System and Method for Storing Files in a Distributed Environment

• US 20120303631A1
• Filed: 05/24/2012
• Published: 11/29/2012
• Est. Priority Date: 05/24/2011
• Status: Active Grant

• Current Assignee: Red Lambda Incorporated
• Original Assignee: Red Lambda Incorporated
• Inventors: Bird, Robert, Whitlock, Matthew
• Granted Patent: US 8,732,140 B2
• Time in Patent Office: Days
• US Class Current: 707/741
• CPC Class Codes: G06F 16/13 File access structures, e.g. distributed indices arrangements of input from, or output to, record carriers G06F3/06
• CPC Class Codes: G06F 16/182 Distributed file systems
• CPC Class Codes: G06F 16/2255 Hash tables
• CPC Class Codes: G06F 16/2272 Management thereof

Abstract

A system and method for storing data-files stored on distributed devices connected to a network. Data-elements of the data-files are allocated to data-blocks stored on the distributed devices. Key-identifiers are calculated for each of the data-blocks based on the allocated data-elements. The key-identifiers are stored in distributed tables stored on the distributed devices. Index-nodes are generated for the data-files based on the data-blocks. A Paxos algorithm is executed for the index-nodes based on the key-identifiers to provide a consensus of the data-files that are stored on the plurality of distributed devices.

Claims

1. A method for storing data, comprising the steps of:

allocating data-elements of stored data-files to data-blocks, wherein the stored data-files are stored on a plurality of distributed devices, wherein the plurality of distributed devices are connected to a network, wherein the data-blocks are stored on the plurality of distributed devices;

calculating key-identifiers for each of the data-blocks, wherein the key-identifiers are based on the allocated data-elements of the corresponding data-blocks;

storing the key-identifiers in distributed tables, wherein the distributed tables are stored on the plurality of distributed devices;

generating index-nodes for the data-files based on the data-blocks, wherein the index-nodes comprise a list of the data-blocks, wherein the index-nodes are stored in a metadata-store; and

,

executing a Paxos algorithm for the index-nodes based on the key-identifiers to provide a consensus of the data-files that are stored on the plurality of distributed devices.

2. A method of claim 1, wherein the key-identifiers comprise hash values, wherein the distributed tables are distributed hash tables.

3. A method of claim 1, wherein the data-blocks are binary large objects, wherein each of the binary large objects is less than a maximum predetermined size.

4. A method of claim 3, wherein each of the binary large objects is associated with a unique address, wherein each unique address is based on the allocated data-elements of the corresponding binary large object.

5. A method of claim 1, wherein the stored data-files are allocated to data-blocks based on a Kademlia algorithm.

6. A method of claim 1, wherein the data-blocks are logically grouped together, and wherein each one of the data-blocks are read and written as a single unit.

7. A method of claim 1, further comprising the step of:

generating directory index-nodes for a plurality of index-nodes, wherein the directory index-nodes are stored in the metadata-store.

8. A method of claim 1, wherein the steps are performed by a plurality of processors operably connected to the distributed devices and adapted to communicate across the network.

9. A method of claim 1, wherein the key-identifiers are globally unique block-identifiers.

10. A method of claim 1, further comprising the steps of:

receiving a request to modify an index-node;

executing a Paxos algorithm for a quorum of nodes that correspond to the data-blocks having near-values in the distributed table, wherein the near-values are hash values that are within a predetermined range of a hash value of the index-node;

modifying the index-node based on an outcome of the Paxos algorithm; and

,

modifying the nodes in the quorum based on the modified index-node, wherein a consensus of the nodes is provided based on the outcome of the Paxos algorithm.

11. A method of claim 10, wherein the consensus is provided by executing voting-rounds for a quorum of nodes.

12. A method of claim 11, wherein the voting-rounds comprise the steps of:

sending round-invitations to the nodes in the quorum, wherein the nodes in the quorum are acceptors;

receiving replies from each acceptor, wherein the replies are round-acceptances comprising a last-voted-value and a round-number, wherein the round-number represents the voting-round in which the corresponding acceptor voted the last-voted-value; and

,

selecting a value based the last-voted-values received from the acceptors.

13. A method of claim 12, further comprising the steps of:

sending vote-invitations to the acceptors, wherein the vote-invitations comprise the selected value.

14. A method of claim 12, wherein the value is selected upon receiving the last-voted-values from a majority of the acceptors.

15. A method of claim 12, further comprising the steps of:

sending a value-announcement from the acceptors to the nodes in the quorum, wherein the value-announcement comprises the selected value, wherein the nodes in the quorum reject a value that is different from the selected value.

16. A method of claim 1, wherein at least one of the index-nodes is encoded based on an Abstract Syntax Notation One (ASN.1) encoding standard.

17. A method of claim 1, wherein at least one of the index-nodes is encoded based on a Distinguished Encoding Rules (DER) message-transfer syntax.

18. A method of claim 17, wherein the Distinguished Encoding Rules (DER) message-transfer syntax is based on a X.690 International Telecommunication Union (ITU)-T recommendation.

19. A method of claim 1, further comprising the steps of:

receiving a round-invitation by an acceptor from a coordinator, wherein the round-invitation comprises a generation-number;

sending an inquiry by the acceptor to the coordinator, only if the generation-number in the round-invitation is greater than a generation-number of the acceptor;

sending a notification by the acceptor to the coordinator, only if the generation-number in the round-invitation is less than the generation-number of the acceptor; and

,

only if the generation-number in the round-invitation is equal to the generation-number of the acceptor, then;

(i) sending a round-begun notification by the acceptor to the coordinator, only if a round-number in the round-invitation is less than the record-number of the acceptor;

(ii) ignoring the round-invitation by the acceptor, only if the round-number in the round-invitation is equal to the round-number of the acceptor; and

,

(iii) updating the round-number in the acceptor to match the round-number in the round-invitation and sending a round-acceptance to the coordinator, only if the round-number in the round-invitation is greater than the round-number in the acceptor.

20. A method of claim 1, further comprising the steps of:

receiving a round-acceptance by a coordinator from an acceptor, wherein the round-acceptance comprises a generation-number;

ignoring the round-acceptance by the coordinator, only if the generation-number in the round-acceptance is different from the generation-number of the coordinator; and

,

only if the generation-number of the coordinator is equal to the generation-number of the round-acceptance, then;

(i) ignoring the round-acceptance by coordinator, only if the round-number in the round-acceptance is different from the round-number on the coordinator; and

,

(ii) only if the round-number in the round-acceptance is equal to the round-number on the coordinator, then;

(a) ignoring the round-acceptance by coordinator, only if the coordinator has already received a round-acceptance from the acceptor in the current generation and round; and

,

(b) tracking by the coordinator a last-voted-value in the round-acceptance received from the acceptor, only if the coordinator has not already received a round-acceptance from the acceptor in the current generation and round.

21. A method of claim 1, further comprising the steps of:

receiving a round-acceptance by a coordinator from a majority of acceptors, wherein the majority of acceptors are acceptors that received round-invitations from the coordinator, wherein the round-acceptance comprises a generation-number;

selecting, by the coordinator, a last-voted-value, wherein the last-voted-value is a value that the coordinator received from the acceptor in the highest-numbered round;

only if the acceptors failed to vote in the current generation, selecting a proposed-value by the coordinator; and

,

sending, by the coordinator, a vote-invitation to each of the acceptors, wherein the vote-invitation is the proposed-value only if the acceptors failed to vote in the current generation, wherein the vote-invitation is the last-voted-value only if the acceptors voted in the current generation.

22. A method of claim 1, further comprising the steps of:

receiving a vote-invitation by an acceptor from a coordinator, wherein the vote-invitation comprises a generation-number;

ignoring the vote-invitation by the acceptor, only if the generation-number in the vote-invitation is different from the generation-number on the acceptor; and

,

only if the generation-number of the acceptor is equal to the generation-number of the vote-acceptance, then;

(i) sending a round-begun notification by the acceptor to the coordinator, only if a round-number in the vote-invitation is less than the record-number of the acceptor; and

,

(ii) only if a round-number in the vote-invitation is equal to the record-number of the acceptor, then;

(a) ignoring the vote-acceptance by acceptor, only if the acceptor has already voted in the current generation and round; and

,

(b) sending a vote-announcement from the acceptor to learners listed in the vote-invitation, only if the acceptor has not already voted in the current generation and round.

23. A method of claim 1, further comprising the steps of:

receiving a vote-announcement by a learner from an acceptor, wherein the vote-announcement comprises a generation-number;

ignoring the vote-announcement by the learner, only if the generation-number in the vote-announcement is different from the generation-number of the learner; and

only if the generation-number of the learner is equal to the generation-number of the vote-announcement, then;

(i) ignoring the vote-announcement by the learner, only if the round-number in the vote-announcement is less than the round-number of the learner; and

,

(ii) only if a round-number in the vote-announcement is equal to or greater than the record-number of the learner, then;

(a) ignoring the vote-acceptance by learner, only if the learner has already received a vote-announcement from the acceptor in the current generation and round; and

,

(b) tracking by the learner the vote received from the acceptor, only if the learner has not already received a vote-announcement from the acceptor in the current generation and round.

24. A method of claim 1, further comprising the steps of:

receiving a vote-announcement by a learner from a majority of acceptors, wherein the vote-announcements have a pre-determined majority-number for determining a quantity needed for the majority of acceptors, wherein the vote-announcements comprise a generation-number and a value; and

,

only if the values of all of the vote-announcements are equal, then the learner tracks the value of the vote-announcements and the learner increments the generation-number for a next-iteration of the Paxos algorithm.

25. A method of claim 1, further comprising the steps of:

receiving a round-begun notification by the coordinator to the acceptor;

ignoring the round-begun notification by the coordinator, only if the generation-number in the round-begun notification is different from the generation-number on the coordinator; and

,

only if the generation-number in the round-begun notification is equal to the generation-number on the coordinator, then;

(i) ignoring the round-begun notification by the coordinator, only if the round-number in the round-begun notification is less than or equal to the round-number on the coordinator; and

,

(ii) selecting, by the coordinator, a new round-number that is greater than the round-number in the round-begun notification and sending a round-invitation to each of the acceptors, only if the round-number in the round-begun notification is greater than the round-number on the coordinator.

26. A method of claim 1, further comprising the steps of:

receiving an inquiry from a sender by a receiver, wherein the sender is a sending-node, wherein the receiver is a receiving-node;

sending a notification by the receiver to the sender, wherein the notification comprises a generation-number and a value of the receiver; and

,

only if the sender receives the notification, then;

(i) ignoring the notification by the receiver, only if the generation-number in the notification is less than the generation-number of the receiver; and

,

(ii) only if the generation-number in the notification is equal to or greater than the generation-number on the receiver, then the receiver tracks the value in the notification and the receiver sets the generation-number in the receiver to a number equal to one value greater than the generation-number in the notification for the next-iteration of the Paxos algorithm.

27. A system for storing data, comprising:

data-blocks stored on a plurality of distributed devices, the plurality of distributed devices are adapted to be connected via a network, the plurality of distributed devices store data-files, the data-files comprising data-elements;

the stored data-elements are allocated to the data-blocks;

key-identifiers calculated for each of the data-blocks, the key-identifiers are based on the allocated data-elements of the corresponding data-blocks;

distributed tables that store the key-identifiers, the distributed tables are stored on the plurality of distributed devices;

index-nodes for the data-files generated based on the data-blocks, the index-nodes comprising a list of the data-blocks, the index-nodes are stored in a metadata-store; and

,

a Paxos algorithm executed for the index-nodes based on the key-identifiers to provide a consensus of the data-files that are stored on the plurality of distributed devices.