[00:00:00] Speaker 03:
First is number 22, 1936, Netflix, Inc.

[00:00:04] Speaker 03:
versus Avago Technologies, Mr. Batts.

[00:00:08] Speaker 00:
Good morning, Your Honors.

[00:00:09] Speaker 00:
May it please the court, Carver Batts, on behalf of Appellant Netflix.

[00:00:12] Speaker 00:
There are two distinct issues that I'd like to discuss with the panel this morning.

[00:00:17] Speaker 00:
And I'd like to start with the issue of Master's Figure 9, the only issue that- Could you back up a moment first?

[00:00:23] Speaker 03:
I mean, the board relied to a large extent on

[00:00:29] Speaker 03:
It's construction of the term performance data.

[00:00:33] Speaker 00:
Yes.

[00:00:33] Speaker 00:
And that's where I was going, Your Honor.

[00:00:35] Speaker 03:
Individual instance data.

[00:00:36] Speaker 03:
Correct.

[00:00:37] Speaker 03:
You disagree with that and based on the one or more language in the claim, you say it's broader than that.

[00:00:44] Speaker 03:
My question is how broad is it in your view?

[00:00:48] Speaker 00:
Yes, Your Honor.

[00:00:48] Speaker 00:
That is the reason I wanted to start with figure nine is because that is the only issue that needs to be addressed for the arguments regarding master's figure nine.

[00:00:57] Speaker 00:
If this court on de novo review rejects the board's claim construction, that's the only issue, the only basis for the rejection.

[00:01:04] Speaker 01:
But in response to Judge Dyck's point, it would be helpful if you would offer what you think is a correct claim construction.

[00:01:12] Speaker 00:
Sure.

[00:01:13] Speaker 00:
So we believe that it is clear from the claim language itself, claim one's claim language,

[00:01:19] Speaker 00:
the dependent claim language, the specification of all those experts' testimony, as well as their infringement contentions, that what refers to performance data includes not only instance-level data, but it also refers to the other, like the tier levels, the other application tier, and the host tier level.

[00:01:36] Speaker 00:
Host and tier-level data.

[00:01:37] Speaker 00:
Correct.

[00:01:38] Speaker 00:
And I think that's consistent with the claim language.

[00:01:40] Speaker 03:
Is tier-level data broader than house-level data?

[00:01:44] Speaker 00:
There's different types of tiers.

[00:01:46] Speaker 00:
I'd say an application tier would be narrower, and then a host tier would be broader, but I think there are actually different tiers.

[00:01:53] Speaker 00:
I don't know.

[00:01:55] Speaker 00:
When you're asking which one is broader, I think they're different.

[00:01:58] Speaker 03:
Well, I mean, you're relying on dependent claim five to say that performance data includes tier-level data, right?

[00:02:07] Speaker 03:
Correct.

[00:02:07] Speaker 03:
So what is tier-level data?

[00:02:10] Speaker 03:
How is it defined?

[00:02:11] Speaker 03:
I know there is reference to it in the specification here.

[00:02:15] Speaker 03:
What is your understanding of how tier level data is defined?

[00:02:19] Speaker 00:
Sure, so application, I'll do both in turn.

[00:02:22] Speaker 00:
So application level data, for example, if you're running Adobe Acrobat, you would have multiple copies of Adobe, that's application level.

[00:02:29] Speaker 00:
Application tier would be a series of applications.

[00:02:32] Speaker 00:
So you could have Adobe, maybe Microsoft Word, other applications.

[00:02:36] Speaker 00:
You're talking about a series of applications, and you're looking at the tier of the applications.

[00:02:42] Speaker 00:
Host-level data then is talking about, like for example, the actual server that's being run on.

[00:02:47] Speaker 00:
You're looking at the server, like the CPU utilization.

[00:02:51] Speaker 03:
Well, how does tier-level data relate to host-level data?

[00:02:54] Speaker 03:
And Master's relates to host-level data, right?

[00:02:57] Speaker 00:
Well, clearly Master's does relate to host-level data, but it has all the sets of data.

[00:03:02] Speaker 00:
It's saying it's collecting the data at all these different blocks.

[00:03:05] Speaker 03:
Okay, so how does host-level data

[00:03:07] Speaker 03:
relate to tier-level data?

[00:03:12] Speaker 03:
How is it different?

[00:03:14] Speaker 00:
Host-level data is looking at

[00:03:18] Speaker 00:
how the application instance, the relationship of the application instance to the host and what is occurring on the host.

[00:03:25] Speaker 00:
So for a CPU... Particular machine.

[00:03:27] Speaker 00:
For a particular machine.

[00:03:28] Speaker 00:
So if you're looking at the CPU utilization of a machine and you're looking at multiple instances of the application running on that machine, you can look at what is the correlation of that data between the application instance and the host.

[00:03:40] Speaker 00:
I don't want to jump right away to figure eight and master's appendix there, but I think that's also good examples of where you can see the actual host data, the CPU usage data, and you see how it's being affected by different application instances either being created, destroyed, or migrated.

[00:03:58] Speaker 00:
So what we have is if you look in the specification, we sign it to the...

[00:04:02] Speaker 03:
I understand your argument about dependent claim five and tier level data, but what's the basis for saying that claim one also covers host level data?

[00:04:17] Speaker 00:
So I would point, Your Honor, to appendix 1428.

[00:04:20] Speaker 00:
This was Avago's expert's testimony on this exact question, and we asked him about what do you mean by characterizing the performance of said application instances.

[00:04:29] Speaker 01:
Is 1428?

[00:04:31] Speaker 00:
Appendix 1428.

[00:04:32] Speaker 00:
Is Dr. Rosenblum?

[00:04:33] Speaker 00:
Correct.

[00:04:34] Speaker 00:
Any states close.

[00:04:35] Speaker 03:
So does your jumping to that suggest that there's no intrinsic evidence to suggest one way or the other whether it applies to house-level data?

[00:04:44] Speaker 00:
Well, I certainly think we look at the claim language itself.

[00:04:49] Speaker 00:
Phillips, you start with the claim language.

[00:04:50] Speaker 00:
Claim one is the language of claim one itself

[00:04:54] Speaker 00:
is not limited to application instance level, because it's saying application instance level and then it has a wherein clause.

[00:05:00] Speaker 03:
It could have just ended at that.

[00:05:01] Speaker 03:
Wait, that's a different issue.

[00:05:03] Speaker 03:
We've gotten beyond that.

[00:05:04] Speaker 03:
We're assuming for the moment that you're right, that because of dependent claim five, it expands beyond individual instance data to tier-level data.

[00:05:13] Speaker 03:
The question is, what's the intrinsic evidence that shows that it applies to host-level data?

[00:05:21] Speaker 00:
It's the specification, then, Your Honor, because the specification has clear examples of host-level data.

[00:05:27] Speaker 00:
So I would point you to appendix page 43, column 4, lines 3 through 8.

[00:05:34] Speaker 00:
that there's collection that can gather and analyze data for particular targets, e.g.

[00:05:40] Speaker 00:
web clients, web servers, networks, application servers.

[00:05:44] Speaker 00:
All these are referring to host level data.

[00:05:47] Speaker 00:
And that's what I was saying was consistent with Dr. Rosenbaum's testimony on Appendix 1428, where he says, the performance data about CPU utilization, so host-level data, could be associated with individual application instances of an application.

[00:06:04] Speaker 00:
So we have, the tie-in is that Avago has conceded that the construction is excluding embodiments.

[00:06:13] Speaker 00:
In their briefing, they conceded the fact that this construction excludes embodiments.

[00:06:18] Speaker 00:
And those embodiments include the application tier as well as the host level data.

[00:06:22] Speaker 00:
And they haven't shown any probative evidence, any explanation for why you would exclude embodiments here.

[00:06:27] Speaker 00:
And we have the dependent claims that are showing that the board's construction is clearly wrong, and it's not limited as they construed it to the application instance level.

[00:06:37] Speaker 01:
Mr. Bass, you cited appendix 43.

[00:06:40] Speaker 01:
What column?

[00:06:41] Speaker 00:
What line?

[00:06:42] Speaker 00:
Column 4, lines 3 through 8, Your Honor.

[00:06:45] Speaker 00:
OK.

[00:06:48] Speaker 00:
And then, I mean, there's more.

[00:06:49] Speaker 00:
There's appendix 43, column 3, lines 56 to 58.

[00:06:54] Speaker 00:
where it discusses the performance metrics on servers running target applications.

[00:07:01] Speaker 00:
So the specification is consistent with Avago's experts testimony.

[00:07:05] Speaker 00:
And it's also consistent with their infringement contentions from the district court where they pointed to CPU usage.

[00:07:10] Speaker 00:
That's the metric that they pointed to in the district court to say that would be satisfied.

[00:07:14] Speaker 01:
Now, are you urging us to the... It's not really clear, I guess, if the... By implication, there was kind of a board claim construction here.

[00:07:25] Speaker 01:
But it's not really clearly spelled out, I think.

[00:07:28] Speaker 01:
Correct me if you think I'm wrong.

[00:07:29] Speaker 01:
But do we need to... Should we construe the claim or...

[00:07:33] Speaker 01:
If we agree with you or should it go back to the board for construction?

[00:07:38] Speaker 00:
I think you can do either, your honor.

[00:07:41] Speaker 00:
I think clearly we set forth in our reply brief, you know, as I mentioned, the infringement contentions, they said very broad.

[00:07:47] Speaker 00:
They raised in their patent on response, a narrowing construction.

[00:07:50] Speaker 00:
In our reply brief, we had a section called claim construction.

[00:07:52] Speaker 00:
This was in the section.

[00:07:53] Speaker 00:
We set performance data.

[00:07:54] Speaker 00:
They're wrong about this narrowing.

[00:07:56] Speaker 00:
The board discussed it at the oral hearing.

[00:07:58] Speaker 00:
So I think there's clearly, and then you see the decision, there's clearly, they even meant, they call it an implicit claim construction.

[00:08:03] Speaker 00:
It's a claim construction, your honor.

[00:08:05] Speaker 00:
So I think it is for de novo review.

[00:08:07] Speaker 00:
And I think it's clear that they erred in limiting it in this fashion so that you can either provide them the construction saying, here's what's wrong and it should be this broad, or you can send it back and remand it back for them to make that determination with your guidance.

[00:08:21] Speaker 00:
I think either way is appropriate.

[00:08:23] Speaker 03:
Okay, if we're past the performance date, unless Judge Schall has further questions about that, let's talk about the correlation issue, which, as I understand it, we would have to decide with respect to Figure 8, because the board's decision with respect to Figure 8 of Masters rests on two grounds, one, performance data, and

[00:08:43] Speaker 03:
The second is the absence of correlation, even if you assume that the figure shows performance data.

[00:08:49] Speaker 00:
Correct.

[00:08:50] Speaker 00:
And I do think that the board clearly abused its discretion in not considering Master's Appendix J. Appendix is on a CD, a physical CD, but it's part of the Master's prior art reference.

[00:09:05] Speaker 00:
You're saying it's intrinsic evidence.

[00:09:07] Speaker 00:
It's intrinsic evidence, Your Honor, and we provided it to them.

[00:09:10] Speaker 00:
There's multiple reasons why this would be rebuttal evidence, proper evidence, intrinsic evidence, whatever.

[00:09:15] Speaker 00:
The Board, literally, the decision is silent as to this evidence.

[00:09:18] Speaker 00:
And Avada's response in the briefing here was, well, they disregarded it.

[00:09:22] Speaker 00:
It was untimely.

[00:09:23] Speaker 00:
There's not even briefing on that issue below, let alone a ruling on that below.

[00:09:28] Speaker 00:
And if you look at Master's, Master's PENIX-J, and I really want to go to PENIX 1338,

[00:09:35] Speaker 00:
through appendix 1341.

[00:09:36] Speaker 00:
It is spot on, blows up every argument Avago has about figure eight.

[00:09:43] Speaker 00:
This is the path display.

[00:09:44] Speaker 00:
The figure eight is the path display.

[00:09:47] Speaker 00:
And on 1338, second paragraph, it talks about the primary purpose of the path display is to show the conditions in the system, and then it goes on to say for scale up events, scale down events, and moves.

[00:10:00] Speaker 00:
Which page is this again?

[00:10:01] Speaker 00:
1338, you're on.

[00:10:02] Speaker 00:
So the second paragraph.

[00:10:04] Speaker 00:
So you see in the second sentence there, it talks about the different actions.

[00:10:07] Speaker 00:
So a creation event, a destruction event, and a move, a migration event.

[00:10:12] Speaker 00:
And it goes on to talk about the QoS metrics, so the performance metrics that are occurring there.

[00:10:18] Speaker 00:
And it keeps going.

[00:10:20] Speaker 00:
The detail on these four pages is key, because the very next page talks about a message, the first line.

[00:10:26] Speaker 00:
The message notifies the past display when a new process is started, i.e.

[00:10:30] Speaker 00:
a creation event.

[00:10:31] Speaker 00:
the path display adds this process to the display.

[00:10:35] Speaker 00:
So there can't be any doubt that when a new creation event occurs, what you see for a new line in this colored figure, this is a colored version of the figure if you look on page 1341, this is a colored version of figure eight from masters.

[00:10:49] Speaker 00:
So you can see each line has a different color and represents a different application instance and what's occurring over time

[00:10:57] Speaker 00:
for the performance metric of review time.

[00:11:00] Speaker 00:
So you have a correlation.

[00:11:02] Speaker 03:
Our 1341 doesn't show any colors.

[00:11:03] Speaker 00:
1341?

[00:11:05] Speaker 03:
Yes, it does, your honor.

[00:11:07] Speaker 03:
Not ours.

[00:11:08] Speaker 03:
Not in the appendix.

[00:11:10] Speaker 00:
Oh, I apologize, your honor.

[00:11:12] Speaker 00:
to see colors there.

[00:11:14] Speaker 00:
Oh, you've got some colors.

[00:11:16] Speaker 00:
There's colors in it, and you can see the tie-in from the specific instance.

[00:11:21] Speaker 00:
But your copy is not the same as ours.

[00:11:23] Speaker 03:
If you want us to look at a colored version, you've got to supply us with a colored version.

[00:11:27] Speaker 00:
I apologize, Your Honor.

[00:11:29] Speaker 00:
I'll look into that immediately.

[00:11:30] Speaker 01:
That would be helpful, yeah, because yours does, but as Judge Dyck was saying... But even without the colors, Your Honor, this language explains the specific

[00:11:38] Speaker 00:
how a creation event or a migration event or a destruction event is being added into the graph.

[00:11:42] Speaker 00:
And what we have here when you ask about your correlation question, you have here the specific, you can see the specific metrics for each application instance.

[00:11:50] Speaker 00:
So this is not, it doesn't matter what the claim construction is, for each application instance, it's showing the correlation.

[00:11:57] Speaker 00:
And this was the exact

[00:11:58] Speaker 00:
example in the specification, because if you ask for correlation, well, what could be correlation?

[00:12:03] Speaker 00:
Well, again, their expert and the specification are consistent, because the specification on column two, lines three through seven, appendix 42, states correlating the performance data to one or more lifecycle events may also comprise determining the change in performance of the application

[00:12:20] Speaker 00:
as the one or more application instances are created and destroyed.

[00:12:23] Speaker 01:
Where is this exactly, Mr. Patz?

[00:12:24] Speaker 00:
Column 2, lines 3 through 7 of the 722 patent, Your Honor.

[00:12:29] Speaker 00:
Column 2, lines what?

[00:12:31] Speaker 00:
3 through 7.

[00:12:32] Speaker 00:
Of the 3, 722.

[00:12:33] Speaker 00:
722, Your Honor.

[00:12:35] Speaker 00:
So what does the patent itself say as an example of correlation?

[00:12:39] Speaker 00:
And that's exactly consistent with Dr. Rosenblum's testimony on Appendix 1453 that we also cited, where when we asked him about what would be for claim one, what would constitute correlating for claim one, and he said, perhaps the most basic kind of correlation would be to say, here's what these performance metrics, here's the values these different performance metrics had at the time of a particular life cycle event.

[00:13:04] Speaker 00:
That's exactly what's being shown in Master.

[00:13:06] Speaker 00:
They said, hey, Master's Figure 8 doesn't really show this.

[00:13:09] Speaker 00:
We have the actual intrinsic evidence from Master's Appendix J that shows their arguments are wrong.

[00:13:17] Speaker 00:
And we discussed this at the oral hearing, Your Honor.

[00:13:19] Speaker 00:
We had the arguments about it.

[00:13:20] Speaker 00:
It was submitted to them before three depositions of our expert.

[00:13:24] Speaker 00:
And the board just doesn't address it in their decision.

[00:13:27] Speaker 00:
I don't know what could be a further example of an abuse of discretion under an APA and an agency not considering

[00:13:34] Speaker 00:
the exact evidence that shows why the arguments were wrong.

[00:13:38] Speaker 03:
So the bottom line is you want us to decide performance data and you want us to decide correlation taking into account Appendix J. That's it, right?

[00:13:49] Speaker 00:
I think those are the two issues.

[00:13:50] Speaker 00:
Yes, Your Honor.

[00:13:51] Speaker ?:
Okay.

[00:13:51] Speaker 03:
All right.

[00:13:51] Speaker 03:
We'll give you two minutes for a bottle.

[00:13:53] Speaker 00:
Thank you, Your Honor.

[00:13:55] Speaker 03:
Mr. Young?

[00:14:06] Speaker 02:
Good morning, Your Honors.

[00:14:07] Speaker 02:
May it please the court, Dan Young from Vago Technologies.

[00:14:10] Speaker 02:
I'm going to address the two issues that Mr. Bass discussed with you in just a second.

[00:14:15] Speaker 02:
One thing I do want to say at the outset is that this is a case where the board specifically found that the petition

[00:14:20] Speaker 02:
does not present an identifiable, consistent, or cohesive theory of unpatentability connecting the antecedent basis for each of the limitations.

[00:14:28] Speaker 02:
In other words, that's an abuse of discretion standard.

[00:14:30] Speaker 02:
They looked at the petition, and they looked at all four of the elements, the collecting, the detecting, the correlating.

[00:14:37] Speaker 03:
Yeah, maybe if they'd stopped at that point, you'd have an argument that they did go on and construe a performance statement.

[00:14:45] Speaker 02:
Your Honor, as you asked Netflix's counsel, what is their construction?

[00:14:52] Speaker 02:
What did they propose?

[00:14:53] Speaker 02:
They didn't propose a specific construction.

[00:14:55] Speaker 02:
It was plain and ordinary meaning.

[00:14:56] Speaker 02:
And what the board said was they looked at the language of the claim.

[00:15:00] Speaker 03:
It's always baffling to me.

[00:15:01] Speaker 03:
This happens constantly at the board.

[00:15:03] Speaker 03:
People are saying plain and ordinary meaning, but that's not really what they mean.

[00:15:07] Speaker 03:
They have different versions of what plain and ordinary meaning is.

[00:15:10] Speaker 03:
This happens constantly.

[00:15:12] Speaker 03:
That seems to be this situation.

[00:15:14] Speaker 02:
Well, Your Honor, what the board

[00:15:15] Speaker 02:
What the board looked at with respect to this particular term is... Just one question.

[00:15:21] Speaker 01:
I don't think I'm throwing this discussion off course, but what do you understand was the board's construction of performance rate?

[00:15:32] Speaker 02:
I think what the board found was that the performance data, and I'll use figure nine as an example, that type of CPU data, where it's just the data for the CPU that's running various applications, that itself

[00:15:45] Speaker 02:
they didn't find was fall within performance data for one or more application instances.

[00:15:50] Speaker 02:
That we're in provision that they rely on, that associated language.

[00:15:53] Speaker 01:
But is there a place in the board's decision where you say, we can go and see the board saying, here's our claim construction?

[00:16:02] Speaker 02:
No, Your Honor, they said that the- It's implicit?

[00:16:04] Speaker 03:
Yes, Your Honor.

[00:16:06] Speaker 03:
It's implicit that it's limited to a single application instance, right?

[00:16:10] Speaker 02:
No, Your Honor.

[00:16:11] Speaker 02:
It says one or more.

[00:16:11] Speaker 02:
The claim line specifically says one or more application instances.

[00:16:15] Speaker 02:
So what do you understand their construction to be?

[00:16:18] Speaker 02:
Well, I understand with respect to what the board did was it looked at the construction in lieu of figures 9A and 9B, which is what we're talking about.

[00:16:27] Speaker 02:
So what the board said there was that there was not an explanation as to how just CPU did it.

[00:16:35] Speaker 03:
Did they have a construction or not?

[00:16:38] Speaker 02:
They did not have an implicit construction.

[00:16:40] Speaker 03:
What is the implicit construction?

[00:16:42] Speaker 02:
The implicit construction is that the CPU data that you see in figure 9b does not fall within the scope of the clean language.

[00:16:49] Speaker 03:
No, no, no.

[00:16:49] Speaker 03:
That's not sufficient.

[00:16:50] Speaker 03:
Articulate in words what you understand their implicit construction to a bit.

[00:16:57] Speaker 02:
I understand their implicit construction to be that the performance data has to be related to one or more application instances.

[00:17:06] Speaker 02:
wherein the performance data is associated with the performance of the application.

[00:17:11] Speaker 03:
Does it include host data?

[00:17:12] Speaker 03:
Does it include tier data?

[00:17:13] Speaker 02:
Potentially, Your Honor, and that's what Dr. Rosenberg... Potentially it does.

[00:17:16] Speaker 02:
I'm sorry, Your Honor?

[00:17:17] Speaker 02:
Potentially.

[00:17:17] Speaker 02:
Potentially it does.

[00:17:18] Speaker 02:
It depends on... But Dr. Rosenberg said specifically in his testimony, this is on page 74, columns, lines one through eight of his testimony, he says,

[00:17:30] Speaker 02:
that the performance data, as that term is used in Claim 1, requires the ability to determine performance of individual application instances.

[00:17:36] Speaker 02:
So for example, if you had all the instance data that were running on a host computer, and then they tally it up, but you could individually go back and figure out, well, this instance is performing this way, this one's doing this, then that would include that data, because it does associate, it is performance data for one or more application instances.

[00:17:57] Speaker 02:
The wherein provision that they rely on was added

[00:17:59] Speaker 02:
during prosecution.

[00:18:01] Speaker 02:
It's added to narrow the claim, not broaden the claim.

[00:18:04] Speaker 03:
So for example... Wait, wait, so in Masters you're saying that you can't go back and parse the data in the way you just described?

[00:18:16] Speaker 02:
I don't know, Your Honor.

[00:18:17] Speaker 02:
It's certainly not described at all in the specification.

[00:18:19] Speaker 02:
The specification mentions figure nine in a single sentence in the collecting limitation.

[00:18:25] Speaker 03:
I don't think there was any

[00:18:27] Speaker 03:
argument between the parties on this point as to whether the claims require that you be able to go back in and aggregate data and pick out the individual instances from that.

[00:18:42] Speaker 03:
I mean, it's really sounding to me as though the two of you don't disagree that much as to what performance data means.

[00:18:50] Speaker 02:
Well, certainly performance data, as it's reflected in the figures that they use as these examples, certainly doesn't show performance data that would satisfy the definition.

[00:18:58] Speaker 02:
Now, is it possible somewhere in MASTERS?

[00:19:01] Speaker 03:
Potentially, but the board's position was... Well, why is it that MASTERS doesn't show performance data that falls within the correct definition?

[00:19:09] Speaker 02:
Well, for example, if we're looking at figures 9A and 9B, what that says is that what is specifically called

[00:19:22] Speaker 02:
is specifically called a decision review display.

[00:19:25] Speaker 02:
That's what it is.

[00:19:26] Speaker 02:
So what Masters is talking about, the entirety of Masters is talking about, the RMA, the Resource Management Architecture.

[00:19:34] Speaker 02:
And what this is is an architecture that creates and manages these instances.

[00:19:39] Speaker 02:
I thought the board said that Figure 9 doesn't show performance data.

[00:19:44] Speaker 02:
It doesn't show performance data as it's related to one or more application instances.

[00:19:48] Speaker 02:
So what you see in Figure 9A.

[00:19:50] Speaker 02:
What do you understand that to mean?

[00:19:52] Speaker 02:
What they're saying is that the performance data that's shown in figures 9A and 9B, that data, what it's showing there, is data related to the RMA.

[00:20:00] Speaker 02:
In other words, at the top of the figure 9A, it shows how fast the RMA made a decision that says, hey, we need to do a scale for them, and then how long it took to implement that decision.

[00:20:13] Speaker 02:
That's data related to the RMA, not related to the individual application instances.

[00:20:20] Speaker 03:
It's not saying... Do you just agree that it could be aggregate data?

[00:20:25] Speaker 02:
Sure, Your Honor.

[00:20:26] Speaker 02:
There could be a situation where you have aggregate data.

[00:20:29] Speaker 03:
I'm not understanding.

[00:20:30] Speaker 03:
I'm sorry.

[00:20:30] Speaker 03:
I'm not understanding.

[00:20:31] Speaker 02:
Okay, so Your Honor, I'll take a step back.

[00:20:34] Speaker 02:
And what the board found in its final written decisions was that the response time data

[00:20:41] Speaker 02:
that you see on the very top of figure 9A that's showing how fast the RMA is making a decision.

[00:20:49] Speaker 02:
That's not performance data related to the application instances.

[00:20:52] Speaker 02:
That's performance data related to how fast, how the RMA is performing.

[00:20:57] Speaker 02:
So like an example, when you say response time in the shame pattern, the 722 pattern, that would be like you have an application running, how fast does it address a client request?

[00:21:08] Speaker 02:
The client comes in and says, I want you to do X.

[00:21:11] Speaker 02:
How fast is that application to perform that task?

[00:21:14] Speaker 02:
Which, what you're seeing in figure nine, A and B, and what the board found, is that's something different.

[00:21:19] Speaker 02:
That's how, that's the overall structure deciding I need to do an event, and then how fast it will take me to implement that event.

[00:21:31] Speaker 02:
So, those are, and the board found, those are kind of two different buckets of data, and the petition doesn't provide any explanation

[00:21:40] Speaker 02:
as to how that data that you see in figures 9a and 9b.

[00:21:44] Speaker 02:
relates to the claim line.

[00:21:45] Speaker 02:
That's what their findings said.

[00:21:48] Speaker 01:
What you've been saying in response to your Judge Deitz question seems to, again, go back to claim construction, the construction of performance data.

[00:21:59] Speaker 01:
Now, correct me if I'm wrong, but it seems that both you and Mr. Batts say that there wasn't an express claim construction here, but there was an implicit claim construction.

[00:22:10] Speaker 01:
Both seem to say that.

[00:22:12] Speaker 01:
Assuming we look at that implicit construction and disagree with it, do we construe the claim ourselves or do we send it back for the board to take a shot at it with some guidance?

[00:22:29] Speaker 02:
If you disagree, the board would have to make that decision.

[00:22:31] Speaker 03:
The board did not come to a conclusion that says... Why does the board have to make the decision?

[00:22:35] Speaker 03:
Claim construction is de novo.

[00:22:36] Speaker 03:
We can construe the claim.

[00:22:38] Speaker 02:
Well, the claim, Your Honor, what the board did was it evaluated the... What's the answer to that?

[00:22:44] Speaker 03:
We can construe the claim, right?

[00:22:46] Speaker 03:
Yes, Your Honor.

[00:22:47] Speaker 02:
But what the board's decision was, was related to the adequacy of the petition.

[00:22:51] Speaker 01:
the petition didn't set forth as to what because of the position position adequacy that that's right you read the board decision it says that but it seems that underlying the boards view on that as to the adequacy of the petition was its underlying view with respect to the claim construction you know so don't we have to i mean if we disagree with that implicit claim construction

[00:23:18] Speaker 01:
we could send it back or we could decide.

[00:23:20] Speaker 01:
But doesn't it make sense, as perhaps Judge Dyke was suggesting, that we should articulate a claim construction?

[00:23:28] Speaker 02:
Yes, Your Honor.

[00:23:28] Speaker 02:
What we've expressly found was that the petition doesn't provide any explanation as to how the data that's reflected in figures 9A and 9B satisfies the all four elements of the claim.

[00:23:42] Speaker 02:
That's what they said.

[00:23:43] Speaker 02:
But that was based on a view of performance data, correct?

[00:23:47] Speaker 02:
it was based on performance data and what was described in the petition.

[00:23:50] Speaker 01:
And the petition doesn't... It was based on the... Excuse me, but it was based, I think, and correct me if I'm wrong, but it was based, as you said, on what was described in the petition and as to the board's view, apparently, of performance data.

[00:24:07] Speaker 01:
Is that correct?

[00:24:09] Speaker 02:
It was absolutely how the board viewed the limitation, the entire limitation, which says performance data for one or more application instances.

[00:24:17] Speaker 02:
And they're like, well, this is just CPU data in a vacuum.

[00:24:21] Speaker 01:
So the view's decision hinges on a claim construction, right?

[00:24:29] Speaker 01:
Yes, Your Honor.

[00:24:29] Speaker 01:
OK.

[00:24:30] Speaker 01:
That's what I'm saying.

[00:24:31] Speaker 01:
If we agree with that construction, fine.

[00:24:34] Speaker 01:
But if we disagree, is there anything to prevent us from saying, OK, here's what we believe is the correct claim construction and sending it back?

[00:24:46] Speaker 02:
Your Honor, I think there's two things, and this may be the source of a little bit of confusion.

[00:24:51] Speaker 02:
The two issues are adequacy of the petition, and then, as you say, client construction.

[00:24:55] Speaker 02:
So the board focused primarily on the adequacy of the petition.

[00:24:59] Speaker 02:
But I understand what you're saying, and there is an underlying construction there.

[00:25:02] Speaker 02:
And what the board said was, again, with figures 9A and 9B, they said that was data related to the RMA.

[00:25:10] Speaker 02:
No, I understand.

[00:25:11] Speaker 03:
Where does this say that?

[00:25:14] Speaker 03:
16 and 17, where is the board set?

[00:25:24] Speaker 02:
So your honor, this would be on, yes, appendix page 17, where it says further.

[00:25:30] Speaker 02:
And again, this is related to what the Netflix cited to certain portions of the specification to support it.

[00:25:37] Speaker 02:
And what the board said further, the cited portions of the 722 patent do not support petitioners' implicit claim construction.

[00:25:44] Speaker 02:
Both the cited portions of the 722 patent refer to application instances response time in servicing a client request, not the system's response time in creating an application instance.

[00:25:54] Speaker 03:
And that seems to be saying that aggregate data is not performance data.

[00:26:01] Speaker 02:
What that's saying is, Your Honor, that the... Am I wrong?

[00:26:05] Speaker 03:
Isn't that saying that aggregate data is not performance data?

[00:26:11] Speaker 02:
Your Honor, you could... I mean, what it expressly says is it says... Answer my question.

[00:26:16] Speaker 03:
Yes, no, maybe... Potentially, Your Honor.

[00:26:18] Speaker 02:
I think potentially would be the answer.

[00:26:20] Speaker 02:
Because what it's talking about is servicing a client request.

[00:26:23] Speaker 02:
That relates to the RMA.

[00:26:25] Speaker 02:
That's what I was trying to say.

[00:26:26] Speaker 02:
Servicing a client request... I apologize.

[00:26:32] Speaker 02:
I got that backwards.

[00:26:33] Speaker 02:
Servicing a client request is what the 722 PAC was talking about.

[00:26:36] Speaker 02:
And it says not the system's response time.

[00:26:38] Speaker 02:
The system's response time is the RMA.

[00:26:40] Speaker 02:
That's what it's talking about.

[00:26:41] Speaker 01:
It does seem, and correct me if I'm wrong, Mr. Young, that your position stands or falls on whether this implicit claim construction was correct.

[00:26:57] Speaker 01:
If it was, you're home free.

[00:26:59] Speaker 01:
If it wasn't, then you have a problem, perhaps.

[00:27:02] Speaker 02:
Well, Your Honor, certainly with respect to the collecting limitation, that's true.

[00:27:06] Speaker 02:
The board didn't make any findings with respect to figures 9A and 9B, with respect to the correlating

[00:27:11] Speaker 03:
I understand that, but let's turn to the correlating limitation for Figure 8, because that's an alternative ground that they articulated with respect to Figure 8, and that seems to ignore Exhibit 9 of Masters.

[00:27:27] Speaker 03:
And that's a problem, right?

[00:27:29] Speaker 02:
I'm sorry, Your Honor, Exhibit 9 of Mastery?

[00:27:31] Speaker 02:
I mean, Exhibit J of Mastery.

[00:27:33] Speaker 02:
Yes, Your Honor, the issue with Exhibit J, and it doesn't add, Exhibit J doesn't add anything to what you see in Figures 8a and 8b.

[00:27:43] Speaker 02:
What that is showing is that, and the board found this, talking about it on page 20 of their opinion,

[00:27:49] Speaker 02:
What the RMA does is it looks at it as a threshold.

[00:27:53] Speaker 02:
That's that horizontal line that you see going across figures 8A and 8B, as well as what you see the horizontal line in the Appendix J figures.

[00:28:00] Speaker 02:
They're really the same situation.

[00:28:02] Speaker 02:
What it does is it looks at a threshold.

[00:28:05] Speaker 02:
It looks at the application instances and how it compares to that threshold.

[00:28:08] Speaker 02:
That's how it makes the decision about whether or not to do a certain action.

[00:28:11] Speaker 02:
That's what the board found.

[00:28:13] Speaker 02:
Appendix J was added so that Appendix J is not in the petition.

[00:28:18] Speaker 02:
the defense he was not relied on by their expert in his declaration of this it's part it's appendix j is intrinsic evidence is it not yes your honor but the with the board look at this decision is the adequacy of the petition appendix j was provided a day before we were going to pose their expert on his never said what it looked at it that's true your honor they did not the the appendix j

[00:28:43] Speaker 02:
Still, at the end of the day, the board's opinions with respect to exhibit 8A and 8B remain the same.

[00:28:52] Speaker 02:
What it said was that all you're showing is continuously looking at an application instance, in this case a couple of them.

[00:28:58] Speaker 02:
And then it says, then they say, oh, well, an application was added in section 13, 34.

[00:29:06] Speaker 02:
but all you're seeing is those lines.

[00:29:09] Speaker 02:
There's not a correlation between, so what the patent will say is that if you add an instance as an example, that the other instances drop or they rise in their performance.

[00:29:21] Speaker 02:
When we ask their expert about this, you'll see those lines going up and down without even a new

[00:29:28] Speaker 02:
a new instance being added.

[00:29:30] Speaker 02:
You're not isolating the change with respect to that particular instance.

[00:29:35] Speaker 02:
All you're doing is tracking application instances.

[00:29:38] Speaker 02:
That's been in the prior art for a long time.

[00:29:40] Speaker 02:
That didn't add anything to the

[00:29:42] Speaker 02:
to what was already in the prior art.

[00:29:45] Speaker 02:
All you're doing is tracking applications.

[00:29:46] Speaker 02:
If you were then to isolate and say, okay, when we add this, this is how it impacts these other instances, well, then they may have an argument.

[00:29:53] Speaker 02:
But that's not what's reflected in exhibit 8A and 8B, and Appendix J doesn't add anything to that disclosure.

[00:29:58] Speaker 03:
Well, okay, but shouldn't we hear from the board as to what it thinks about Appendix J?

[00:30:04] Speaker 03:
I just don't understand why they didn't mention it.

[00:30:07] Speaker 02:
Well, because, Your Honor, I think what the Board did was they focused on the adequacy of the petition.

[00:30:11] Speaker 02:
The entire decision that you read is all about the adequacy of the petition.

[00:30:15] Speaker 02:
And the appendix J was not mentioned in the petition.

[00:30:18] Speaker 02:
And they did not file a motion.

[00:30:19] Speaker 02:
They didn't file a motion to supplement.

[00:30:21] Speaker 03:
I'm sorry, Your Honor.

[00:30:22] Speaker 03:
They held repeatedly that arguments develop in the course of the IPR.

[00:30:29] Speaker 03:
And that's something the Board has to accommodate itself in.

[00:30:32] Speaker 02:
Well, the Board noted in the oral argument that they did not file a motion to supplement the information.

[00:30:37] Speaker 02:
from the exhibit J. And when they provided it, they didn't provide any explanations to how they were going to even use it.

[00:30:42] Speaker 02:
They just said, here's 100 and some odd pages of.

[00:30:45] Speaker 01:
But if it's intrinsic evidence and part of the patent, isn't it necessarily there?

[00:30:51] Speaker 01:
And can't anyone look at it, particularly in response to the ebb and flow of arguments before the board?

[00:30:58] Speaker 02:
I think the proper procedure, Your Honor, would be to get the application J, file a motion to supplement authority, and say why you're going to have it.

[00:31:05] Speaker 02:
Did they mention it in the reply?

[00:31:09] Speaker 02:
And they did mention it in the reply.

[00:31:11] Speaker 02:
And you had a sir reply.

[00:31:12] Speaker 02:
We did, Your Honor.

[00:31:13] Speaker 02:
And the board's decision on figures 8A and 8B, the exact fines that they made would be applicable to 8A and 8B as well as Appendix J. All right.

[00:31:27] Speaker 03:
I think we're out of time.

[00:31:28] Speaker 03:
Thank you.

[00:31:28] Speaker 03:
Thank you, Your Honor.

[00:31:30] Speaker 03:
Mr. Batch, you have two minutes.

[00:31:36] Speaker 00:
Thank you, Your Honor.

[00:31:37] Speaker 00:
I just have a few points here.

[00:31:38] Speaker 00:
I think you've heard now the parties agree that there was an implicit construction that was performed below with respect to performance data.

[00:31:46] Speaker 00:
I think when you ask what would our construction be, I think the one that I have is that, and it sounds like we're very close to them, any data that is related to the performance of one or more application instances.

[00:31:57] Speaker 00:
He used the term related to.

[00:31:59] Speaker 00:
I think that's similar to associated with.

[00:32:01] Speaker 00:
But if you want to do a specific construction, but clearly.

[00:32:04] Speaker 03:
Do you think the board's implicit construction

[00:32:06] Speaker 00:
Well, the board specifically said that they were denying rejecting our arguments with respect to based upon performance data.

[00:32:13] Speaker 00:
So they specifically said on page 11 of their page 12 of the final written decision, they said, well, it's a broad term.

[00:32:23] Speaker 00:
Petitioner does not adequately establish that.

[00:32:25] Speaker 00:
And then it goes on to say that this language, the claim language, includes such a broad variety of data as application-level data, tier-level data, and host-level data.

[00:32:34] Speaker 00:
So they're clearly rejecting that the term included any of those sets of data.

[00:32:39] Speaker 00:
And then on page 17 of the final written decision, they said, quote, further, the cited portions of the 722 patent do not support petitioners' implicit claim interpretation.

[00:32:50] Speaker 00:
So they clearly made a determination based upon claim construction here.

[00:32:55] Speaker 00:
The next issue I want to point out is that there were several new arguments that were made by Avago's counsel here that are not made below.

[00:33:03] Speaker 00:
But if we go to the Masters Appendix J, he said Appendix J doesn't add anything to Masters Figure 8.

[00:33:15] Speaker 00:
And I have to strenuously disagree with that, Your Honor.

[00:33:18] Speaker 00:
Just the pages that I pointed out to you make clear that there are a lot of additional teachings in Appendix J that are direct rebuttals to their arguments below.

[00:33:28] Speaker 00:
And for him to conjecture here about what they could decide or what they could have thought or what they might have thought doesn't go to the question of we have nothing by the board, not a single reference to Appendix J.

[00:33:40] Speaker 00:
in the final written decision to even appeal to you.

[00:33:42] Speaker 00:
So he mentioned supplemental evidence or... I think we're out of time.

[00:33:46] Speaker 03:
Thank you, Mr. Batts.

[00:33:47] Speaker 00:
Any further questions, sir?

[00:33:48] Speaker 03:
No.

[00:33:49] Speaker 03:
Thank you, Mr. Yann.

[00:33:50] Speaker 03:
Thank you.