Board finds disclosure that memory access time period exists is not enough to anticipate decision based on this time period

Posted by: Karen G. Hazzah

Takeaway:  At issue in this appeal was the limitation "threshold defining both a number of times a segment is accessed and a given time period for the number of times." The Applicant did not dispute that the counter in the secondary reference tracked page accesses exceeding a threshold, but did argue that the reference did so without regard to a time period. The Examiner asserted that the threshold in the reference implicitly defined a time period, in that "when the threshold is reached the amount of time it took to reach it is also defined." The Board found this interpretation to be unreasonable. The Board acknowledged that the threshold determination "inevitably involve[d] a period of time to count those events," which was "simply the time elapsed during the counting process." However, the Board found that the reference "does not measure, let alone account for [this time period] as a factor for migration" as claimed. The Board reversed the rejection. (Ex parte Jones, PTAB 2011.)

Details:

Ex parte Jones
Appeal 2009011684; Appl. No. 11/066,038; Tech Center 2100
Decided:  August 9, 2011

The application on appeal was directed to a multiprocessor memory system that caches frequently used data in a centralized switch memory.

     1. A system comprising:
     a plurality of nodes, each node comprising a processor coupled to a memory; and
     a switching device coupled to the plurality of nodes and containing
          migration logic configured to migrate segments of each memory to a memory of the switching device so that segments are accessible to each of the processors. 

To overcome an obviousness rejection, the Applicant amended claim 1 to pull in dependent claim 4, as follows: 

migration logic that is configured to identify segments of memory that are each accessed above a threshold number of times over a given time period and that is configured to migrate segments of each memory so identified to a memory of the switching device.

The Applicant distinguished the secondary reference by arguing that Schonias did not teach tracking memory access "over a threshold number with regard to a given time period." (Emphasis in original.)

The Examiner maintained the rejection, explaining that the claim language "does not require that the threshold must be met within a specific time period instead it just requires that the threshold is met over some time period, which of course is true." (Emphasis added.)

In response, the Applicant amended again as follows:

migration logic that is configured to identify segments of memory that each accessed above a threshold, said threshold defining both a number of times a segment is accessed over and a given time period for the number of times, and that is configured to migrate segments of each memory so identified to a memory of the switching device"

The Applicant again argued that Schonias tracked page accesses that exceed a threshold, but without regard to a time period.

The Examiner maintained the rejection and took the Applicant final. According to the Examiner, the claim "merely states that 'a time period' is defined by the threshold" but not "how the time is defined by the threshold or what variable the time is defined with respect to, or that the time can't vary." The Examiner then explained that in Schonias, "the time period is defined by how long is takes to reach the count threshold of the counter, therefore the threshold being reached defines the amount of time it took to reach it."

The Applicant appealed and argued that the Examiner's claim construction was erroneous:

The Examiner apparently takes the position that because it will take some finite period of time for the counter threshold in Schoinas to eventually be reached, that the count threshold reaches does define an amount of time. However, the claim requires the migration logic to implement a threshold which defines both the number of times and the time period. Schoinas' counter threshold only defines a number of accesses, not a corresponding time period. The Examiner's reading of the claim is disingenuous and counter to what one ordinary skill in the art would understand upon reading Appellant's disclosure.

In the Answer, the Examiner first noted that the claim does not require a predefined time period, since neither the claims nor the specification provided "any specific explanation as to how the time period of these claims are defined." The Examiner then explained that the broadest reasonable interpretation covered a time period defined in terms of a number of memory accesses – which does read on a time period, even if that period is not predefined. Applying this to Schoinas, "when the threshold is reached the amount of time it took to reach it is also defined."

The Applicant filed a Reply Brief and again distinguished the claim from Schoinas:

In Schoinas, it will not matter whether it took 5 minutes or 5 hours to reach the terminal count for accessing a given memory area. In claim 1, it will matter. In Schoinas, a memory area that is accessed, say, 50 times is migrated even if it took a week to be accessed that number of times. The number of accesses alone is not enough for claim 1; the corresponding time period is also considered.

The Board reversed the rejection. The Board framed the dispositive issue as whether Schoinas' threshold defines a time period that is associated with the number of times a memory segment is accessed. The Board found that the Examiner had acknowledged that Schoinas' threshold did not explicitly define the time period. The Board further found that the Examiner relied on a defined time period that was "implicit in the length of time take to reach the specific count threshold." The Board found this reading of Schoinas was unreasonable in light of the specification:

     To be sure, counting discrete events, such as memory accesses, to determine whether those events reach a predetermined threshold amount will inevitably involve a period of time to count those events. But that hardly means that this inexorable passage of time is accounted for in this determination. It is simply the time elapsed during the counting process – a period of time that Schoinas does not measure, let alone account for as a factor for migration.
     By specifically defining both a count and a time period, the claimed threshold sets particular limits on the rate at which counts occur (i.e. their frequency) – a key temporal restriction that Schoinas simply does not contemplate.

My two cents: The Board recognized the Examiner's flawed logic, which boils down to what I call a "mere existence" analysis. The Examiner took a teaching of the existence of X (here, a time period) and asserted that the mere existence of X satisfied the claim limitation. Here, the claim as properly interpreted required not simply the existence of a time period during which memory was accessed, but also an action – identifying a segment – based on this property.