SAMSUNG ELECTRONICS CO., LTD.

Patent Trials and Appeals Board

Jun 25, 2021

2020000894 (P.T.A.B. Jun. 25, 2021)

UNITED STATES PATENT AND TRADEMARK OFFICE
UNITED STATES DEPARTMENT OF COMMERCE
United States Patent and Trademark Office
Address: COMMISSIONER FOR PATENTS
P.O. Box 1450
Alexandria, Virginia 22313-1450
www.uspto.gov
APPLICATION NO. FILING DATE FIRST NAMED INVENTOR ATTORNEY DOCKET NO. CONFIRMATION NO.
15/061,873 03/04/2016 Byoung Young AHN SSIM00009 US 8832
136129 7590 06/25/2021
Innovation Counsel LLP
2880 Zanker Road
Suite 201
San Jose, CA 95134
EXAMINER
KNIGHT, PAUL M
ART UNIT PAPER NUMBER
2139
NOTIFICATION DATE DELIVERY MODE
06/25/2021 ELECTRONIC
Please find below and/or attached an Office communication concerning this application or proceeding.
The time period for reply, if any, is set in the attached communication.
Notice of the Office communication was sent electronically on above-indicated "Notification Date" to the
following e-mail address(es):
docket@innovationcounsel.com
PTOL-90A (Rev. 04/07)
UNITED STATES PATENT AND TRADEMARK OFFICE
________________

BEFORE THE PATENT TRIAL AND APPEAL BOARD
________________ 

Ex parte BYOUNG YOUNG AHN, YANG SEOK KI, and
INSEOK STEPHEN CHOI

Appeal 2020-000894
Application 15/061,873
Technology Center 2100
________________

Before JOHN P. PINKERTON, JASON J. CHUNG, and
BETH Z. SHAW, Administrative Patent Judges.

CHUNG, Administrative Patent Judge.

DECISION ON APPEAL
Pursuant to 35 U.S.C. § 134(a), Appellant1 appeals the Final Rejection
of claims 1–19.2 We have jurisdiction under 35 U.S.C. § 6(b).
We AFFIRM.

INVENTION
The invention relates generally to storage devices and, more
particularly, to a storage device including a key-value integrated translation

1 We use the word “Appellant” to refer to “applicant” as defined in
37 C.F.R. § 1.42. According to Appellant, Samsung Electronics Co., Ltd. is
the real party in interest. Appeal Br. 3.
2 The Examiner withdrew the § 101 rejection of claims 1–19. Ans. 3.
Appeal 2020-000894
Application 15/061,873

2
layer. Spec. ¶ 2. Claim 1 is illustrative of the subject matter on appeal and
is reproduced below:
1. A storage device comprising:
a computer readable memory;
a non-volatile memory; and
a memory controller including a host interface for
interfacing with a host system and a memory interface for
interfacing with the non-volatile memory,
wherein a query received from the host system over the
host interface includes a key of a key-value data that is stored in
the non-volatile memory, wherein the memory controller
further includes a key-value translation layer including a table
indexer tree, one or more mapper tables, and one or more
location mappers,
wherein the table indexer tree contains first mapping
information for translating the key received over the host
interface to an index,
wherein the one or more mapper tables contain second
mapping information for obtaining a location of a location
mapper that contains an entry associated with the index, and
wherein the location mapper contains an address of the
key-value data or the key-value data associated with the entry in
the non-volatile memory.
Appeal Br. 56 (Claims Appendix).

REJECTIONS
Claims 1–3, 5–10, 12–15, and 17–19 are rejected under 35 U.S.C.
§103 as being unpatentable over Pundir (US 2016/0077744 A1, pub. March
17, 2016) in view of Marmol et al., NVMKV: A Scalable, Lightweight,
FTLaware Key-Value Store, USENIX Annual Technical Conference July 8–
10, 2015). Final Act. 5–12.
Appeal 2020-000894
Application 15/061,873

3
Claims 4, 11, and 16 are rejected under 35 U.S.C. § 103 as being
unpatentable over Pundir and Marmol in further view of Boyle (US
8,412,881 B2, iss. Apr. 2, 2003). Final Act. 13–14.

ANALYSIS
Claim 1 Rejected Under 35 U.S.C. § 103
A. The combination of Pundir and Marmol teaches the limitation “a key of a
key-value data that is stored in the non-volatile memory”
The Examiner finds that Marmol teaches Key-value (KV) stores that
employ a first layer to translate keys to Logical Block Addresses (LBAs)
and a second layer to translate the LBAs to physical locations in flash
devices, which teaches the limitation “a key of a key-value data.” Final Act.
5–6; Ans. 9–10 (citing Marmol § 3.1, second paragraph; § 5.1, first and
fourth paragraphs; § 5.3, last paragraph). The Examiner determines that
whether or not Pundir teaches a key is moot because Marmol teaches a key
that falls within Appellant’s desired scope for the term. Ans. 10. The
Examiner concludes that a person of ordinary skill of art would have been
motived to combine Marmol with Pundir’s three level data structure that
maps an LBA to physical block address (PBA) to achieve the claimed key
value translation layer to translate between a key and a PBA, because key-
value stores are ubiquitous in high performance data-intensive, scale out and
NoSQL environment and many KV stores are used in flash devices to meet
performance requirements. Final Act. 6–7; Ans. 9–10 (citing Marmol,
Abstract).
Appellant argues that Pundir’s logical unit number identifier (LUN
ID) or Logical Block Address (LBA) is not, and cannot be interpreted as the
claimed “key” of a key-value pair because the claimed storage device stores
Appeal 2020-000894
Application 15/061,873

4
and manages data records in a format of a data pair consisting of a key and a
corresponding value, while Pundir’s storage device merely stores data in
data blocks and accesses and retrieves the stored data based on its block
address. Appeal Br. 36–37 (citing Pundir ¶¶ 33, 39, 41). According to
Appellant, the stored data in Pundir does not include its own data address
(LUN ID or LBA) as a pair of the data address itself and the associated data.
Appeal Br. 37. Appellant also argues that Marmol’s mapping of a key to an
LBA does not transform Pundir’s input such as LBA or LUN ID into the
claimed key because Marmol teaches the key is a distinct entity that must be
translated into an LBA. Id. at 36. We disagree with Appellant.
At the outset, Appellant’s arguments in the preceding paragraph are
not persuasive because the Examiner relies on Marmol to teach the
limitation “a key of a key-value data that is stored in the non-volatile
memory.” In particular, the cited portion of Marmol teaches KV stores (i.e.,
this teaches Key-value database) employ a first layer to translate keys (i.e.,
this teaches key) to LBA and a second layer to translate the LBAs to
physical locations in flash devices (i.e., this teaches the non-volatile
memory), which teaches the limitation “a key of a key-value data that is
stored in the non-volatile memory.” Final Act. 5–6; Ans. 9–10 (citing
Marmol § 3.1, second paragraph; § 5.1, first and fourth paragraphs; § 5.3,
last paragraph). Furthermore, Marmol teaches “[m]ultiple KV stores on a
single flash device” and “the KV value pairs store in the flash device,”
which explicitly teaches the limitation “a key of a key-value data that is
stored in the non-volatile memory” recited in claim 1. See Marmol § 3.1,
second paragraph and § 5.1, fifth paragraph. We agree with the Examiner
Appeal 2020-000894
Application 15/061,873

5
that whether Pundir teaches a key is moot because Marmol teaches a key
that falls within the scope of the claimed limitation. Ans. 10.
Appellant presents for the first time in the Reply Brief arguments
against the rejection of claim 1. Reply Br. 3–4. Appellant has not shown
good cause for raising these new arguments, and we decline to consider an
argument raised for the first time in the Reply Brief, because the Examiner
has not been provided a chance to respond. See 37 C.F.R. § 41.41(b)(2)
(2018).
In particular, Appellant argues that even if Marmol teaches a key, the
combination of Marmol and Pundir still fails to teach the limitation “a query
received from the host system over the host interface includes a key of a
key-value data that is stored in non-volatile memory.” Reply Br. 10.
According to Appellant, it would be impossible to reasonably combine
Pundir with Marmo because Pundir does not contain any concept of a “key”
of a key value data received from the host system as a part of query, and
therefore the two disparate systems of Pundir and Marmol work in
completely different ways from the beginning of a query process. Id. 10–11.
Nonetheless, Appellant’s arguments above have ignored the combined
teachings and suggestions of Pundir and Marmol. Specifically, we find no
error with the Examiner’s determination that an ordinarily skilled artisan
would have had sufficient motivation to apply the teachings of Pundir—a
query received from the host system over the host interface including LUN
or LBA (see Final Act. 5–6)—to the teachings of Marmol—a key of a key-
value data that is stored in non-volatile memory and translate the key to the
LBA (id. at 6–7) collectively to arrive at the limitation “a query received
Appeal 2020-000894
Application 15/061,873

6
from the host system over the host interface includes a key of a key-value
data that is stored in non-volatile memory” recited in claim 1.
We agree with the Examiner’s that a person of ordinary skill of art
would have been motived to combine Marmol with Pundir’s three level data
structure that maps an LBA to physical block address (PBA) to achieve the
claimed key value translation layer to translate between a key and a PBA,
because key-value stores are ubiquitous in high performance data-intensive,
scale out and NoSQL environment and many KV stores are used in flash
devices to meet performance requirements. Final Act. 6–7; Ans. 9–10
(citing Marmol Abstract). The combination involves the predictable use of
prior art elements according to their established functions. The Supreme
Court has held that in analyzing the obviousness of combining elements, a
court need not find specific teachings, but rather may consider “the
background knowledge possessed by a person having ordinary skill in the art”
and “the inferences and creative steps that a person of ordinary skill in the art
would employ.” KSR Int’l Co. v. Teleflex Inc., 550 U.S. 398, 418(2007). To be
nonobvious, an improvement must be “more than the predictable use of prior
art elements according to their established functions,” id. at 417, and the basis
for an obviousness rejection must include an “articulated reasoning with some
rational underpinning to support the legal conclusion of obviousness.” Id. at
418 (citation omitted). Here, the Examiner has provided a sufficient rationale
for the combination, which is to use key-value stores to meet performance
requirements in a high performance data-intensive, scale out and NoSQL
environment. Final Act. 6–7 (citing Marmol, Abstract). Accordingly, we
find that the Examiner provides sufficient motivation for combining the
teachings of Pundir with the teachings of Marmol.
Appeal 2020-000894
Application 15/061,873

7
Moreover, Appellant does not provide persuasive evidence that
combining the respective teachings of the references (as proffered by the
Examiner – Final Act. 6–7) would have been “uniquely challenging or
difficult for one of ordinary skill in the art,” or that such a combination
would have “represented an unobvious step over the prior art.” Leapfrog
Enters., Inc. v. Fisher-Price, Inc., 485 F.3d at 1162. Nor does Appellant
provide any objective evidence of secondary considerations, which, as our
reviewing court explains, “operate[] as a beneficial check on hindsight.”
Cheese Systems, Inc. v. Tetra Pak Cheese and Powder Systems, Inc., 725
F.3d at 1352.
Accordingly, Appellant does not persuade us of error in the
Examiner’s conclusion that a person having of ordinary skill in the art at the
time of the invention would have combined Marmol with Pundir to teach the
limitation “a key of a key-value data that is stored in the non-volatile
memory” recited in claim 1.
B. The combination of Pundir and Marmol teaches the claimed storage
device and Pundir performs Key-to address mapping in the storage device
The Examiner finds that Pundir teaches a storage controller and a
storage I/O stack including a three-level mapping dense tree operate on a
node to access information requested by a host, and Marmol teaches a key is
sent to Flash Translation Layers (FTLs), and the combination of Pundir and
Marmol teaches the limitation “[a] storage device comprising: . . . the
memory controller includes a key-value translation layer.” Final Act. 5–8;
Ans. 10–11 (citing Pundir ¶¶ 34, 55 and Marmol § 3.1, second paragraph;
§ 5.1, first and fourth paragraphs; § 5.3, last paragraph). The Examiner
determines that Marmol teaches the key being sent to FTLs (FTL is the
software layer run in the controller and drives some logic to physical
Appeal 2020-000894
Application 15/061,873

8
mapping) and the dense tree of Pundir performs the operation of FTL, the
combination of Pundir and Marmol teaches the key-translation mapping
occurred in the memory controller. Id. at 12. The Examiner further
concludes that modifying the type of key taught in Pundir based on the
teaching of Marmol does not require moving the memory management
software FTL to the location used in Marmol. Id.
Appellant argues that Pundir and Marmol do not teach the claimed
storage device because both Pundir and Marmol perform their key-to-
address mapping at locations outsides of the storage device. Appeal Br. 43.
According to Appellant, the block-based storage devices disclosed in Pundir
and Marmol are fundamentally different from the claimed storage device
because the Examiner finds that Pundir’s storage device does not store data
as a key-value pair. Id. at 38 (citing Final Act. 8). Appellant also argues
that even if Pundir’s system disclose key-value pairs, Pundir failed to teach
the claimed storage device that stores data in as a key-value pair because the
key-value pairs are stored and maintained in a cluster database separately
and independently from the storage controller. Id.
In addition, Appellant argues that Pundir fails to teach the claimed
storage device because the claim subject matter entirely eliminates “double
mapping” scheme (i.e., key-value pairs to LBAs on the host, and LBAs to
PBAs on the storage device) using the claimed key-value translation layer
that provides improvements in term of efficiency and the size of
conventional “double mapping” schemes such as Pundir. Id. Appellant
further argues Marmol does not cure the deficiencies of Pundir because
Marmol’s host system uses a block-based address scheme to access the
Appeal 2020-000894
Application 15/061,873

9
block based flash device by perform all the computation and translation in
the host system. Id. at 39–43. We disagree with Appellant.
Pundir teaches a storage controller and a storage I/O stack (i.e., this
teaches a memory controller) including a three-level mapping dense tree
(i.e., this teaches translation layer) operates on a node (i.e., storage device)
to access information requested by a host. Final Act. 5–8; Ans. 10–11
(citing Pundir ¶¶ 34, 55); see also Pundir ¶ 32 (“FIG. 2 is a block diagram of
node 200 that is illustratively embodied as a storage system” (emphasis
added)). The cited portion of Marmol teaches a key is sent to FTLs. Final
Act. 6–7; Ans. 10–11 (citing Marmol § 5.1, first paragraph).
Appellant’s arguments above have ignored the combined teachings
and suggestions of Pundir and Marmol. Specifically, we find no error with
the Examiner’s finding that an ordinarily skilled artisan would have had
sufficient motivation to apply the teachings of Pundir—a storage device
comprises a memory controller with a translation layer (see Final Act. 5–
8)—to the teachings of Marmol—a key-translation layer (id. at 6–7)
collectively to arrive at the limitation “[a] storage device comprising: . . . the
memory controller further includes a key-value translation layer” recited in
claim 1.
We also agree with the Examiner’s conclusion that the combination of
Pundir and Marmol teaches the key-translation mapping occurred in the
memory controller and modifying the type of key taught in Pundir based on
the teaching of Marmol does not require moving the memory management
software FTL to the location used in Marmol. Id. at 12. Pundir discloses the
node 200 is a storage system (i.e., this teaches a storage device) and a
storage I/O stack with a three-level mapping dense tree operates on the node
Appeal 2020-000894
Application 15/061,873

10
200, which explicitly teach to perform translation mapping inside the
storage device. See Pundir ¶¶ 32, 54–55 (emphasis added).
We disagree with Appellant’s argument that because Pundir does not
entirely eliminate the “double mapping” scheme, Pundir fails to teaches the
claimed storage device. Appeal Br. 38–39. “[A]ppellant’s arguments fail
from the outset because . . . they are not based on limitations appearing in
the claims.” See In re Self, 671 F.2d at 1348 (CCPA 1982). In this case, the
phrase “eliminate the double mapping scheme” is not recited in claim 1.
Therefore, Appellant does not persuade us of error in the Examiner’s
conclusion that the combination of Pundir and Marmol teaches the limitation
“[a] storage device comprising: . . . the memory controller . . . includes a
key-value translation layer” recited in claim 1.
C. Appellant’s arguments pertaining to caching and eliminating double
mapping are misplaced because these features are not recited in the claims
Appellant argues that the combination of Pundir and Marmol does not
teach the three layer index structure as claimed because the traditional block-
based protocols in Pundir and Marmol require double mapping that involves
translation from a key to an LBA, and a translation from the LBA to a PBA.
Appeal Br. 44 (citing Final Act. 8). According to Appellant, the claimed
“three-layer index structure” eliminates at least a second mapping for the
block-to-block translation (i.e., from LBAs to PBAs) as required by
conventional block-based protocols because the claimed storage device can
directly receive “a query” including “a key of a key-value data” from the
host system over the host interface bypassing at least the block-to-block
translations from LBAs to PBAs. Appeal Br. 44 (citing Spec. Fig. 3).
Appellant also argues, in comparison to “an LBA space” that has a limitation
of a physical storage, a “key Space” employed by the claimed three-layer
Appeal 2020-000894
Application 15/061,873

11
index structure does not have the limitation of the physical storage capacity
because the Specification recites “[d]ue to the limited size of the memory of
the storage device, a tree index can be split into smaller sized subtrees, and
those subtrees may be dynamically loaded (cached) in the memory of the
memory controller.” Id. (citing Spec. ¶¶ 40–44). We disagree with
Appellant.
We agree with the Examiner that cacheability is not recited in the
claim. Ans. 13. In addition, we disagree with Appellant’s argument that
because Pundir and Marmol require double mapping scheme including a
second mapping for the block-to-block translation, the combination of
Pundir and Marmol fails to teach the claimed “three-layer index structure.”
Appeal Br. 38–39. “[A]ppellant’s arguments fail from the outset because . .
. they are not based on limitations appearing in the claims.” See In re Self,
671 F.2d at 1348 (CCPA 1982). In this case, “eliminate at least the second
mapping for the block-to-block translation” is not recited in claim 1.
Accordingly, Appellant does not persuade us of error in the
Examiner’s conclusion that the combination of Pundir and Marmol teaches
the three layer index structure as claimed.
D. Pundir teaches the limitation “a table indexer tree, one or more
mapper tables, and one or more location mappers”
The Examiner finds Pundir’s level 0 of the dense tree structure
contains information for translating the received data to an index, which
teaches the limitation “the table indexer tree contains first mapping
information for translating the key received over the host interface to an
index” recited in claim 1. Final Act. 8; Ans. 14 (citing Pundir ¶ 55). The
Examiner also finds Pundir’s level 1 of the dense tree structure points to
level 2 of the dense tree structure, which teaches the limitation “the one or
Appeal 2020-000894
Application 15/061,873

12
more mapper tables contain second mapping information for obtaining a
location of a location mapper that contains an entry associated with the
index” recited in claim 1. Final Act. 8; Ans.14 (citing Pundir ¶ 64). In
addition, the Examiner finds Pundir’s level 2 of the dense tree structures
contains data entries that include an address called data entry, which teaches
the limitation “the location mapper contains an address of the key-value data
or the key-value data associated with the entry in the non-violate memory”
recited in claim 1. Final Act. 8; Ans. 14 (citing Pundir ¶¶ 61, 64). The
Examiner concludes that Pundir teaches the claimed three level key valued
structure because all three levels of the recited “key-value translation layer”
read on the three levels of the dense tree structure of Pundir. Ans. 5.
Appellant argues that the combination of Pundir and Marmol fails to
teach or suggest the claimed key-value translation layer including “a table
indexer tree, one or more mapper tables, and one or more location mappers”
because the Examiner merely recites a single tree (such as the multi-level
dense tree) and fails to explain how each level of the multi-level dense tree
respectively corresponds to the claimed “table indexer tree,” “one or more
mapper tables,” and “one or more location mappers.” Appeal Br. 45–46.
According to Appellant, the plain meaning of the term “table” is “an
arrangement of data in rows and columns,” and the plain meaning of “tree”
is a “data type that simulates a hierarchical tree structure, with a root value
and subtrees of children with a parent node, represented as a set of linked
nodes” in accordance with Wikipedia and these Wikipedia definitions are
consistent with the Specification. Reply Br. 5–6 (citing Spec. ¶¶ 9–10, 43–
Appeal 2020-000894
Application 15/061,873

13
44, 55).3 Appellant argues Pundir does not teach key-value translation layer
because the claimed “table indexer tree” and “mapper tables” are structurally
different from those provided in Pundir based on the plain meanings of the
claim terms “tree” and “table.” Reply Br. 5–8, 18–19. In particular,
Appellant argues that Pundir’s multiple levels of nodes in a single tree
structure are not, and cannot be read onto, the claimed “mapper table”
because a tree (or nodes of a tree) is not a table. Id. at 6. We disagree with
Appellant.
Pundir teaches level 0 of the dense tree structure contains information
for translating the received data to an index, which teaches the limitation
“the table indexer tree contains first mapping information for translating the
key received over the host interface to an index” recited in claim 1. Final
Act. 8; Ans. 14 (citing Pundir ¶ 55). Pundir also teaches level 1 of the dense
tree structure points (i.e., this teaches mapping) to level 2 of the dense tree
structure (i.e., this teaches a location of a location mapper), which teaches
the limitation “the one or more mapper tables contain second mapping
information for obtaining a location of a location mapper that contains an
entry associated with the index” recited in claim 1. Final Act. 8; Ans. 14
(citing Pundir ¶ 64). In addition, Pundir teaches level 2 of the dense tree
structures contains data entries that includes address called data entry (i.e.,

3 We determine that Appellant’s argument in the Reply Brief pertaining to
the plain meaning of the term “table” and “tree” being executed sequentially
is permissible because the Examiner clarifies the table indexer tree, the
mapper table, and the location mapper in Pundir for the first time in the
Answer. Compare Ans. 13–14 (the Examiner clarifies the table indexer tree,
the mapper table, and the location mapper in Pundir), with Final Act. 7–9
(lacking clarification of the table indexer tree, the mapper table, and the
location mapper).
Appeal 2020-000894
Application 15/061,873

14
this teaches address associated with the entry), which teaches the limitation
“the location mapper contains an address of the key-value data or the key-
value data associated with the entry in the non-volatile memory” recited in
claim 1. Final Act. 8; Ans. 14 (citing Pundir ¶¶ 61, 64). We agree with the
Examiner’s finding that Pundir teaches the claimed three level key value
structure because all three levels of the recited “key-value translation layer”
read on the three levels of the dense tree structure of Pundir. Ans. 5.
We also note that paragraph 65 of Pundir teaches “the top level (level
0) of the dense tree 700 is maintained in-core as a balanced tree.” See
Pundir ¶ 65. Further, Figure 7 of Pundir teaches level 1 of the dense tree
structure includes a five-column one row table, which teaches the limitation
“mapper tables.” See Pundir Figure 7.
Accordingly, Appellant does not persuade us of error in the
Examiner’s findings that the Pundir teaches “table indexer tree,” “one or
more mapper tables,” and “one or more location mappers” as recited in
claim 1.
Appellant does not argue claims 2–19 separately with particularity.
Therefore, we sustain the Examiner’s rejection of: (1) independent claims 1,
8 and 13; and (2) dependent claims 2–7, 9-12, 14–19 under 35 U.S.C. § 103.
We have only considered those arguments that Appellant actually
raised in the Briefs. Arguments Appellant could have made, but chose not to
make, in the Briefs have not been considered and are deemed to be waived.
See 37 C.F.R. § 41.37(c)(1)(iv).

CONCLUSION
The Examiner’s rejections for claims 1–19 is Affirmed.
Appeal 2020-000894
Application 15/061,873

15
DECISION SUMMARY
In summary:
TIME PERIOD FOR RESPONSE
No time period for taking any subsequent action in connection with
this appeal may be extended under 37 C.F.R. § 1.136(a)(1)(iv). See
37 C.F.R. § 1.136(a)(1)(iv).

AFFIRMED

Claim(s)
Rejected
35 U.S.C.
§
Reference(s)/Basis Affirmed Reversed
1–3, 5–10,
12–15, 17–
19
103 Pundir, Marmol 1–3, 5–10,
12–15, 17–
19

4, 11, 16 103 Pundir, Marmol,
Boyle
4, 11, 16
Overall
Outcome
1–19