System and method for facilitating mitigation of read/write amplification in data compression

Abstract

The system can receive data to be written to a non-volatile memory in the distributed storage system. The received data can include a plurality of input segments. The system can assign consecutive logical block addresses (LBAs) to the plurality of input segments. The system can then compress the plurality of input segments to generate a plurality of fixed-length compressed segments, with each fixed-length compressed segment aligned with a physical block address (PBA) in a set of PBAs. The system compresses the plurality of input segments to enable an efficient use of storage capacity in the non-volatile memory. Next, the system can write the plurality of fixed-length compressed segments to a corresponding set of PBAs in the non-volatile memory. The system can then create, in a data structure, a set of entries which map the LBAs of the input segments to the set of PBAs. This data structure can be used later by the system when processing a read request including a LBA.

Description

BACKGROUND

Field

This disclosure is generally related to the field of data storage. More specifically, this disclosure is related to a system and method for facilitating mitigation of read/write amplification when performing data compression in a data storage system.

Related Art

The proliferation of the Internet and e-commerce continues to create a vast amount of digital content. Today, various distributed storage systems have been created to access and store the ever-increasing amount of digital content. However, network bandwidth and storage capacity of physical resources are two characteristics of distributed storage systems which can greatly impact their performance, cost, and efficiency.

Even with the addition of storage capacity to a distributed storage system, the physical bandwidth can still only support a limited number of users while meeting the requirements of a Service Level Agreement (SLA). For example, when a storage system experiences a heavy load of simultaneous incoming traffic, some drives may become non-responsive due to a lack of sufficient bandwidth, even if sufficient storage capacity is available.

Data compression techniques have been used in distributed storage systems to save storage capacity and to reduce the amount of data transferred, thus enabling the efficient use of storage capacity and communication bandwidth. However, efficiency of the compression techniques has become increasingly critical with the increase in amount of digital content. The existing data compression techniques are inefficient due to the overhead from high read and write amplifications inherent in their data processing operations. Therefore, some challenges still remain in designing an efficient data compression technique that is capable of providing an improved performance of storage systems with regards to latency and an improved efficiency with respect to resource consumption, network load, read/write amplification, etc.

SUMMARY

One embodiment of the present disclosure provides a system and method for facilitating data compression in a distributed storage system. During operation, the system can receive data to be written to a non-volatile memory in the distributed storage system. The received data can include a plurality of input segments. The system can assign consecutive logical block addresses (LBAs) to the plurality of input segments. The system can then compress the plurality of input segments, e.g., by applying a data compression technique, to generate a plurality of fixed-length compressed segments, with each fixed-length compressed segment aligned with a physical block address (PBA) in a set of PBAs. The system compresses the plurality of input segments to enable an efficient use of storage capacity in the non-volatile memory. Next, the system can write the plurality of fixed-length compressed segments to a corresponding set of PBAs in the non-volatile memory. The system can then create, in a data structure, a set of entries which map the LBAs of the input segments to the set of PBAs. This data structure can be used later by the system when processing a read request including a LBA.

In some embodiments, the system can compress the plurality of input segments by: reading sequentially a subset of the plurality of input segments into a sliding window, wherein the subset includes one or more of the input segments; incrementally compressing data in the sliding window until compressed data aligns with a PBA; in response to determining that the compressed data aligns with the PBA: identifying an offset and/or a length of data input corresponding to the compressed data; and writing the compressed data to the PBA in the non-volatile memory; and moving the sliding window consecutively along the plurality of input segments based on the offset and/or the length of the data input.

In some embodiments, the data structure can include: an index field which can include a LBA as an index of the data structure; a PBA field; and a cross-bit field which can indicate whether data in one LBA is written into one PBA or more than one PBA.

In some embodiments, the system can create the set of entries in the data structure by performing the following operations: determining that compressed data associated with a LBA is written into two consecutive PBAs; and setting a flag in a cross-bit field corresponding to the LBA in the data structure.

In some embodiments, the system can receive, from a client, a data read request including the LBA. The system can identify, in the data structure, one or more PBAs corresponding to the LBA. The system can then read compressed data from the one or more PBAs. Next, the system may decompress the compressed data to generate decompressed data. The system can then provide, to the client, requested data based on the decompressed data.

In some embodiments, the system can read the compressed data from the one or more PBAs by: reading the compressed data from one PBA when a flag in a cross-bit field corresponding to the LBA in the data structure is not set; and reading the compressed data from two or more consecutive PBAs when the flag in the cross-bit field corresponding to the LBA in the data structure is set.

In some embodiments, the system can apply erasure coding to the plurality of compressed segments prior to writing the compressed segments to a journal drive.

In some embodiments, the system can apply erasure coding to the plurality of compressed segments prior to writing the compressed segments to different storage nodes in a distributed storage system.

BRIEF DESCRIPTION OF THE FIGURES

FIG. 1 illustrates an exemplary Input/Output (I/O) amplification in a data compression scheme, in accordance with the prior art.

FIG. 2 illustrates an exemplary read amplification in a data compression scheme, in accordance with the prior art.

FIG. 3 illustrates an exemplary data compression scheme with fixed-length output, in accordance with an embodiment of the present disclosure.

FIG. 4 illustrates an exemplary example of a mapping table used in a data compression scheme, in accordance with an embodiment of the present disclosure.

FIG. 5 illustrates an exemplary data compression scheme for reducing read amplification, in accordance with an embodiment of the present disclosure.

FIG. 6A illustrates an exemplary system architecture, in accordance with the prior art.

FIG. 6B illustrates an exemplary modified system architecture, in accordance with an embodiment of the present disclosure.

FIG. 7A presents a flowchart illustrating a method for facilitating a data compression scheme, in accordance with an embodiment of the present disclosure.

FIG. 7B presents a flowchart illustrating a method for facilitating a data compression scheme, in accordance with an embodiment of the present disclosure.

FIG. 7C presents a flowchart illustrating a method for facilitating a data compression scheme to process a read request, in accordance with an embodiment of the present disclosure.

FIG. 8 illustrates an exemplary computer system that facilitates a data compression scheme, in accordance with an embodiment of the present disclosure.

FIG. 9 illustrates an exemplary apparatus that facilitates a data compression scheme, in accordance with an embodiment of the present disclosure.

In the figures, like reference numerals refer to the same figure elements.

DETAILED DESCRIPTION

The following description is presented to enable any person skilled in the art to make and use the embodiments, and is provided in the context of a particular application and its requirements. Various modifications to the disclosed embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be applied to other embodiments and applications without departing from the spirit and scope of the present disclosure. Thus, the embodiments described herein are not limited to the embodiments shown, but are to be accorded the widest scope consistent with the principles and features disclosed herein.

Overview

Data-intensive operations performed by the existing data compression techniques can result in an increase in the utilization of storage resources. Specifically, existing data compression schemes generate irregular-sized compression results. Aligning these irregular-sized compression results with the LBA can involve additional data-intensive operations resulting in increased read/write amplifications and suboptimal usage of the storage resources. Furthermore, due to the inefficient data compression, decompressing the compressed data can also result in increased read amplification. Therefore, the existing data compression schemes can increase the processing burden on the storage system, increase latency, and can result in wearing out of the storage media or can decrease the life span of the storage media. Such a data compression scheme is described below in relation to FIG. 1 and FIG. 2. A system architecture including a data compression scheme is described below in relation to FIG. 6A.

Embodiments described herein address the above-mentioned drawbacks associated with the existing data compression schemes. Specifically, a system can generate a plurality of PBA-aligned fixed-length compressed data segments from LBA-aligned input data segments by applying a sliding window to these input data segments to mitigate read and write amplification of the storage system. Furthermore, the manner in which the data compression scheme is incorporated within a system architecture of a storage cluster can be changed to result in a reduction in the amount of data transferred within the storage cluster, thereby saving the communication bandwidth. Therefore, by applying the data compression scheme described in the present disclosure, the system can reduce latency, save storage system resources, and enhance the efficiency of the distributed storage system. Such a novel data compression scheme is described below in relation to FIG. 3, FIG. 4, FIG. 5, and FIG. 7. A modified system architecture with the data compression scheme, a computer system, and an apparatus facilitating the data compression scheme are described below in relation to FIG. 6B, FIG. 8, and FIG. 9, respectively.

The term “distributed storage system” refers to a set of compute nodes (or client servers) interacting through a set of storage servers (or storage nodes) via a network, such as a data center network.

The term “storage cluster” refers to a group of storage servers.

The term “storage server” refers to a server in a distributed storage system. A storage server can have multiple drives, where data may be written on to a drive for persistent storage. A storage server can also include a journal associated with the journaling file system. A drive can also include a storage, storage medium, or other storage means associated with the drive.

Data Compression Scheme

FIG. 1 illustrates an exemplary Input/Output (I/O) amplification in a data compression scheme, in accordance with the prior art. A system implementing a data compression scheme 100, receives as input a group of files, e.g., File X 102 and File Y 104, to be compressed. Files 102 and 104 are concatenated and divided into fixed-length segments, e.g., 108-114. Specifically, File X 102 is mapped to segments 108 and 110, while File Y 104 is mapped to segments 110-114. Data compression scheme 100 can compress the segments 108-114 individually to generate variable-length compressed segments C1116, C2118, C3120, and C4122. Since compressed segments 116-122 are not aligned with LBAs, they are further subject to additional processing.

For example, to write compressed segment 116 into its corresponding LBA1162, the system implementing data compression scheme 100 can split the compressed segment C1116 into two portions C11124 and C12126. Note that C11124 aligns with LBA1162 and hence can be written into a LBA1162 in a storage drive. However, since length of C12126 does not align with LBA2164, i.e., length of C12126 is less than length of LBA2164, C12126 is padded with a certain bit pattern P 128 so that length of 126 is equal to length of LBA2164. Next, when C2118 is obtained, the system implementing data compression scheme 100 can reload LBA2164 from memory, i.e., {C12130,P 128} is reloaded, and the system can drop P 128 to concatenate with a portion of C2118. Specifically, the system implementing data compression scheme 100 can split C2118 into two portions, i.e., C21132 and C22134, in a manner that a combination of the first portion of C2, i.e., C21132, and a last portion of reloaded C1, i.e., C12130, aligns with LBA2164. Then the system can write {C12130,C21132} into the same LBA2164. Note that during the process of compressing data, the compressed data, i.e., {C12130,P 128}, is first written to LBA2164 only to be read out and re-written with {C12130,C21132}. These read and write operations can increase overhead of read/write amplification during the data-intensive compression.

Next, a second portion of C2, i.e., C22134, is left to be written into subsequent LBA3166. This can be done by padding a certain bit pattern P 136 to C22134 so that a combination of 134 and 136 aligns with LBA 3166. The system can then write C22134 and P 136 into LBA3166. When C3120 is received, the system can reload {C22138, P 136} and combine it with C31140 and can drop P 136 prior to writing the new combination into LBA 3166. The system can continue to perform the reloading of compressed portions from previous LBA and dropping of the padding in this previous LBA before combining the reloaded compressed portion with a consecutive compressed portion. In other words, the system implementing such data-intensive data compression scheme 100 can perform frequent writing, reading, and dropping operations that can result in suboptimal consumption of resources.

Such suboptimal consumption of the resources is also observed when processing a read request for a portion of a file that has been compressed and stored using data compression scheme 100. Specifically, when the system receives a read request to read a part of File Y 104, e.g., portion REQ 106 in File Y 104, the system implementing data compression scheme 100 can first identify where the compressed version of File Y 104 is stored in memory. Then the system may identify that File Y 104 containing the requested portion REQ 106 is included in segment 114 and the compressed version of segment 114 is in segment C4122. Therefore, data compression scheme reads the whole segment C4122 from LBA4-LBA6, e.g., LBA4168, LBA5170, and LBA6172. Note that C32146 in LBA4168 does not correspond to File Y 104 and hence the system can drop C32146 and padding P 154 in LBA6172 at a first stage of processing the read request (operation 156).

The system can then perform decompression of compressed segments: C41148, C42150, and C43152 (operation 158). The system may then select (operation 160) the requested data and send the requested data 106 to the requesting client or source while a remaining portion of segment 114 that does not include the requested data can be dropped at a second stage of processing the read request. Note that when processing a read request, irrelevant data, e.g., C32146, C41148, C42150, and P 154, are read and dropped resulting in read amplification. The system implementing data compression scheme 100 can result in I/O amplification which can increase the processing burden, latency, and can result in a reduced lifespan of the storage media.

FIG. 2 illustrates an exemplary read amplification in a data compression scheme, in accordance with the prior art. A system implementing a data compression scheme 200 can receive as input a group of files 202-212, e.g., File A 202, File B 204, File C 206, File D 208, and File E 208, to be compressed. In practice, in order to improve a compression efficiency, it is desirable that an input data be of a large size. Therefore, to improve the compression efficiency, the system can merge or concatenate the files 202-212 to form a large file or segment 214. The system implementing data compression scheme 200 can then apply data compression to segment 214 with a high compression ratio to generate a compressed segment 216. Compressed segment 216 can then be divided to align with LBA1218, LBA2220, and LBAx 222.

However, the problems associated with such data compression become evident when processing a read request for a specific file. For example, when a request for data (REQ 210) in File D 208 is received, the system implementing data compression scheme 200 may have to read the entire compressed segment 216 to perform decompression 224. During the process of reading the requested data, the system can drop 226 a large amount of decompressed data before providing the requested data to a requesting client. Therefore, in the traditional data compression scheme 200, the entire segment 214 is read and decompressed irrespective of the size of requested data 210, e.g., size of segment 214 could be much larger than the size of requested data 210. Since a large amount of decompressed data is read and then dropped 226, the resources used for the read and decompression operations can be used inefficiently.

Embodiments disclosed herein describe an efficient data compression scheme that is capable of overcoming the disadvantages of existing data compression schemes described in relation to FIG. 1 and FIG. 2. Specifically, the data compression scheme described in the present disclosure is capable of mitigating the read and write amplifications in the storage system. Furthermore, the compression and decompression engines described in the present disclosure can be incorporated into the storage system architecture in a way that can reduce the amount of data transferred and reduce the bandwidth consumption. In the following paragraphs the implementation details of a novel data compression scheme in accordance with the present disclosure are addressed.

FIG. 3 illustrates an exemplary data compression scheme with fixed-length output, in accordance with an embodiment of the present disclosure. During operation, a data compression system implementing data compression scheme 300 can receive data to be compressed. For example, the received data could be represented as equal length data portions, i.e., A 302, B 304, C 306, D 308, E 310, F 312, G 314, H 316, I 318, and J 320. Each data portion in the group of data portions A 302-J 320 can be aligned with an LBA in LBA range 344. The system may continuously and incrementally perform compression on the received data portions until the compressed output of the system is aligned with a PBA in a PBA range 346. Data compression system implementing data compression scheme 300 can ensure that the compressed output is of fixed-length and aligned with PBA.

For example, the data compression system may identify that the data portions A 302, B 304, C 306, and a part of D 308 when compressed may result in output P1338 that aligns with a PBA. The system may use these input data portions as one input data chunk and mark them with an offset and length, e.g., {O1, L1} 326. Note that this input data chunk 326 may not align with LBA. For example, input data chunk 326 may end in the middle of a LBA, e.g., in the middle of D 308. The remaining part of data in the LBA, i.e., remaining part of data in D 308, can be grouped into a later input data chunk 328, i.e., {O2, L2} 328, to generate compressed output P2340 that is aligned with a PBA. Similarly, data portions G 314, H 316, I 318, and a part of J 320, i.e., J1322, can be grouped to form an input data chunk 330 with offset O3 and length L3. The system can then compress input data chunk 330 to generate a PBA-aligned compressed output P3342.

In one embodiment of the present disclosure, the LBAs associated with each of data portions 302-306 and part of 308 can be mapped to one PBA P1338. Similarly, LBAs of data portions including remaining part of D 308, E 310, and F 312 can be mapped to PBA P2340. The LBAs of data portions G 314, H 316, I 318, and part of J 320 (i.e., J1322) can be mapped to PBA P3342. Note that the system does not apply any data padding to the data portions 302-322 during the data compression process, except at the end of the incoming data J2324 where data padding can be applied. Since the system stores the compressed data in memory without performing any additional processing on the input data chunks 326, 328, and 330, e.g., reloading and dropping of data, the system implementing data compression scheme 300 can facilitate the mitigation of read and write amplification. The manner in which a mapping 348 between LBAs in LBA range 344 and PBAs in PBA range 346 is built is described below in relation to FIG. 4.

FIG. 4 illustrates an exemplary example of a mapping table used in a data compression scheme, in accordance with an embodiment of the present disclosure. FIG. 4 shows an implementation 400 of a mapping table which can include mappings between LBAs and PBAs. Mapping table 442 may include the following fields: a LBA 402 field, a PBA 404 field, and a cross-bit 406 field. Mapping table 442 may use LBA as an index of the table to map with a PBA. Cross-bit field 406 may indicate whether compressed data crosses a PBA boundary. In other words, cross-bit field 406 indicates whether only one PBA is to be read or two consecutive PBAs are to be read when processing a read request associated with a LBA.

The data compression system may build mapping table 442 based on the data portions used for performing data compression and the corresponding compressed data portions, e.g., mapping 428. For example, the data compression system may generate compressed data Cx 414 after applying data compression to a data portion LBAx 412. Note that compressed data Cx 414 aligns with PBAi 416 and can be stored in a non-volatile memory. The system may then include an entry 408 in mapping table 442 to denote a mapping between LBAx 412 and PBAi 416. In entry 408 of mapping table 442, cross-bit field 406 is set to “0” to indicate that LBAx 412 after compression generates Cx 414 which aligns with just one PBAi 416. However, when a consecutive data portion aligned with LBAy 418 is compressed, the compressed data portion Cy 422 can be written into both PBAi 424 and PBA i+1 426. The system can include mapping 430 into entry 410 in mapping table 442. Note that since Cy 422 is written into both PBAi 424 and PBAi+1 426, cross-bit field 406 corresponding to entry 410 in mapping table 442 can be set to “1.” With cross-bit field 406 set to “1” the system may have to read both PBAi and PBAi+1 when processing a read request for LBAy.

For example, when the system receives a read request for data associated with LBAy 418, the system may first look-up mapping table 442 to identify entry 410 for LBAy. Note that the LBA itself can also be saved in memory along with its data. Since cross-bit field in entry 410 is set to “1” the system may read out both PBAi and PBA i+1 from memory. The system may then first scan PBAi to find a data block 440 including a header 432 and endian 438 associated with LBAx 434 and Cx 436. Header 432 can mark a start position of a space where information associated with LBAx 434 and Cx 436 are stored, and endian 438 can mark an end position of this space. Next, the system may scan PBAi+1 to find another header and endian associated with LBAy and Cy. The system may identify that PBAi+1 includes the compressed form of the requested data LBAy. The system may then only decompress Cy and send the decompressed data associated with LBAy to a requesting client.

FIG. 5 illustrates an exemplary data compression scheme 500 for reducing read amplification, in accordance with an embodiment of the present disclosure. During operation, the data compression system implementing data compression scheme 500 can receive a plurality of files to be compressed. To generate fixed-length compressed outputs, the system may apply a sliding window to the received files, e.g., File A 502, File B 504, File C 506, File D 508, File E 510, and File F 512. The system can move the sliding window consecutively from left to right along the received files. When compressed data for a specific data portion within the sliding window aligns or reaches the size of a PBA, the system may temporarily halt the movement of the sliding window. The system may then mark the specific data portion within the sliding window with an offset/length and truncate this portion from the received files to represent a data input to a data compression engine. The system may then resume the movement of the sliding window from the end of the specific data portion marked with offset/length and can continue to seek data to be compressed into one or more aligned PBAs. The system can continue to move the sliding window to the end of the received files for generating PBA-aligned compressed data.

For example, sliding window 514 can include entire File A 502 and a portion of File B 504. The data compression system may start compressing in increments data within sliding window 514. In other words, instead of compressing all the data available within sliding window 514, the system may compress incremental amounts of data until a total length of compressed output aligns with a PBA1544. When the system determines that the compressed output length aligns with a PBA, the system may truncate a data chunk 516 from the files included within sliding window 514. The system may mark data chunk 516 with a length and an offset pair {O1, L1}. The system may then use data chunk 516 as an input to a compression engine 534 for generating compressed output that is aligned with PBA1544.

The system can then move sliding window 514 to a position where data chunk 516 ends. This new position of sliding window 514 is indicated by dashed block 518 in FIG. 5. The system may continue to move sliding window 514 along the set of received files 502-512 until the end of the received files is reached. This movement of sliding window 514 is denoted by blocks 518, 522, 526, and 530. As the sliding window moves consecutively along the received files 504-512, the system can provide different data chunks to a compression engine to generate multiple PBA-aligned compressed data.

Specifically, the system may apply sliding window 518 to truncate data chunk {O2,L2} 520 from the received files and apply compression 536 to data chunk 520 to obtain compressed data aligned with PBA2546. Similarly, the system may use the different positions 522, 526, and 530 of sliding window along the received files to identify data chunks {O3,L3} 524, {O4,L4} 528, and {O5,L5} 532, respectively. The system may apply compression 538, 540, and 542 to data chunks 524, 528, and 532, respectively. The output of compression 538, 540, and 542 can include fixed-length PBA-aligned compressed data, i.e., PBA3548, PBA4550, and PBA5540, respectively.

When the system receives a read request, e.g., a request to read a portion of data REQ 554 from File D 508, the system may process the read request for REQ 554 by first locating a corresponding PBA containing compressed data. For example, the system may use LBA associated with the read request to look-up a mapping table (shown in FIG. 4) to identify a PBA corresponding to the LBA in the read request. Specifically, the system may identify PBA4550 to be storing the compressed data associated with the LBA in the read request. The system may apply a decompression engine 554 to decompress compressed data in PBA4550 to obtain decompressed data chunk {O4, L4} 528. Next, the system may read requested data REQ 554 from data chunk 528 and provide data REQ 554 to a client that sent the read request.

Data compression schemes shown in FIG. 1 and FIG. 2, apply data compression to fixed-length input data to generate variable-length compressed outputs resulting in an increased read/write amplification during the data-intensive compression process. Further, when processing a read request, the existing data processing schemes can deliver a suboptimal performance because large compressed data chunks are decompressed and a significant portion of the decompressed data are dropped before actually sending the requested data. Such a suboptimal performance of the existing data compression schemes can result in high read/write amplification, high latency, and inefficient use of communication bandwidth.

In the exemplary embodiments shown in FIG. 5, the system can generate fixed-length PBA-aligned compressed data from varied-length input data chunks. Further, the system can apply a sliding window to the received files to mitigate I/O amplification during the data compression process. Moreover, when processing a read request the data compression system described in the present disclosure is capable of satisfying a small-sized read by reading just one or two PBAs, thereby reducing read amplification and the amount of data dropped when processing the read request. The reduction in read/write amplification can have an improved performance impact on the storage system in terms of latency, the amount of data transferred, and communication bandwidth consumption.

System Architecture

FIG. 6A illustrates an exemplary system architecture 600, in accordance with the prior art. In a distributed storage system, data compression can be applied at multiple places for generating different data formats. For example, system memory 602 can store original data 604 to be compressed, and original data 604 can be divided into data chunks 606-610 prior to applying compression. Specifically, compression 612 is applied to original data 604 to generate compressed original data. The system can then apply erasure coding (EC) 624 to the compressed original data prior to storing them in multiple journal drives 614. The system can also apply compression 616-620 to data chunks 606-610, respectively to generate compressed data chunks. These compressed data chunks are then subject to EC 626-630 prior to being distributed and stored in multiple storage drives 622. Note that in system architecture 600 since compression engines 612, 616-620 are placed after system memory 602, the system may require considerable amount of memory to store uncompressed original data 604. Further, uncompressed original data 604 and data chunks 606-610 are sent towards journal drives 614 and storage drives 622, which may result in a high consumption of network bandwidth.

FIG. 6B illustrates an exemplary modified system architecture 640, in accordance with an embodiment of the present disclosure. System architecture 640 can integrate data compression and decompression engines 646 at the data input in network interface card (NIC) 644. The data received via NIC 644 is compressed at 646 before transferring 648 and storing compressed data 650 in system memory 642. Compressed data 650 can then be treated as original data of the storage cluster. For example, in system memory 642 compressed data 650 can be divided into data chunks 652-656. The system may then apply erasure coding (EC) 658 to compressed data 650 prior to storing them in multiple journal drives 666. The system may also apply EC 660-664 to compressed data chunks 652-656 prior to distributing and storing them in multiple storage drives 668. Note that in system architecture 640, since compression and decompression engines 646 are placed at the entrance of the storage cluster, the amount of data transfer within system architecture 640 can be reduced. Therefore, an overall burden on the data transfer and data processing operations can be mitigated.

Exemplary Method for Facilitating a Data Compression Scheme

FIG. 7A presents a flowchart 700 illustrating a method for facilitating a data compression scheme, in accordance with an embodiment of the present disclosure. During operation, the system can receive data to be written to a non-volatile memory in the storage system (operation 702). The received data can include a plurality of input segments. The system can assign consecutive logical block addresses (LBAs) to the plurality of input segments (operation 704). The system can then compress the plurality of input segments to generate a plurality of fixed-length compressed segments, with each compressed segment aligned with a physical block address (PBA) in a set of PBAs. Specifically, the system can first sequentially read a subset of the plurality of input segments into a sliding window (operation 706). Next, the system can compress in increments data within the sliding window (operation 708). The system may determine whether the compressed data aligns with a PBA (operation 710).

In response to the system determining that the compressed data does not align with the PBA, the system may continue to incrementally compress data in the sliding window (e.g., by applying operations 708 and 710 to incrementally compress data). In response to the system determining that the compressed data aligns with the PBA, i.e., a length of compressed data is equal to a length of the PBA, the system can continue operation at Label A of FIG. 7B.

FIG. 7B presents a flowchart 720 illustrating a method for facilitating a data compression scheme, in accordance with an embodiment of the present disclosure. During operation, the system may write the compressed data to the PBA in the non-volatile memory (operation 722). The system may identify and save an offset and/or a length of data input corresponding to the compressed data (operation 724). The system can then create, in a data structure, an entry that maps the LBAs of the input segments used for generating the compressed data to one or more PBAs (operation 726). The system can then determine whether the compressed output associated with the LBA is written to at least two consecutive PBAs (operation 728). If the condition in operation 728 is true, then the system can set a flag in a cross-bit field in the data structure entry (operation 730). If the condition in operation 728 is false, then the system may continue to operation 732.

The system can proceed to move the sliding window consecutively along the plurality of input segments based on the offset and/or the length of the data input (operation 732). In other words, the system may move the sliding window to start from a position where the previous data input ended. Next, the system can determine whether the sliding window has reached the end of the plurality of input segments in the received data (operation 734). In response to the system determining that the condition in operation 734 is satisfied, the operation returns. In response to the system determining that the condition in operation 734 is not satisfied, the operation continues to operation 706 of FIG. 7A (e.g., by applying operations 706, 708, and 710 to the remaining plurality of input segments that are yet to be compressed).

FIG. 7C presents a flowchart 750 illustrating a method for facilitating a data compression scheme to process a read request, in accordance with an embodiment of the present disclosure. During operation, the system may receive, from a client, a data read request including an LBA (operation 752). The system can then identify in a data structure an entry corresponding to the LBA (operation 754). The system can then determine whether a flag in a cross-bit field in the data structure entry is set (operation 756). When the flag is set, the system may read compressed data from two or more consecutive PBAs (operation 758). When the flag is not set, then the system can read the compressed data from just one PBA (operation 760). Next, the system can decompress the compressed data to generate decompressed data (operation 762). The system can then provide to the requesting client the requested data based on the decompressed data (operation 764) and the operation returns.

Exemplary Computer System and Apparatus

FIG. 8 illustrates an exemplary computer system that facilitates a data compression scheme, in accordance with an embodiment of the present disclosure. Computer system 800 includes a processor 802, a memory 804, and a storage device 806. Computer system 800 can be coupled to a plurality of peripheral input/output devices 832, e.g., a display device 810, a keyboard 812, and a pointing device 814, and can also be coupled via one or more network interfaces to network 808. Storage device 806 can store an operating system 818 and a content processing system 820.

In one embodiment, content processing system 820 can include instructions, which when executed by processor 802 can cause computer system 800 to perform methods and/or processes described in this disclosure. During operation of computer system 800, content processing system 820 can include instructions for receiving a set of files for performing data compression (communication module 822). Content processing system 820 may further include instructions for applying a sliding window to the set of received files (sliding window module 824). Sliding window module 824 can move the sliding window consecutively along the set of received files until the end of the set of received files is reached. At each position of the sliding window, sliding window module 824 may truncate only that portion of data within the sliding window which would produce a fixed-length PBA-aligned compressed data.

Content processing system 820 can include instructions for compressing data within the sliding window to generate PBA-aligned compressed data (data compression module 826). Content processing system 820 can include instructions for storing the PBA-aligned compressed data in a PBA of a non-volatile memory (data storing module 828). Content processing system 820 can further include instructions to build a mapping table that includes a set of mappings between a set of LBAs of plurality of input segments associated with the received files to one or more PBAs of the corresponding compressed data (mapping module 830).

Content processing system 820 can further include instructions for processing a read request from a client (communication module 822). Specifically, the read request can be processed by first looking-up the mapping table to identify a LBA that was included in the read request. After the LBA has been identified in the mapping table, corresponding PBAs can be determined from the mapping table based on a flag setting in a cross-bit field. Decompression of compressed data is performed on the one or more PBAs identified in the mapping table to generate decompressed data (data decompression module 832). Content processing system 820 can include instructions to send the requested data to the client based on the decompressed data (communication module 822).

FIG. 9 illustrates an exemplary apparatus that facilitates a data compression scheme, according to one embodiment of the present disclosure. Apparatus 900 can include a plurality of units or apparatuses that may communicate with one another via a wired, wireless, quantum light, or electrical communication channel. Apparatus 900 may be realized using one or more integrated circuits, and may include fewer or more units or apparatuses than those shown in FIG. 8. Further, apparatus 900 may be integrated in a computer system, or realized as a separate device that is capable of communicating with other computer systems and/or devices. Specifically, apparatus 900 can include units 902-912, which perform functions or operations similar to modules 822-832 of computer system 800 in FIG. 8. Apparatus 900 can include: a communication unit 902, a sliding window unit 904, a data compression unit 906, a data storing unit 908, a mapping unit 910, and a data decompression unit 912.

The methods and processes described in the detailed description section can be embodied as code and/or data, which can be stored in a computer-readable storage medium as described above. When a computer system reads and executes the code and/or data stored on the computer-readable storage medium, the computer system performs the methods and processes embodied as data structures and code and stored within the computer-readable storage medium.

The data structures and code described in this detailed description are typically stored on a computer-readable storage medium, which may be any device or medium that can store code and/or data for use by a computer system. The computer-readable storage medium includes, but is not limited to, volatile memory, non-volatile memory, magnetic and optical storage devices such as disk drives, magnetic tape, CDs (compact discs), DVDs (digital versatile discs or digital video discs), or other media capable of storing computer-readable media now known or later developed.

Furthermore, the methods and processes described above can be included in hardware modules or apparatus. The hardware modules or apparatus can include, but are not limited to, application-specific integrated circuit (ASIC) chips, field-programmable gate arrays (FPGAs), dedicated or shared processors that execute a particular software module or a piece of code at a particular time, and other programmable-logic devices now known or later developed. When the hardware modules or apparatus are activated, they perform the methods and processes included within them.

The foregoing descriptions of embodiments of the present disclosure have been presented for purposes of illustration and description only. They are not intended to be exhaustive or to limit the present disclosure to the forms disclosed. Accordingly, many modifications and variations will be apparent to practitioners skilled in the art. Additionally, the above disclosure is not intended to limit the present disclosure. The scope of the present disclosure is defined by the appended claims.

Claims

1. A computer-implemented method, comprising: receiving data to be written to a non-volatile memory, wherein the data includes a plurality of input segments, which are assigned with consecutive logical block addresses (LBAs);compressing the plurality of input segments to generate a plurality of fixed-length compressed segments, with each fixed-length compressed segment aligned one or more physical block addresses (PBAs) in a set of PBAs, wherein compressing the plurality of input segments comprises: reading sequentially a subset of the plurality of input segments into a sliding window, wherein the subset includes one or more of the input segments;incrementally compressing data in the sliding window until compressed data length satisfies a fixed length associated with a PBA;in response to determining that the compressed data length satisfies the fixed length: identifying an offset and/or a length of data input corresponding to the compressed data; andwriting the compressed data to the PBA in the non-volatile memory; andmoving the sliding window consecutively along the plurality of input segments based on the offset and/or the length of the data input;writing a respective compressed segment into one or more memory blocks addressed with respective PBAs in the non-volatile memory; andcreating, in a data structure, a set of entries which map the LBAs of the input segments to the set of PBAs, wherein creating the set of entries comprises: determining that the compressed segment associated with a LBA and stored in the non-volatile memory crosses a boundary between the two consecutive memory blocks addressed with respective PBAs; andsetting a field corresponding to the LBA in the data structure.

2. The method of claim 1, wherein the data structure includes: an index field which includes a LBA as an index of the data structure;a PBA field; anda cross-bit field which indicates whether data in one LBA is written into one PBA or more than one PBA.

3. The method of claim 1, further comprising: receiving, from a client, a data read request including the LBA;identifying, in the data structure, one or more PBAs corresponding to the LBA;reading compressed data from the one or more PBAs;decompressing the compressed data to generate decompressed data; andproviding, to the client, requested data based on the decompressed data.

4. The method of claim 3, wherein reading the compressed data from the one or more PBAs further comprises: reading the compressed data from one PBA when a flag in a cross-bit field corresponding to the LBA in the data structure is not set; andreading the compressed data from two or more consecutive PBAs when the flag in the cross-bit field corresponding to the LBA in the data structure is set.

5. The method of claim 1, further comprising: applying erasure coding to the plurality of compressed segments prior to writing the compressed segments to a journal drive.

6. The method of claim 1, further comprising: applying erasure coding to the plurality of compressed segments prior to writing the compressed segments to different storage nodes in a distributed storage system.

7. The method of claim 1, wherein compressing the plurality of input segments further comprises: continuously performing the incrementally compressing the data in the sliding window until the compressed data length satisfied the fixed length associated with the PBA.

8. The method of claim 1, wherein compressing the plurality of input segments is performed by a compression/decompression engine associated with a network interface card while receiving the data to be written to the non-volatile memory, wherein the compression/decompression engine resides at an entrance to different storage nodes in a distributed storage system.

9. A computer system, comprising: a co-processor; anda storage device coupled to the processor and storing instructions, which when executed by the co-processor cause the co-processor to perform a method, the method comprising: receiving data to be written to a non-volatile memory, wherein the data includes a plurality of input segments, which are assigned with consecutive logical block addresses (LBAs);compressing the plurality of input segments to generate a plurality of fixed-length compressed segments, with each fixed-length compressed segment aligned with a one or more physical block addresses (PBAs) in a set of PBAs, wherein compressing the plurality of input segments comprises: reading sequentially a subset of the plurality of input segments into a sliding window, wherein the subset includes one or more of the input segments;incrementally compressing data in the sliding window until compressed data length satisfies a fixed length associated with a PBA;in response to determining that the compressed data length satisfies the fixed length: identifying an offset and/or a length of data input corresponding to the compressed data; andwriting the compressed data to the PBA in the non-volatile memory; andmoving the sliding window consecutively along the plurality of input segments based on the offset and/or the length of the data input;writing a respective compressed segment into one or more memory blocks addressed with respective PBAs in the non-volatile memory; andcreating, in a data structure, a set of entries which map the LBAs of the input segments to the set of PBAs, wherein creating the set of entries comprises: determining that the compressed segment associated with a LBA and stored in the non-volatile memory crosses a boundary between the two consecutive memory blocks addressed with respective PBAs; andsetting a field corresponding to the LBA in the data structure.

10. The computer system of claim 9, wherein the data structure includes: an index field which includes a LBA as an index of the data structure;a PBA field; anda cross-bit field which indicates whether data in one LBA is written into one PBA or more than one PBA.

11. The computer system of claim 9, wherein the method further comprises: receiving, from a client, a data read request including the LBA;identifying, in the data structure, one or more PBAs corresponding to the LBA;reading compressed data from the one or more PBAs;decompressing the compressed data to generate decompressed data; andproviding, to the client, requested data based on the decompressed data.

12. The computer system of claim 11, wherein reading the compressed data from the one or more PBAs further comprises: reading the compressed data from one PBA when a flag in a cross-bit field corresponding to the LBA in the data structure is not set; andreading the compressed data from two or more consecutive PBAs when the flag in the cross-bit field corresponding to the LBA in the data structure is set.

13. The computer system of claim 9, wherein the method further comprises: applying erasure coding to the plurality of compressed segments prior to writing the compressed segments to a journal drive.

14. The computer system of claim 9, wherein the method further comprises: applying erasure coding to the plurality of compressed segments prior to writing the compressed segments to different storage nodes in a distributed storage system.

15. The computer system of claim 9, wherein compressing the plurality of input segments is performed by a compression/decompression engine associated with a network interface card while receiving the data to be written to the non-volatile memory, wherein the compression/decompression engine resides at an entrance to different storage nodes in a distributed storage system.

16. An apparatus, comprising: a co-processor; anda storage medium storing instructions, which when executed by the co-processor cause the co-processor to perform a method, the method comprising: receiving data to be written to a non-volatile memory, wherein the data includes a plurality of input segments, which are assigned with consecutive logical block addresses (LBAs);compressing the plurality of input segments to generate a plurality of fixed-length compressed segments, with each fixed-length compressed segment aligned with one or more physical block address (PBA) in a set of PBAs, wherein compressing the plurality of input segments comprises: reading sequentially a subset of the plurality of input segments into a sliding window, wherein the subset includes one or more of the input segments;incrementally compressing data in the sliding window until compressed data length satisfies a fixed length associated with a PBA;in response to determining that the compressed data length satisfies the fixed length: identifying an offset and/or a length of data input corresponding to the compressed data; andwriting the compressed data to the PBA in the non-volatile memory; andmoving the sliding window consecutively along the plurality of input segments based on the offset and/or the length of the data input;writing a respective compressed segment to one or more memory blocks addressed with respective PBAs in the non-volatile memory; andcreating, in a data structure, a set of entries which map the LBAs of the input segments to the set of PBAs, wherein creating the set of entries comprises: determining that the compressed segment associated with a LBA and stored in the non-volatile memory crosses a boundary between the two consecutive memory blocks addressed with respective PBAs; andsetting a field corresponding to the LBA in the data structure.

17. The apparatus of claim 16, wherein the data structure includes: an index field which includes a LBA as an index of the data structure;a PBA field; anda cross-bit field which indicates whether data in one LBA is written into one PBA or more than one PBA.

18. The apparatus of claim 16, wherein the method further comprises: receiving, from a client, a data read request including the LBA;identifying, in the data structure, one or more PBAs corresponding to the LBA;reading compressed data from the one or more PBAs;decompressing the compressed data to generate decompressed data; andproviding, to the client, requested data based on the decompressed data.

19. The apparatus of claim 18, wherein reading the compressed data from the one or more PBAs further comprises: reading the compressed data from one PBA when a flag in a cross-bit field corresponding to the LBA in the data structure is not set; andreading the compressed data from two or more consecutive PBAs when the flag in the cross-bit field corresponding to the LBA in the data structure is set.

20. The apparatus of claim 16, wherein the method further comprises: applying erasure coding to the plurality of compressed segments prior to writing the compressed segments at least one of: a journal drive; anddifferent storage nodes in a distributed storage system.