`
`UNITED STATES PATENT AND TRADEMARK OFFICE
`
`_____________
`
`BEFORE THE PATENT TRIAL AND APPEAL BOARD
`
`_____________
`
`MICRON TECHNOLOGY, INC.,
`Petitioner
`
`v.
`
`VERVAIN, LLC,
`Patent Owner
`_____________
`
`Case: IPR2021-01547
`U.S. Patent No. 8,891,298
`_____________
`
`
`
`PATENT OWNER’S RESPONSE
`
`
`
`
`
`
`
`
`
`
`
`

`

`IPR2021-01547
`U.S. Patent No. 8,891,298
`
`
`
`
`
`I.
`
`II.
`
`TABLE OF CONTENTS
`
`INTRODUCTION ........................................................................................... 1
`
`OVERVIEW OF THE ’298 PATENT AND THE CHALLENGED
`CLAIMS .......................................................................................................... 2
`
`A.
`
`B.
`
`C.
`
`SLC AND MLC FLASH ........................................................................... 3
`
`ADDRESS TABLE ..................................................................................... 5
`
`DATA INTEGRITY TESTS ......................................................................... 6
`
`D. HOT AND COLD DATA ............................................................................ 6
`
`E.
`
`CLAIM 1 .................................................................................................. 7
`
`III.
`
`PERSON OF ORDINARY SKILL IN THE ART .......................................... 9
`
`IV. OVERVIEW OF THE ALLEGED PRIOR ART ............................................ 9
`
`A. DUSIJA (EX. 1010) ................................................................................10
`
`B.
`
`SUTARDJA (EX. 1011) ..........................................................................14
`
`C. MOSHAYEDI (EX. 1012) .......................................................................19
`
`V.
`
`CLAIM CONSTRUCTION ..........................................................................22
`
`A.
`
`B.
`
`C.
`
`“BLOCKS” (CLAIMS 1, 2) .......................................................................23
`
`“DATA INTEGRITY TEST” (CLAIM 1) ......................................................27
`
`“ON A PERIODIC BASIS” (CLAIM 11) ......................................................30
`
`VI. THE CITED REFERENCES DO NOT RENDER CLAIMS 1-5, 8-9, AND
`11 UNPATENTABLE ...................................................................................30
`
`A. DUSIJA IN VIEW OF SUTARDJA DOES NOT RENDER OBVIOUS CLAIMS 1-
`5 AND 11 (GROUND 1) ..........................................................................31
`
`
`
`DUSIJA IN VIEW OF SUTARDJA DOES NOT DISCLOSE OR SUGGEST
`LIMITATION [1.F] (GROUND 1)...................................................32
`
`A)
`
`THE PETITION RELIES ON SUTARDJA’S DISCLOSURES
`REGARDING LOGICAL ADDRESSES, NOT “BLOCKS” AS
`PROPERLY CONSTRUED ...................................................32
`
`i
`
`

`

`IPR2021-01547
`U.S. Patent No. 8,891,298
`
`
`B)
`
`THE PETITION HAS NOT ESTABLISHED THAT SUTARDJA
`DISCLOSES DETERMINING WHICH OF THE BLOCKS ARE
`ACCESSED MOST FREQUENTLY .......................................39
`
`
`
`DUSIJA IN VIEW OF SUTARDJA DOES NOT DISCLOSE OR SUGGEST
`LIMITATION [1.G] (GROUND 1) ..................................................44
`
`A)
`
`B)
`
`C)
`
`D)
`
`THE PETITION DOES NOT ANALYZE “BLOCKS” FOR
`LIMITATION [1.G] UNDER THE PROPER CONSTRUCTION ..44
`
`THE PETITION’S RELIANCE ON SUTARDJA’S DISCLOSURE
`OF SWAPPING DATA FOR LIMITATION [1.G] IS FLAWED ..45
`
`THE “SECOND WAY” DISCUSSED IN THE PETITION FOR
`LIMITATION [1.G] DOES NOT DISCLOSE “TRANSFERRING
`THE RESPECTIVE CONTENTS OF THOSE BLOCKS TO THE AT
`LEAST ONE SLC NON-VOLATILE MEMORY MODULE.”...48
`
`SUTARDJA’S DISCLOSURE OF MAPPING ADDRESSES TO THE
`SECOND NVS MEMORY DOES NOT DISCLOSE OR SUGGEST
`TRANSFERRING CONTENTS OF BLOCKS TO SLC AS IN
`LIMITATION [1.G] ............................................................50
`
`
`
`CLAIMS 2-5 AND 11 (GROUND 1) ...............................................56
`
`B.
`
`DUSIJA IN VIEW OF SUTARDJA AND LI DOES NOT RENDER OBVIOUS
`CLAIMS 8 AND 9 (GROUND 2) ...............................................................56
`
`C. MOSHAYEDI IN VIEW OF DUSIJA DOES NOT RENDER OBVIOUS CLAIMS
`1-5 AND 11 (GROUND 3) .......................................................................56
`
` MOSHAYEDI IN VIEW OF DUSIJA DOES NOT DISCLOSE OR
`SUGGEST TRANSFERRING CONTENTS OF BLOCKS TO SLC AS IN
`LIMITATION [1.G] ......................................................................57
`
`
`
`
`
`PETITIONER’S RELIANCE ON MOSHAYEDI’S DISCLOSURES
`REGARDING LOGICAL BLOCK ADDRESSES IS ERRONEOUS FOR
`THE “BLOCKS” OF LIMITATIONS [1.F] AND [1.G] .......................63
`
`CLAIMS 2-5 AND 11 (GROUND 3) ...............................................67
`
`D. MOSHAYEDI IN VIEW OF DUSIJA AND SUTARDJA DOES NOT RENDER
`OBVIOUS CLAIM 11 (GROUND 4) ..........................................................67
`
`E. MOSHAYEDI IN VIEW OF DUSIJA AND LI DOES NOT RENDER OBVIOUS
`CLAIMS 8 AND 9 (GROUND 5) ...............................................................67
`
`VII. CONCLUSION ..............................................................................................68
`
`ii
`
`

`

`IPR2021-01547
`U.S. Patent No. 8,891,298
`
`
`TABLE OF AUTHORITIES
`
` Page(s)
`
`Cases
`
`Amazon Web Services, Inc. v. Saint Regis Mohawk Tribe,
`IPR2019-00103, Paper No. 22 (PTAB May 10, 2019) .................... 37, 38, 42, 43
`
`Belden Inc. v. Berk-Tek,
`805 F.3d 1064 (Fed. Cir. 2015) .......................................................................... 53
`
`Bicon, Inc. v. Straumann Co.,
`441 F.3d 945 (Fed. Cir. 2006) ............................................................................ 28
`
`Corning Incorp. v. Danjou’s DSM IP Assets B.V.,
`Case No. IPR2013-00043, Paper No. 95 (PTAB May 1, 2014) ................... 38, 43
`
`DePuy Spine, Inc. v. Medtronic Sofamor Danek, Inc.,
`469 F.3d 1005 (Fed. Cir. 2006) .......................................................................... 23
`
`Hill-Rom Servs., Inc. v. Stryker Corp.,
`755 F.3d 1367 (Fed. Cir. 2014) .......................................................................... 28
`
`Innova/Pure Water, Inc. v. Safari Water Filtration Sys., Inc.,
`381 F.3d 1111 (Fed. Cir. 2004) .......................................................................... 28
`
`Merck & Co. v. Teva Pharm. USA, Inc.,
`395 F.3d 1364 (Fed. Cir. 2005) .......................................................................... 28
`
`Phillips v. AWH Corp.,
`415 F.3d 1303 (Fed. Circ. 2005) (en banc) ........................................................ 22
`
`Samsung SDI Co., Ltd. v. Ube Industries, Inc.,
`IPR2017-02116, Paper No. 8 (March 12, 2018) ........................................... 38, 48
`
`Toyota Motor Corp. v. Cellport Systems, Inc.,
`IPR2015-00633, Paper No. 11 (Aug. 14, 2015) ................................................. 23
`
`Vervain, LLC v. Micron Tech., Inc.,
`No. W-21-CV-00487-ADA, Dkt. 42 (W.D. Tex. Jan. 24, 2022) ................. 28, 30
`
`iii
`
`

`

`IPR2021-01547
`IPR2021-01547
`U.S. Patent No. 8,891,298
`U.S. Patent No. 8,891,298
`
`
`Statutes
`Statutes
`
`35 U.S.C. § 316(€)
`ccccssesccsssesssssscccssssessssssccessssecsssusesssssesessssessssnsesessueesssssessssneessssseessse 1
`35 U.S.C. § 316(e) ..................................................................................................... 1
`
`
`
`
`
`
`1V
`iv
`
`

`

`IPR2021-01547
`U.S. Patent No. 8,891,298
`
`
`Exhibit
`
`EXHIBIT LIST
`
`Description
`
`Ex. 2001 Declaration of Dr. Sunil Khatri
`
`Previously
`Submitted
`X
`
`Ex. 2002 Chen et al., Ultra MLC Technology Introduction, Advantech
`Technical White Paper (Oct. 5, 2012) (“Chen”)
`
`Ex. 2003 Excerpts from Micheloni et al., Inside NAND Flash Memories
`(1st ed. 2010) (“Micheloni”)
`
`Ex. 2004 U.S. Patent No. 10,950,300 to G.R. Mohan Rao (“’300
`Patent”)
`
`Ex. 2005 Microsoft Computer Dictionary definition for “data integrity”
`
`Ex. 2006 Hargrave’s Communications Dictionary definition for “data
`integrity”
`
`Ex. 2007 https://www.law360.com/articles/1381597/albright-says-he-ll-
`very-rarely-put-cases-on-hold-for-ptab
`
`Ex. 2008 Docket Sheet for Case. No. 6:21-cv-487-ADA; Vervain v.
`Micron Technology et al.; U.S. District Court, Western District
`of Texas.
`
`Ex. 2009 Exhibit A-3, Invalidity Claim Chart for the ’298 Patent based
`on U.S. Patent Application Pub. No. 2011/0099460 (“Dusija”)
`
`Ex. 2010 Exhibit A-18, Invalidity Claim Chart for the ’298 Patent based
`on U.S. Patent Application Pub. No. US 2008/0140918
`(“Sutardja”)
`
`Ex. 2011 Exhibit A-20. Invalidity Claim Chart for the ’298 Patent based
`on U.S. Patent Application Pub. No. 2009/0327591
`(“Moshayedi”)
`
`Ex. 2012 Claim Construction Order in Vervain v. Micron Tech., Inc.,
`No. 6:21-cv-487-ADA (W.D. Tex.) and Vervain v. Western
`Digital Corp., No. 6:21-cv-488-ADA (W.D. Tex.) (Jan. 24,
`
`X
`
`X
`
`X
`
`X
`
`X
`
`X
`
`X
`
`X
`
`X
`
`X
`
`
`
`v
`
`

`

`IPR2021-01547
`U.S. Patent No. 8,891,298
`
`
`2022)
`
`Ex. 2013 Micron’s Preliminary Invalidity Contentions for U.S. Patent
`Nos. 8,891,298; 9,196,385; 9,997,240; and 10,950,300; Case.
`No. 6:21-cv-487-ADA; Vervain v. Micron Technology et al.;
`U.S. District Court, Western District of Texas.
`
`Ex. 2014 Declaration of Dr. Sunil Khatri in Support of Patent Owner’s
`Response
`
`Ex. 2015 Transcript of June 10, 2022 Deposition of Dr. David Liu
`
`Ex. 2016 U.S. Patent No. 8,285,940
`
`X
`
`
`
`
`
`
`
`vi
`
`

`

`IPR2021-01547
`U.S. Patent No. 8,891,298
`
`
`I.
`
`INTRODUCTION
`
`Vervain, LLC (“Patent Owner”) respectfully submits this Response to the
`
`Board’s decision to institute inter partes review (Paper No. 13, the “Decision”) and
`
`to the petition for inter partes review (Paper No. 1, the “Petition”) filed by Micron
`
`Technology, Inc. (“Petitioner”). The Board instituted review of U.S Patent No.
`
`8,891,298 (Ex. 1001, “the ’298 patent” or “the challenged patent”) on five grounds
`
`that challenge claims 1-5, 8, 9, and 11 (“the challenged claims”) of the ’298 patent.
`
`Decision, 7-8, 40. Petitioner has not, however, carried its burden of proving
`
`unpatentability by a preponderance of the evidence (35 U.S.C. § 316(e)).
`
`As explained below and in the accompanying declaration of Patent Owner’s
`
`expert, Dr. Khatri, Petitioner has not established that the cited prior art discloses or
`
`suggests all of the limitations of the challenged claims.1 For example, the Petition
`
`does not analyze the claims and prior art under the proper construction of “blocks,”
`
`which the Petitioner improperly maps to teachings in the prior art regarding logical
`
`(not physical) blocks/addresses. Additionally, the Petition makes several errors in
`
`its analysis of limitations [1.F] and [1.G] of claim 1 for both the Dusija-Sutardja
`
`ground (Ground 1) and the Moshayedi-Sutardja ground (Ground 3). Moreover, a
`
`
`1 Patent Owner submits the declaration of Dr. Khatri (Ex. 2014), an expert in the field
`
`of the ’298 patent. (Ex. 2014, ¶¶1-19.)
`
`1
`
`

`

`IPR2021-01547
`U.S. Patent No. 8,891,298
`
`person of ordinary skill in the art (POSA) would not have configured the system of
`
`the Dusija reference (Ex. 1010) in view of Sutardja (Ex. 1011), and would not have
`
`configured Sutardja’s system in view of Moshayedi (Ex. 1012), in the manners
`
`proposed by Petitioner.
`
`Petitioner has not met its burden in this proceeding and has not established
`
`that the challenged claims are unpatentable. Accordingly, the patentability of the
`
`challenged claims should be confirmed.
`
`II. OVERVIEW OF THE ’298 PATENT AND THE CHALLENGED
`CLAIMS
`
`The ’298 Patent, entitled “Lifetime Mixed Level Non-Volatile Memory
`
`System” was filed on April 25, 2012 and has an effective filing date of July 19, 2011.
`
`Ex. 1001. Dr. Mohan Rao is the sole named inventor of the ’298 Patent.
`
`At a high level, the ’298 Patent describes, among other things, a reliable flash
`
`memory storage system combining both single-level cell (SLC) and multi-level cell
`
`(MLC) non-volatile memories.2 Ex. 1001, Abstract; Ex. 2014, ¶¶24-39. Prior to the
`
`’298 Patent, Dr. Rao recognized that “MLC NAND flash SSDs are slowly replacing
`
`
`2 Non-volatile memories can store information even after the system is powered off.
`
`Flash memory is a specific type of non-volatile memory, where data is stored in
`
`“blocks” of “pages.” Ex. 1001, 2:31-48.
`
`2
`
`

`

`IPR2021-01547
`U.S. Patent No. 8,891,298
`
`and/or coexisting with SLC NAND flash in newer SSD systems” because “MLC
`
`flash memory is less expensive than SLC flash memory[] on a cost per bit basis.”
`
`Ex. 1001, 3:14-15, 5:24-26. However, while “MLC NAND flash enjoys greater
`
`density than SLC NAND flash” it comes “at the cost of a decrease in access speed
`
`and lifetime (endurance).” Id., 3:19-21. As a result, various hybrid systems
`
`combining SLC and MLC (among others) have been developed to combine the
`
`benefits of both types of non-volatile flash storage at a low cost. Id., 3:43-45.
`
`The ’298 Patent addresses improvements and solutions for managing the
`
`writing of data optimally for improved reliability and lifetime (endurance) of such
`
`hybrid memory systems. Id., 3:38-45. Specifically, the Challenged Claims are
`
`directed to specific techniques for efficiently using SLC and MLC flash to improve
`
`the overall performance of the memory. Id., claim 1. For example, if certain data is
`
`used more frequently, then it is transferred to higher-performance SLC. Id. By
`
`doing so, the number of errors is reduced, and overall endurance of the memory is
`
`increased. Id., 3:43-45.
`
`A.
`
`SLC and MLC Flash
`
`SLC memory stores 1 bit per cell, and MLC memory stores more than 1 bit
`
`per cell. Id., 1:64-67; Ex. 2014, ¶¶30-32. As noted above, there are pros and cons
`
`to SLC and MLC flash. In general, SLC is faster and less prone to errors, but
`
`requires more space and power to store a given amount of data. Ex. 1001, 1:38-43.
`
`3
`
`

`

`IPR2021-01547
`U.S. Patent No. 8,891,298
`
`The opposite is true of MLC. MLC flash is slower and more prone to errors, but
`
`stores data more densely with less power consumption. Id., 3:19-21.
`
`SLC and MLC flash memories both use the same type of transistor called a
`
`floating gate transistor. Id., 3:29. They both store a charge in the floating gate of
`
`each transistor (cell), which changes the threshold voltage of the transistor. The
`
`memory uses the threshold voltage to determine what bit, or bits, were stored in the
`
`transistor. The MLC cell in the figure below illustrates threshold voltages for a 2-
`
`bit MLC cell.
`
`Ex. 2002, 5.
`
`
`
`The primary difference between SLC and MLC is what data each threshold
`
`voltage represents. With SLC flash, the transistor stores only a 1 or 0, so a wide
`
`range of threshold voltages can be allotted to a single bit. This allows for faster and
`
`more reliable memory access. On the other hand, MLC flash must be slowly and
`
`carefully programmed to a narrower, more precise range of threshold voltages, with
`
`each threshold voltage range representing a specific pair of bits (see figure above,
`
`4
`
`

`

`IPR2021-01547
`U.S. Patent No. 8,891,298
`
`which shows four pairs of bits—11, 10, 01, and 00—corresponding to smaller ranges
`
`of threshold voltages compared to the SLC). Ex. 1001, 3:15-21.
`
`B. Address Table
`
`To provide wear leveling, garbage collection, and bad block management, a
`
`translation layer is used to map logical addresses to actual physical memory
`
`locations. Ex. 2003, 9-11; Ex. 1001, 2:49-3:13; Ex. 2014, ¶33. As part of this
`
`translation layer, “tables are widely used in order to map sectors and pages from
`
`logical to physical.” Ex. 2003, 9; Ex. 1001, 2:64-3:1. These tables map logical
`
`blocks to physical blocks. Ex. 2003, 9-11; Ex. 1001, 2:64-3:1. Using a “block” or
`
`similar granularity is important, since flash memory is arranged so that when erasing
`
`and rewriting data, a whole block is “erased together.” Ex. 2003, 6; Ex. 1001, 2:38-
`
`48. Dr. Rao explained that “[t]he address ranges within the translation table will
`
`assume some minimum quantum, such as, for example, one block…” Ex. 1001,
`
`5:27-31. Dr. Rao further explained that memory is written and mapped on the
`
`granularity of a “quantum,” such as a block or page. Id., 5:27-31; FIGS. 3A-3B.
`
`During operation of the flash memory, logical addresses are frequently
`
`remapped to new physical locations. Id., 2:65-3:31, 3:67-4:10, 5:20-40. Over time,
`
`a particular logical address may be mapped or associated with many different
`
`physical locations (blocks). Ex. 2014, ¶¶33, 73-74. And multiple logical addresses
`
`5
`
`

`

`IPR2021-01547
`U.S. Patent No. 8,891,298
`
`may point to the same block over time, so there is not a one-to-one correspondence
`
`between the logical addresses and the blocks over time. Id.
`
`C. Data Integrity Tests
`
`As mentioned above, when data is stored in MLC memory, it is more prone
`
`to errors than when data is stored in SLC memory. One reason for this is that the
`
`threshold voltage intervals for MLC memory are smaller than the intervals for SLC
`
`memory, and thus, errors can occur when writing or reading the data. Ex. 2014, ¶34.
`
`Errors can also be caused by the data stored in neighboring cells. Id. A data integrity
`
`test is a test that checks the integrity of the data (i.e., whether errors have occurred).
`
`This test can be run immediately after data is written, or at a later time. If the test
`
`reveals a problem such as corrupt data, the data can be remapped to SLC (which is
`
`less error-prone), and the address table modified accordingly. Ex. 1001, 4:4-10.
`
`Alternately, MLC data can be remapped to other MLC blocks, and the address table
`
`then modified accordingly. Id., 2:59-3:13.
`
`D. Hot and Cold Data
`
`One can distinguish between “hot” blocks (which receive more frequent
`
`writes), and “cold” blocks (which receive less frequent writes). Id., 6:24-29.
`
`Because SLC has greater endurance, “hot” blocks can be allocated to SLC to
`
`increase the lifetime of the system. Id. “Cold” blocks, on the other hand, can be
`
`allocated to MLC to take advantage of its higher density storage. Ex. 2014, ¶35.
`
`6
`
`

`

`IPR2021-01547
`U.S. Patent No. 8,891,298
`
`As the ’298 Patent explains, the contents of the “hot” blocks can be transferred
`
`to SLC “on a periodic basis, such as, for example every 1000 writes or every 10,000
`
`writes.” Ex. 1001, 6:30-35. Transferring groups of “hot” blocks on a periodic basis
`
`allows the controller to transfer the data from MLC blocks to SLC as a background
`
`process in-between write commands. Ex. 2014, ¶35.
`
`E. Claim 1
`
`In claim 1, the MLC and SLC comprise “erasable blocks” (highlighted red).
`
`These are the physical locations that must be erased before data can be written to
`
`them. See [1.A] and [1.B] below. Meanwhile, an address map comprises a list of
`
`“logical address ranges” (highlighted purple); these logical address ranges are
`
`mapped to the physical address ranges for the blocks. [1.D].
`
`[1.PRE]
`
`A system for storing data comprising:
`
`Claim 1
`
`[1.A]
`
`at least one MLC…module comprising a plurality of individually
`
`erasable blocks;
`
`[1.B]
`
`at least one SLC…module comprising a plurality of individually
`
`erasable blocks; and
`
`[1.C]
`
`a controller coupled to the at least one MLC…module and the at
`
`least one SLC…module wherein the controller is adapted to:
`
`[1.D]
`
`a) maintain an address map of at least one of the MLC and
`
`SLC…modules, the address map comprising a list of logical
`
`7
`
`

`

`IPR2021-01547
`U.S. Patent No. 8,891,298
`
`
`address ranges accessible by a computer system, the list of
`
`logical address ranges having a minimum quanta of addresses,
`
`wherein each entry in the list of logical address ranges maps to
`
`a similar range of physical addresses within either the at least
`
`one SLC…module or within the at least one MLC…module;
`
`[1.E]
`
`b) determine if a range of addresses listed by an entry and
`
`mapped to a similar range of physical addresses within the at
`
`least one MLC…module, fails a data integrity test, and, in the
`
`event of such a failure, the controller remaps the entry to the next
`
`available equivalent range of physical addresses within the at
`
`least one SLC…module;
`
`[1.F]
`
`c) determine which of the blocks of the plurality of the blocks
`
`in the MLC and SLC…modules are accessed most frequently by
`
`maintaining a count of the number of times each one of the
`
`blocks is accessed; and
`
`[1.G]
`
`d) allocate those blocks that receive the most frequent writes by
`
`transferring the respective contents of those blocks to the at least
`
`one SLC…module.
`
`
`
`As can be seen above, claim 1 uses the claim terms “blocks” and “logical
`
`address ranges” to refer to two different things. The blocks are the physical locations
`
`in the MLC and SLC where the data is stored. [1.A-B]. Each block has a fixed
`
`“range of physical addresses.” [1.D]. Meanwhile, the address map contains a list of
`
`logical address ranges that are mapped to the physical address ranges. As the claim
`
`8
`
`

`

`IPR2021-01547
`U.S. Patent No. 8,891,298
`
`indicates, the logical address ranges are remapped to new physical address ranges.
`
`[1.E]. Thus, a logical address range does not permanently point to a specific physical
`
`address range. Rather the corresponding physical address range may change over
`
`time.
`
`Turning to [1.F], the claim refers to “the blocks,” where the antecedent basis
`
`is “erasable blocks” in [1.A-B]. Thus, the controller is adapted to “determine which
`
`of the [erasable blocks]…are accessed most frequently by maintaining a count of the
`
`number of times each one of the blocks is accessed.”
`
`Finally, in [1.G], the controller is adapted to transfer the contents of those
`
`blocks that receive the most frequent writes to SLC memory.
`
`III. PERSON OF ORDINARY SKILL IN THE ART
`
`For purposes of this proceeding only, Patent Owner adopts Petitioner’s
`
`definition of a person of ordinary skill in the art (POSA). Petition, 27; Ex. 2014,
`
`¶20-23.
`
`IV. OVERVIEW OF THE ALLEGED PRIOR ART
`
`For the purposes of this Response, only Dusija, Sutardja, and Moshayedi are
`
`relevant because they are the only references asserted against the sole independent
`
`claim—claim 1.
`
`9
`
`

`

`IPR2021-01547
`U.S. Patent No. 8,891,298
`
`
`A. Dusija (Ex. 1010)
`
`Dusija addresses a problem that occurs with flash memory—as it ages, its
`
`error rate increases, which requires a resource intensive ECC to correct errors. Ex.
`
`1010, [0012-0017]; Ex. 2014, ¶¶53-59. As Dusija explains, in order to ensure data
`
`integrity in such situations, a complex, computationally intensive ECC is utilized
`
`which results in memory performance degradation. Ex. 1010, [0014]. To address
`
`this problem, Dusija “provid[es] a mechanism to control and limit the errors arising
`
`after writing to high density memory [i.e., MLC] … and a second chance to rewrite
`
`data with less error if the copy in the high density memory has excessive errors.”
`
`Id., [0024]. By using the disclosed mechanism the ECC complexity is reduced and
`
`an “advantage is gained at the slight expense of an additional post-write read and
`
`infrequent additional rewrites to a less [sic, lower] density memory portion [i.e.,
`
`SLC].” Id.
`
`In contrast to the ’298 Patent, the SLC in Dusija is primarily used for
`
`enhancing memory life by starting an error management process “after the memory
`
`has aged to a predetermined amount.” Id., Abstract. Hence, the primary teaching of
`
`Dusija is to extend memory life when the memory is aging, and not slow it down by
`
`doing the “error management … when a memory is new with little or no errors.” Id.
`
`Thus, Dusija teaches a MLC memory chip where a nominal amount of SLC is
`
`10
`
`

`

`IPR2021-01547
`U.S. Patent No. 8,891,298
`
`deployed only late in the lifetime of the MLC memory in order to avoid ECC
`
`processing at the controller ASIC. Id., [0016], [0024], [0155].
`
`To reduce ECC complexity and increase memory performance, Dusija
`
`describes a flash memory device 90 including a controller ASIC chip 102
`
`(highlighted purple) and a memory chip 100 (highlighted blue).
`
`
`
`11
`
`

`

`IPR2021-01547
`U.S. Patent No. 8,891,298
`
`Id., FIG. 1, [0016], [0059], [0106].3 The memory array 200 is shown in more detail
`
`in Figures 14A-14B, 16A-16C, and 20A-20C. Exemplary Figure 20A is reproduced
`
`below.
`
`Id., FIG. 20A. As seen above, the memory array 200 comprises “a first portion
`
`operating with less error but of lower density storage” and a “second portion
`
`
`
`
`3 Unless otherwise noted, Patent Owner added coloring to Figures.
`
`12
`
`

`

`IPR2021-01547
`U.S. Patent No. 8,891,298
`
`operating with a higher density but less robust storage” (highlighted blue). Id.,
`
`Abstract.
`
`For the low density storage portion of the memory, Dusija discloses memory
`
`referred to as “D1” which it describes as memory cells storing 1 bit of data—i.e.,
`
`SLC. For the high density storage portion of the memory, Dusija discloses memory
`
`referred to as “D3” which it describes as memory cells storing 3 bits of data—i.e.,
`
`MLC. Id., [0025], [0028].
`
`In the embodiment shown in Figure 20A, Dusija describes that input data from
`
`the host is first cached in a first section 411 (highlighted yellow) of the D1 memory
`
`and subsequently folded into D3 memory. Id., [0162-0163]. The D1 (SLC) memory
`
`is primarily used for staging and caching incoming data from the host.
`
`Further, Dusija describes “post-write-read” error management processes
`
`where, when enabled, a filled D3 block is read back and compared to the data in the
`
`D1 block to determine whether the error rate exceeds a predetermined threshold. Id.,
`
`[0028], [0162-0163]. If so, the currently written D3 block is rejected and a retry
`
`takes place with data being refolded into a new D3 block. Id. The new D3 block is
`
`again read back and checked for excessive errors. Id. If the new D3 block passes,
`
`then it is determined to have good data and the original data in the D1 block is made
`
`obsolete. Id. If the new D3 block still shows excessive error, the process is repeated
`
`13
`
`

`

`IPR2021-01547
`U.S. Patent No. 8,891,298
`
`for a predetermined number of retries after which time the D1 to D3 folding
`
`operation is abandoned with the original data kept at D1. Id.
`
`In further contrast to the ’298 Patent, as Petitioner acknowledges, Dusija does
`
`not describe maintaining a write count for SLC and MLC blocks and allocating those
`
`blocks that receive the most frequent writes to SLC memory. This is not surprising
`
`given Dusija’s limited secondary use of SLC to reduce ECC complexity. Id., [0024]
`
`(describing the “advantage gained at the slight expense of…infrequent additional
`
`rewrites to a less density memory portion [i.e., SLC]”) (emphasis added).
`
`B.
`
`Sutardja (Ex. 1011)
`
`Sutardja describes a solid-state memory system having a controller and two
`
`separate solid-state nonvolatile memory referred to as the first and second NVS
`
`memories. Ex. 1011, Abstract; Ex. 2014, ¶¶60-64. According to Sutardja, “the first
`
`NVS memory has a first storage capacity that is greater than a second storage
`
`capacity of the second NVS memory.” Ex. 1011, [0012]. Both memories “may
`
`include single-level cell (SLC) flash memory or multi-level cell (MLC) flash
`
`memory.” Id., [0108].
`
`In contrast to the ’298 Patent, Sutardja states that the memories are treated “as
`
`if all the blocks formed a single memory.” Id., [0160-0162]. For example, Sutardja
`
`describes that “[w]hen a write request arrives from the host, the wear leveling
`
`module may select the block of memory that has been written to the least from
`
`14
`
`

`

`IPR2021-01547
`U.S. Patent No. 8,891,298
`
`among the available blocks” and “then maps the incoming logical address to the
`
`physical address of this block.” Id., [0111]. Sutardja further explains that “[o]ver
`
`time, this may produce a nearly uniform distribution of write operations across
`
`memory blocks.” Id.
`
`In further contrast to the ’298 Patent, Sutardja does not determine which of
`
`the blocks are accessed most frequently by maintaining a count of the number of
`
`times each one of the blocks is accessed. Instead, Sutardja maintains “normalized
`
`write … and/or erase cycle counts.” Id., [0122], [0160-0165]. Further, Sutardja
`
`relies on “write frequencies for logical addresses” to manage the solid-state memory
`
`system. Id., [0146]. As seen in Figure 7A below, every mention of “frequency” is
`
`referring to the “logical addresses.”
`
`15
`
`

`

`IPR2021-01547
`U.S. Patent No. 8,891,298
`
`
`Id., FIG. 7A. There is no mention in Figure 7A or the corresponding paragraphs
`
`[0145-0147] of determining which physical blocks are accessed most frequently. At
`
`the bottom of Figure 7A is a circle “B” to indicate that the flowchart continues to the
`
`
`
`next Figure 7B.
`
`16
`
`

`

`IPR2021-01547
`U.S. Patent No. 8,891,298
`
`
`
`
`Id., FIG. 7B. In step 514, the control determines whether it is time to perform data
`
`shift analysis. Id., [0148]. But Sutardja never explains when, if ever, it is time to
`
`perform the data shift analysis. Id.
`
`To the right of step 514, is a circle “C” to indicate that the flowchart continues
`
`to the next Figure 7C.
`
`17
`
`

`

`IPR2021-01547
`U.S. Patent No. 8,891,298
`
`
`
`
`Id., FIG. 7C. Figure 7C is the first time that the word “block” is used. But as before,
`
`there is no mention of determining which of the blocks are accessed most frequently
`
`by maintaining a count of the number of times each one of the blocks is accessed.
`
`In other words, these operations are performed on a per-block basis, and not on
`
`blocks. In step 520, the control determines if a number of write operations to a first
`
`block during a predetermined time is greater than or equal to a predetermined
`
`18
`
`

`

`IPR2021-01547
`U.S. Patent No. 8,891,298
`
`threshold. Id., [0149]. Sutardja never explains, however, what the “predetermined
`
`time” or “predetermined threshold” are.
`
`If the number of writes exceeds the threshold, the control maps the
`
`corresponding logical addresses to a block of the second NVS memory. Id., [0149],
`
`FIG. 7C (step 522). There is no mention in Sutardja of transferring the contents of
`
`the first block to SLC. In fact, Sutardja states that the second NVS memory may
`
`include SLC or MLC. Id., [0108]. To the extent the second NVS memory even
`
`includes SLC, there is no mention of transferring the data to SLC.
`
`As mentioned above, Sutardja maps data to blocks that have the least
`
`normalized wear, regardless of whether they are SLC or MLC. In essence, Sutardja
`
`attempts to create a memory system which is operated “as if all the blocks formed a
`
`single memory,” where data is written to either SLC and MLC (but not specifically
`
`SLC or MLC), based on the normalized wear metric. Id., [0161-0163]. As a result,
`
`Sutardja teaches away from giving MLC or SLC special treatment. Instead Sutardja
`
`is agnostic to whether the data is written to SLC or MLC, and only concerned with
`
`the normalized wear of the memory where the data is written to.
`
`C. Moshayedi (Ex. 1012)
`
`Moshayedi describes a flash memory drive having both single-level cell
`
`(SLC) and multi-level cell (MLC) memory with the described goals of
`
`“accommodat[ing] application memory needs at desirable prices, in addition to
`
`19
`
`

`

`IPR2021-01547
`U.S. Patent No. 8,891,298
`
`increasing read/write performance.” Ex. 1012, Abstract, [0008]; see also supra
`
`Section II.A (overview of SLC and MLC flash memory); Ex. 2014, ¶¶66-69. To
`
`accomplish these goals, the flash drive device of Moshayedi “keeps track of the
`
`number of times that data for each logical block address (LBA) has been written to
`
`the flash memory, and determines whether to store newly received data associated
`
`with a particular LBA in SLC flash or in MLC flash depending on the number of
`
`writes that have occurred for that particular LBA.” Id., Abstract (emphasis added).
`
`“For each logical block sent to the flash drive by the host, the host computer
`
`compares the write count of the associated LBA against a threshold.” Id., [0009],
`
`[0024]. “If the write count is above the threshold, the logical block is written to SLC
`
`flash.” Id. Further, “[t]he threshold may be set at 0 initially resulting in all data
`
`being written to SLC flash, and then increased as needed.” Id. Alternatively,
`
`Moshayedi describes that “[f]or some applications, e.g., when a drive h[a]s little or
`
`no stored data, the whole NAND flash area (SLC and MLC) may be used for writes
`
`… to give better performance on ‘new’ drives.” Id., [0050].
`
`In conjunction with this logical address approa