The method of any of aspects 12-14, further comprising identifying the first plurality of network nodes by identifying one or more network nodes responsible for the first object among network nodes identified by the list of participating network nodes.

[0111] Aspect 16. The method of any of aspects 12-15, wherein identifying the one or more network nodes responsible for the first object comprises: determining a pseudodistance with respect to the first object for each of the network nodes identified by the list of participating network nodes; and identifying the one or more network nodes responsible for the first object as the network nodes identified by the list of participating network nodes having a pseudo-distance with respect to the first object below a pseudo-distance threshold.

[0112] Aspect 17. The method of any of aspects 12-16, wherein the pseudo-distance with respect to the first object, for a respective network node of each of the network nodes identified by the list of participating network nodes, is generated by a hash function based on the first object and the respective network node.

[0113] Aspect 18. The method of any of aspects 12-17, further comprising receiving, in addition to the tag number, an identifier of each of the one or more network nodes responsible for the first object.

[0114] Aspect 19. The method of any of aspects 12-18, further comprising: receiving a signature from each of the one or more network nodes responsible for the first object, in addition to the tag number and the identifier; and verifying that the signature was validly generated.- 29 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026

[0115] Aspect 20. The method of any of aspects 12-19, further comprising generating the plurality of coded elements of the first object by encoding the value of the first object via random linear network coding.

[0116] Such alterations, modifications, and improvements are intended to be part of this disclosure and are intended to be within the spirit and scope of the disclosure. Further, though advantages of the present disclosure are indicated, it should be appreciated that not every embodiment of the technology described herein will include every described advantage. Some embodiments may not implement any features described as advantageous herein and in some instances one or more of the described features may be implemented to achieve further embodiments. Accordingly, the foregoing description and drawings are by way of example only.

[0117] Aspects of the above-described embodiments of the technology described herein can be implemented in any of numerous ways. For example, aspects of the embodiments may be implemented using hardware, software, or a combination thereof. When implemented in software, the software code can be executed on any suitable processor or collection of processors, whether provided in a single computer or distributed among multiple computers. Such processors may be implemented as integrated circuits, with one or more processors in an integrated circuit component, including commercially available integrated circuit components known in the art by names such as CPU chips, GPU chips, microprocessor, microcontroller, or co-processor. Alternatively, a processor may be implemented in custom circuitry, such as an ASIC, or semi-custom circuitry resulting from configuring a programmable logic device. As yet a further alternative, a processor may be a portion of a larger circuit or semiconductor device, whether commercially available, semicustom or custom. As a specific example, some commercially available microprocessors have multiple cores such that one or a subset of those cores may constitute a processor. Though, a processor may be implemented using circuitry in any suitable format.

[0118] Various aspects of the present disclosure may be used alone, in combination, or in a variety of arrangements not specifically described in the embodiments described in the- 30 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026foregoing and is therefore not limited in its application to the details and arrangement of components set forth in the foregoing description or illustrated in the drawings. For example, aspects described in one embodiment may be combined in any manner with aspects described in other embodiments.

[0119] Also, aspects of the disclosure may be embodied as a method, of which examples have been provided. The acts performed as part of the method may be ordered in any suitable way. Accordingly, embodiments may be constructed in which acts are performed in an order different than illustrated, which may include performing some acts simultaneously, even though shown as sequential acts in illustrative embodiments.

[0120] Use of ordinal terms such as “first,” “second,” “third,” etc., in the claims to modify a claim element does not by itself connote any priority, precedence, or order of one claim element over another or the temporal order in which acts of a method are performed, but are used merely as labels to distinguish one claim element having a certain name from another element having a same name (but for use of the ordinal term) to distinguish the claim elements.

[0121] The terms “approximately” and “about” may be used to mean within ±20% of a target value in some embodiments, within ±10% of a target value in some embodiments, within ±5% of a target value in some embodiments, and yet within ±2% of a target value in some embodiments. The terms “approximately” and “about” may include the target value. The term “substantially equal” may be used to refer to values that are within ±20% of one another in some embodiments, within ±10% of one another in some embodiments, within ±5% of one another in some embodiments, and yet within ±2% of one another in some embodiments.

[0122] The term “substantially” may be used to refer to values that are within ±20% of a comparative measure in some embodiments, within ±10% in some embodiments, within ±5% in some embodiments, and yet within ±2% in some embodiments. For example, a first direction that is “substantially” perpendicular to a second direction may refer to a first direction that is within ±20% of making a 90° angle with the second direction in some - 31 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026embodiments, within ±10% of making a 90° angle with the second direction in some embodiments, within ±5% of making a 90° angle with the second direction in some embodiments, and yet within ±2% of making a 90° angle with the second direction in some embodiments.

[0123] Also, the phraseology and terminology used herein is for the purpose of description and should not be regarded as limiting. The use of “including,” “comprising,” or “having,” “containing,” “involving,” and variations thereof herein, is meant to encompass the items listed thereafter and equivalents thereof as well as additional items.

[0124] What is claimed is:- 32 - ACTIVE 710940979v1

Claims

Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026CLAIMS1. A method comprising:reading a value of a first object from a distributed memory by:identifying one or more network nodes responsible for the first object; receiving a tag-value pair from each of the one or more network nodes responsible for the first object, thereby receiving a plurality of tag- value pairs, wherein each tag-value pair comprises a tag number and a plurality of coded elements of the first object;identifying one or more of the plurality of tag-value pairs that comprise a most recent tag number among the plurality of tag-value pairs; anddecoding the value of the first object from the plurality of coded elements of the first object in the one or more of the plurality of tag- value pairs that comprise the most recent tag number.

2. The method of claim 1, wherein identifying the one or more network nodes responsible for the first object comprises:receiving one or more network changes from one or more of a first plurality of network nodes; andupdating a list of participating network nodes based on the one or more network changes.

3. The method of claim 2, wherein identifying the one or more network nodes responsible for the first object comprises:querying the first plurality of network nodes for the value of the first object; and subsequent to updating the list of participating network nodes, querying a second plurality of network nodes for the value of the first object, the second plurality of network nodes being different from the first plurality of network nodes.- 33 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 20264. The method of claim 2, further comprising identifying the first plurality of network nodes by identifying one or more network nodes responsible for the first object among network nodes identified by the list of participating network nodes.

5. The method of claim 4, wherein identifying the one or more network nodes responsible for the first object comprises:determining a pseudo-distance with respect to the first object for each of the network nodes identified by the list of participating network nodes; andidentifying the one or more network nodes responsible for the first object as the network nodes identified by the list of participating network nodes having a pseudo-distance with respect to the first object below a pseudo-distance threshold.

6. The method of claim 5, wherein the pseudo-distance with respect to the first object, for a respective network node of each of the network nodes identified by the list of participating network nodes, is generated by a hash function based on the first object and the respective network node.

7. The method of claim 1, wherein decoding the value of the first object from the plurality of coded elements of the first object comprises decoding the value of the first object via random linear network coding.

8. The method of claim 1, further comprising determining that a count of the one or more of the plurality of tag-value pairs that comprise the most recent tag number is greater than a quorum threshold, before decoding the value of the first object.

9. The method of claim 1, further comprising, in response to determining that a count of the one or more of the plurality of tag-value pairs that comprise the most recent tag- 34 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026number is less than a quorum threshold, sending a plurality of coded elements of the first object to each of the one or more network nodes responsible for the first object.

10. The method of claim 1, further comprising receiving, in addition to the tag- value pair, an identifier of each of the one or more network nodes responsible for the first object.

11. The method of claim 10, further comprising:receiving a signature from each of the one or more network nodes responsible for the first object, in addition to the tag-value pair and the identifier; andverifying that the signature was validly generated.

12. A method comprising:writing a value of a first object to a distributed memory by:identifying one or more network nodes responsible for the first object; receiving a tag number from each of the one or more network nodes responsible for the first object, thereby receiving a plurality of tag numbers; and sending, to each of the one or more network nodes responsible for the first object:a plurality of coded elements of the first object; anda new tag number generated based on a most recent tag number among the plurality of tag numbers.

13. The method of claim 12, wherein identifying the one or more network nodes responsible for the first object comprises:receiving one or more network changes from one or more of a first plurality of network nodes; andupdating a list of participating network nodes based on the one or more network changes.- 35 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 202614. The method of claim 13, herein identifying the one or more network nodes responsible for the first object comprises:querying the first plurality of network nodes for the value of the first object; and subsequent to updating the list of participating network nodes, querying a second plurality of network nodes for the value of the first object, the second plurality of network nodes being different from the first plurality of network nodes.

15. The method of claim 13, further comprising identifying the first plurality of network nodes by identifying one or more network nodes responsible for the first object among network nodes identified by the list of participating network nodes.

16. The method of claim 15, wherein identifying the one or more network nodes responsible for the first object comprises:determining a pseudo-distance with respect to the first object for each of the network nodes identified by the list of participating network nodes; andidentifying the one or more network nodes responsible for the first object as the network nodes identified by the list of participating network nodes having a pseudo-distance with respect to the first object below a pseudo-distance threshold.

17. The method of claim 16, wherein the pseudo-distance with respect to the first object, for a respective network node of each of the network nodes identified by the list of participating network nodes, is generated by a hash function based on the first object and the respective network node.

18. The method of claim 12, further comprising receiving, in addition to the tag number, an identifier of each of the one or more network nodes responsible for the first object.

19. The method of claim 18, further comprising:- 36 - ACTIVE 710940979v1Attorney Docket No. 234495-700120 / PCT Electronic Filing Date: May 15, 2026receiving a signature from each of the one or more network nodes responsible for the first object, in addition to the tag number and the identifier; andverifying that the signature was validly generated.

20. The method of claim 12, further comprising generating the plurality of coded elements of the first object by encoding the value of the first object via random linear network coding.- 37 - ACTIVE 710940979v1