`
`·2· · · · · BEFORE THE PATENT TRIAL AND APPEAL BOARD
`
`·3· · · · · · · · · · · · · ---oOo---
`
`·4· ·INTEL CORP. and CAVIUM INC.,
`
`·5· · · · · · · Petitioners,
`
`·6· · · · · · · · · vs.
`
`·7· ·ALACRITECH, INC.,
`
`·8· · · · · · · Patent owner.
`· · ·____________________________/
`·9
`
`10
`
`11· · · ·Case IPR2017-013931; U.S. Patent No. 9,055,104
`· · · · ·Case IPR2017-013911; U.S. Patent No. 7,237,036
`12· · · ·Case IPR2017-014091; U.S. Patent No. 8,131,880
`· · · · ·Case IPR2017-014101; U.S. Patent No. 8,131,880
`13· · · ·Case IPR2017-014051; U.S. Patent No. 7,124,205
`· · · · ·Case IPR2017-014061; U.S. Patent No. 7,673,072
`14· · · ·Case IPR2017-013921; U.S. Patent No. 7,337,241
`
`15
`
`16· · · · · · · · · · · ·DEPOSITION OF
`
`17· · · · · · · · · · KEVIN ALMEROTH, PH.D.
`
`18· · · · · · · _________________________________
`
`19· · · · · · · · · VOLUME 2; PAGES 256 - 478
`
`20· · · · · · · · · · ·FRIDAY, MAY 4, 2018
`
`21
`
`22
`
`23· ·REPORTED BY:· HOLLY THUMAN, CSR No. 6834, RMR, CRR
`
`24· · · · · · · · · · · · · · · · · · · · · ·(SF-173597)
`
`25
`
`INTEL EX. 1225.001
`
`
`
`·1· · · · · · · · · · · · · I N D E X
`
`·2· · · · · · · · · · INDEX OF EXAMINATIONS
`
`·3· ·EXAMINATION BY:· · · · · · · · · · · · · · · · · · PAGE
`
`·4· ·MR. STEPHENS· · · · · · · · · · · · · · · · · · · · 260
`
`·5· · · · · · · · · · · · · ·--o0o--
`
`·6· · · · · · ·EXHIBITS MARKED FOR IDENTIFICATION
`
`·7· ·NO.· · · · · · · · ·DESCRIPTION· · · · · · · · · · ·PAGE
`
`·8· ·Exhibit 10· · ·United States Patent 7,337,241,· · · ·260
`· · · · · · · · · · Boucher et al.
`·9
`· · ·Exhibit 11· · ·Document headed "Chapter 10:· · · · · 292
`10· · · · · · · · · Interrupt Handling"
`
`11· ·Exhibit 12· · ·Corrected Patent Owner's Exhibit· · · 388
`· · · · · · · · · · 2026, Declaration of Kevin
`12· · · · · · · · · Almeroth, Ph.D., IPR2017-1409,
`· · · · · · · · · · U.S. Patent 8,131,880
`13
`· · ·Exhibit 13· · ·Corrected Patent Owner's Exhibit· · · 388
`14· · · · · · · · · 2026, Declaration of Kevin
`· · · · · · · · · · Almeroth, Ph.D.,
`15· · · · · · · · · IPR2017-1410U.S. Patent
`· · · · · · · · · · 8,131,880
`16
`· · ·Exhibit 14· · ·United States Patent 8,131,880,· · · ·390
`17· · · · · · · · · Boucher et al.
`
`18· ·Exhibit 15· · ·Excerpt from Appendix A to the· · · · 401
`· · · · · · · · · · Report of Alacritech's expert
`19· · · · · · · · · Dr. Kevin C. Almeroth Concerning
`· · · · · · · · · · Intel's Infringement,
`20· · · · · · · · · Alacritech, Inc. vs. CenturyLink
`· · · · · · · · · · Communications, LLC, et al. and
`21· · · · · · · · · related actions
`
`22· ·Exhibit 16· · ·Declaration of Dr. Kevin· · · · · · · 431
`· · · · · · · · · · Almeroth in Support of
`23· · · · · · · · · Alacritech's Motion for
`· · · · · · · · · · Preliminary Injunction of
`24· · · · · · · · · Microsoft's Infringement of
`· · · · · · · · · · Claim 1 of U.S. Patent 6,697,868
`25· ·(Cont'd)
`
`INTEL EX. 1225.002
`
`
`
`·1· ·(Exhibits, cont'd)
`
`·2· ·Exhibit 17· · ·Declaration of Dr. Kevin· · · · · · · 436
`· · · · · · · · · · Almeroth in Support of
`·3· · · · · · · · · Alacritech's Reply to
`· · · · · · · · · · Microsoft's Opposition to
`·4· · · · · · · · · Alacritech's Motion for
`· · · · · · · · · · Preliminary Injunction
`·5
`· · ·Exhibit 18· · ·Document headed "14: A Reduced· · · · 452
`·6· · · · · · · · · Operation Protocol Engine (ROPE)
`· · · · · · · · · · for a multiple-layer bypass
`·7· · · · · · · · · architecture"
`
`·8· ·Exhibit 19· · ·Patent Owner's Exhibit 2026,· · · · · 460
`· · · · · · · · · · Declaration of Kevin Almeroth,
`·9· · · · · · · · · Ph.D., Case IPR2017-01405, U.S.
`· · · · · · · · · · Patent 7,124,205
`10
`· · ·Exhibit 20· · ·Curriculum Vitae, Kevin C.· · · · · · 470
`11· · · · · · · · · Almeroth
`
`12· ·Exhibit 21· · ·CS 176 -- Introduction to· · · · · · ·474
`· · · · · · · · · · Computer Communication Networks,
`13· · · · · · · · · Fall 1997
`
`14
`
`15
`
`16
`
`17
`
`18
`
`19
`
`20
`
`21
`
`22
`
`23
`
`24
`
`25
`
`INTEL EX. 1225.003
`
`
`
`·1· · · · · · · · · · · · · ·--o0o--
`
`·2· · · · Deposition of KEVIN ALMEROTH, PH.D., taken by
`
`·3· ·Petitioner Intel Corp., at QUINN EMANUEL, 50 California
`
`·4· ·Street, 22nd Floor, San Francisco, California 94111,
`
`·5· ·commencing at 9:04 A.M., on FRIDAY, MAY 4, 2018, before
`
`·6· ·me, HOLLY THUMAN, CSR, RMR, CRR.
`
`·7· · · · · · · · · · · · · ·--o0o--
`
`·8· · · · · · · · · · · · ·APPEARANCES
`
`·9· ·FOR PETITIONER INTEL CORP.:
`
`10· · · · WEIL, GOTSHAL & MANGES LLP
`· · · · · 700 Louisiana, Suite 1700
`11· · · · Houston, Texas 77002-2755
`· · · · · By:· GARLAND STEPHENS, Attorney at Law
`12· · · · · · ·garland.stephens@weil.com
`
`13· ·FOR PETITIONER CAVIUM,INC.:
`
`14· · · · DUANE MORRIS
`· · · · · 2475 Hanover Street
`15· · · · Palo Alto, California 94304-1194
`· · · · · By:· NIKOLAUS A. WOLOSZCZUK, Attorney at Law
`16· · · · · · ·nawoloszczuk@duanemorris.com
`
`17· ·FOR THE PATENT OWNER:
`
`18· · · · QUINN EMANUEL
`· · · · · 50 California Street, 22nd Floor
`19· · · · San Francisco, California 94111
`· · · · · By:· BRIAN E. MACK, Attorney at Law
`20· · · · · · ·brianmack@quinnemanuel.com
`
`21
`
`22
`
`23
`
`24
`
`25
`
`INTEL EX. 1225.004
`
`
`
`·1· · · ·SAN FRANCISCO, CALIFORNIA; FRIDAY, MAY 4, 2018
`
`·2· · · · · · · · · · · · · 9:04 A.M.
`
`·3· · · · · · · · · · · · · ·--o0o--
`
`·4· · · · · · · · · ·KEVIN ALMEROTH, PH.D.,
`
`·5· · · · · · · _________________________________
`
`·6· ·called as a witness, who, having been first duly sworn,
`
`·7· ·was examined and testified as follows:
`
`·8· · · · · · · · · · · · · ---oOo---
`
`·9· · · · · · · · ·EXAMINATION BY MR. STEPHENS
`
`10· · · · · · (Deposition Exhibit 10 was marked for
`
`11· · · · · · identification.)
`
`12· · · · · · MR. STEPHENS:· We're back on the record.
`
`13· · · · Q.· Dr. Almeroth, you understand you're still
`
`14· ·under oath today?
`
`15· · · · A.· Yes, sir.
`
`16· · · · Q.· Thank you.· The court reporter has handed you
`
`17· ·what's been marked as Almeroth Exhibit 10, which is the
`
`18· ·same as Intel Exhibit 1001, and that is the '241
`
`19· ·patent.
`
`20· · · · · · Do you recognize that?
`
`21· · · · A.· I do.
`
`22· · · · Q.· Yesterday we talked about the discussion, the
`
`23· ·background discussion -- excuse me -- the "Background
`
`24· ·of the Invention" discussion which is on Column 35 in
`
`25· ·the '241, which I believe is essentially the same as
`
`INTEL EX. 1225.005
`
`
`
`·1· ·the similar discussion in the '072 patent.· I wanted to
`
`·2· ·come back to that a little bit.
`
`·3· · · · · · I want to walk through the discussion of "Too
`
`·4· ·Many Interrupts" here to make sure that I understand
`
`·5· ·what your view of it is.· It says -- the first sentence
`
`·6· ·says, "A 64K SMB request (write or read-reply)."
`
`·7· · · · · · What does that mean?
`
`·8· · · · A.· I'm not sure I understand the question.
`
`·9· · · · Q.· Well, a 64K SMB request refers to a system
`
`10· ·message block message sent over TCP/IP that's 64K long.
`
`11· ·Right?
`
`12· · · · A.· It does.
`
`13· · · · Q.· And a write would be sent by a host system
`
`14· ·including 64K bytes to a network storage device to
`
`15· ·store that 64K.· Right?
`
`16· · · · A.· From a network perspective, just to a remote
`
`17· ·host.· But --
`
`18· · · · Q.· Okay.
`
`19· · · · A.· But sure.
`
`20· · · · Q.· And a 64K SMB read-reply would be a response
`
`21· ·to read storage, and that would be received by the
`
`22· ·device.· Right?
`
`23· · · · A.· It would be.
`
`24· · · · Q.· So it's talking about both 64K transmissions
`
`25· ·and 64K receptions in that phrase.· Right?
`
`INTEL EX. 1225.006
`
`
`
`·1· · · · A.· It is.
`
`·2· · · · Q.· And then it says that that would typically be
`
`·3· ·made up of 44 TCP segments when running over Ethernet
`
`·4· ·with a 1500-byte MTU.
`
`·5· · · · · · And I think you agreed yesterday that seems
`
`·6· ·roughly right.
`
`·7· · · · A.· Yes.· And -- well, so the point to make is,
`
`·8· ·the request could be one of two things.· So it's only
`
`·9· ·talking about one 64K SMB.· It's not talking about the
`
`10· ·two together.
`
`11· · · · Q.· Right.· It's saying one or the other.· Right?
`
`12· · · · A.· That's correct.
`
`13· · · · Q.· But is it your understanding that the rest of
`
`14· ·the paragraph applies to either one?
`
`15· · · · A.· It may.· I mean, we'll certainly walk through
`
`16· ·it.
`
`17· · · · Q.· Okay.· Let's do that.
`
`18· · · · · · Now, it says each of the segments may result
`
`19· ·in an interrupt to the CPU, and it's referring to 44
`
`20· ·TCP segments there.· Right?
`
`21· · · · A.· It is.
`
`22· · · · Q.· And there's no suggestion, in that sentence at
`
`23· ·least, that any segment might result in more than one
`
`24· ·interrupt to the CPU.· Right?
`
`25· · · · · · MR. MACK:· Objection.· Form.
`
`INTEL EX. 1225.007
`
`
`
`·1· · · · · · THE WITNESS:· I certainly don't think that
`
`·2· ·sentence should be interpreted as saying there would
`
`·3· ·only be at most one interrupt.· If anything, I think
`
`·4· ·it's saying that there's at least one interrupt.· It's
`
`·5· ·focusing on at least one interrupt.· I think the
`
`·6· ·reality is there certainly would be more than that.
`
`·7· ·BY MR. STEPHENS:
`
`·8· · · · Q.· Well, in fact, it says almost the opposite of
`
`·9· ·what you just said.· Right?· It says each of these
`
`10· ·segments may result in an interrupt to the CPU.· It
`
`11· ·doesn't say that it will result in at least one.
`
`12· · · · A.· I would disagree with that characterization.
`
`13· ·I see where it says may result in an interrupt, but I
`
`14· ·think, reasonably speaking, within a protocol stack you
`
`15· ·would certainly expect it to have multiple interrupts,
`
`16· ·for lots of the reasons we talked about yesterday.
`
`17· · · · Q.· Okay.· I understand that's your view.
`
`18· · · · · · Now, let's back up a second.· This is part of
`
`19· ·the "Background of the Invention" section.· Right?
`
`20· · · · A.· It does list this as the "Background of the
`
`21· ·Invention" section.
`
`22· · · · Q.· And it says "Network processing as it exists
`
`23· ·today is a costly and inefficient use of system
`
`24· ·resources."· Right?
`
`25· · · · A.· Yes.· It looks like you're reading at
`
`INTEL EX. 1225.008
`
`
`
`·1· ·Column 35, about line 13.
`
`·2· · · · Q.· Yeah, I'm sorry.· I should have pointed that
`
`·3· ·out.
`
`·4· · · · · · And then the last sentence of that
`
`·5· ·paragraph says, "The reasons that this processing is so
`
`·6· ·costly are described here."· Right?
`
`·7· · · · A.· Yes.
`
`·8· · · · Q.· And you've testified, and you've said in your
`
`·9· ·declarations, that the inventions in the patents at
`
`10· ·issue here solve these problems.· Right?
`
`11· · · · · · MR. MACK:· Objection to form.
`
`12· · · · · · THE WITNESS:· I think in a broad context, some
`
`13· ·of the claims do solve some of the problems.· I don't
`
`14· ·know that every claim solves every problem that's
`
`15· ·listed here.
`
`16· ·BY MR. STEPHENS:
`
`17· · · · Q.· Okay.· There's nothing to suggest here in the
`
`18· ·"Background of the Invention" section that Alacritech
`
`19· ·failed to account for a huge quantity of interrupts, is
`
`20· ·there?
`
`21· · · · · · MR. MACK:· Objection.· Form.
`
`22· · · · · · THE WITNESS:· I don't understand the question.
`
`23· ·When you say "failed to account for," do you mean in
`
`24· ·writing this section, or --
`
`25
`
`INTEL EX. 1225.009
`
`
`
`·1· ·BY MR. STEPHENS:
`
`·2· · · · Q.· Yes.
`
`·3· · · · A.· -- in their design, or --
`
`·4· · · · Q.· I mean in writing this section.
`
`·5· · · · · · They don't say, "By the way, there's many more
`
`·6· ·interrupts besides the ones we list here as being too
`
`·7· ·many."
`
`·8· · · · · · MR. MACK:· Objection.· Form.
`
`·9· · · · · · THE WITNESS:· Well, I mean, I think this goes
`
`10· ·to an understanding of writing a specification so it
`
`11· ·could be understood by a person of skill in the art.
`
`12· ·And I don't think that it would be necessary to
`
`13· ·identify all of the possible sources of interrupts that
`
`14· ·can be generated during or caused by a network
`
`15· ·transmission.
`
`16· · · · · · I think that, as I testified to yesterday,
`
`17· ·it's laying out a high-level perspective of what the
`
`18· ·problems are, and then, in some cases, giving specific
`
`19· ·examples.· Which is why I think care should be taken in
`
`20· ·trying to read some of the sentences you're reading
`
`21· ·individually and suggest that all Alacritech was saying
`
`22· ·in writing the sentence at Column 36, line 27, is that
`
`23· ·there would only sometimes be at most one interrupt for
`
`24· ·each sender-received TCP segment.
`
`25· · · · Q.· Okay.· Well, let's go back to that.· The next
`
`INTEL EX. 1225.010
`
`
`
`·1· ·sentence says:
`
`·2· · · · · · "Furthermore, since TCP must acknowledge
`
`·3· · · · all of this incoming data, it's possible to
`
`·4· · · · get another 44 transmit-complete interrupts as
`
`·5· · · · a result of sending out the TCP
`
`·6· · · · acknowledgments."
`
`·7· · · · · · Do you see that?
`
`·8· · · · A.· I do.
`
`·9· · · · Q.· So what that's referring to is receiving 44
`
`10· ·TCP segments and acknowledging each one of them
`
`11· ·individually.· Right?
`
`12· · · · A.· No.· There's both the receiving of the data
`
`13· ·and the generation of acknowledgments, which may be
`
`14· ·piggybacked on data.· But also in handling the
`
`15· ·acknowledgments at the place where the data was either
`
`16· ·written or requested.
`
`17· · · · · · So if I send out 44 TCP segments, I will get
`
`18· ·acknowledgments for those 44 TCP segments back.· Those
`
`19· ·44 acknowledgments have to individually be processed
`
`20· ·just like a TCP segment, but they wouldn't have any
`
`21· ·data necessarily.
`
`22· · · · · · So, for example, if I'm writing a 64K SMB
`
`23· ·request, I send out 44 TCP segments.· That generates a
`
`24· ·wave of interrupts.· There's a wave of interrupts at
`
`25· ·the receiver.· The receiver generates acknowledgments
`
`INTEL EX. 1225.011
`
`
`
`·1· ·that come back, and then processing those
`
`·2· ·acknowledgments generates another wave of interrupts.
`
`·3· · · · Q.· But that's not what it's talking about here.
`
`·4· ·Right?· Because it's talking about transmit-complete
`
`·5· ·interrupts.
`
`·6· · · · A.· Yeah.· That's absolutely what it's talking
`
`·7· ·about.
`
`·8· · · · Q.· So it's talking about sending acknowledgments
`
`·9· ·to 44 received TCP segments.· Right?· It's not talking
`
`10· ·about receiving acknowledgments to 44 TCP segments,
`
`11· ·because it's talking about another 44 transmit-complete
`
`12· ·interrupts.
`
`13· · · · A.· Yeah --
`
`14· · · · · · MR. MACK:· Objection.· Form.
`
`15· · · · · · THE WITNESS:· The transmit complete is an --
`
`16· ·is the reception of an acknowledgment.· So it's a
`
`17· ·reception of the acknowledgment at the sender saying
`
`18· ·that the transmit was complete.· And you get one of
`
`19· ·those for each TCP segment, unless you're doing
`
`20· ·something like ACK delays.
`
`21· ·BY MR. STEPHENS:
`
`22· · · · Q.· So it's your testimony under oath here today
`
`23· ·that a transmit-complete interrupt is generated to a
`
`24· ·host processor when it receives an acknowledgment?
`
`25· · · · A.· Ask that again?
`
`INTEL EX. 1225.012
`
`
`
`·1· · · · Q.· It's your testimony that a host computer
`
`·2· ·receives a transmit-complete interrupt when it receives
`
`·3· ·an acknowledgment?
`
`·4· · · · A.· No, no.· You're using the wrong terms.· It --
`
`·5· ·the sender sends out 44 segments.
`
`·6· · · · Q.· Okay.
`
`·7· · · · A.· The receiver gets those segments.· There's
`
`·8· ·interrupts associated with sending the 44 segments.
`
`·9· · · · Q.· Those are on one side.· Right?
`
`10· · · · A.· Right.
`
`11· · · · Q.· Okay.
`
`12· · · · A.· At the receiving side, there's a whole set of
`
`13· ·interrupts with receiving those 44 segments.
`
`14· · · · · · Now, possibly as part of that processing, it
`
`15· ·may generate additional interrupts.· But as part of
`
`16· ·that processing at the receiver, you will send
`
`17· ·acknowledgments back to the transmitter telling the
`
`18· ·transmitter that the reception has been completed at
`
`19· ·the receiver.
`
`20· · · · · · After the host sends the 44 TCP segments, it
`
`21· ·sits there and waits.· It has timers running.· There's
`
`22· ·a lot of processing involved, which is part of what's
`
`23· ·being described here.· And then the acknowledgments
`
`24· ·will start to trickle in.
`
`25· · · · · · The acknowledgments are TCP packets.· They may
`
`INTEL EX. 1225.013
`
`
`
`·1· ·or may not have data.· But now the sender is a receiver
`
`·2· ·of those acknowledgments, and each one of those have to
`
`·3· ·be processed.· It could potentially have data, but you
`
`·4· ·still have to do full processing on the Ethernet, the
`
`·5· ·IP, and the TCP header, and that will generate
`
`·6· ·interrupts as well.
`
`·7· · · · · · So there's really four sets of interrupts that
`
`·8· ·will be created by sending one TCP segment: at the
`
`·9· ·sending side, at the receiving side, generation of
`
`10· ·acknowledgment at the receiver, and then receiving of
`
`11· ·the acknowledgment at the sender.
`
`12· · · · Q.· Okay.· But this isn't talking about -- this
`
`13· ·isn't trying to count interrupts on both sides.· Right?
`
`14· ·It's trying to count interrupts on one side.
`
`15· · · · A.· So --
`
`16· · · · · · MR. MACK:· Objection.· Form.
`
`17· · · · · · THE WITNESS:· Even focusing on one side, the
`
`18· ·sender sends out the data that generates a set of
`
`19· ·interrupts, and then receiving the acknowledgments back
`
`20· ·will require a processing of those interrupts and will
`
`21· ·generate -- or processing of those packets and will
`
`22· ·generate interrupts as well.
`
`23· ·BY MR. STEPHENS:
`
`24· · · · Q.· But that's not what's being discussed in this
`
`25· ·sentence.· Let me just read it so we make sure we're
`
`INTEL EX. 1225.014
`
`
`
`·1· ·talking about the same sentence:
`
`·2· · · · · · "Furthermore, since TCP must acknowledge all
`
`·3· ·of this incoming data" -- so it's receiving the 44 TCP
`
`·4· ·segments.· Right?
`
`·5· · · · A.· Uh-huh.
`
`·6· · · · Q.· -- "it's possible to get another 44
`
`·7· ·transmit-complete interrupts as a result of sending out
`
`·8· ·the TCP acknowledgments."
`
`·9· · · · A.· Yeah.
`
`10· · · · Q.· So this -- these 44 interrupts that are being
`
`11· ·discussed here are on the side of the write or
`
`12· ·read-reply that's receiving 44 TCP segmenting and
`
`13· ·sending out as many as 44 acknowledgments.· Right?
`
`14· · · · · · MR. MACK:· Objection to form.
`
`15· · · · · · THE WITNESS:· I disagree.
`
`16· ·BY MR. STEPHENS:
`
`17· · · · Q.· Okay.· Well, what is it, then?
`
`18· · · · A.· Because it says as a result of sending out the
`
`19· ·acknowledgments.· So the result of sending out the
`
`20· ·acknowledgments is the reception of those
`
`21· ·acknowledgments.
`
`22· · · · · · Now, frankly, whether or not you want to parse
`
`23· ·it your way or not, my description of where you have
`
`24· ·interrupts on sending, interrupts on receiving,
`
`25· ·interrupts on generating acknowledgments, and
`
`INTEL EX. 1225.015
`
`
`
`·1· ·interrupts on receiving the acknowledgments, is still
`
`·2· ·accurate.· I mean, it still happens.· And so whether or
`
`·3· ·not the statement is focused on that or not is really
`
`·4· ·irrelevant.
`
`·5· · · · Q.· Well, that's for the PTAB to decide.· And I'm
`
`·6· ·asking about this sentence, not about things that are
`
`·7· ·not discussed in this sentence.
`
`·8· · · · · · So when it says that TCP must acknowledge all
`
`·9· ·of this incoming data, it's talking about the side
`
`10· ·that is receiving 44 TCP segments.· Do you agree with
`
`11· ·that?
`
`12· · · · A.· Well, if it used that language.· It's
`
`13· ·different than what's here, but -- I mean --
`
`14· · · · Q.· TCP must acknowledge all of this incoming
`
`15· ·data.· Right?· It's not talking about outgoing data;
`
`16· ·it's talking about incoming data.
`
`17· · · · A.· It is.
`
`18· · · · Q.· Okay.· So that's the 44 TCP segments that are
`
`19· ·being received.
`
`20· · · · A.· After they are sent -- I mean -- okay.· That's
`
`21· ·fine.· I'm trying to answer your questions.· But just
`
`22· ·as long as it's clear that -- I mean, I would be
`
`23· ·surprised if you or Dr. Horst would disagree with the
`
`24· ·fact that receiving an acknowledgment would also
`
`25· ·require a --
`
`INTEL EX. 1225.016
`
`
`
`·1· · · · · · MR. MACK:· Garland, we're going to object to
`
`·2· ·this entire line of questioning, because you know this
`
`·3· ·is not a ground of instituted rejection.· It's not --
`
`·4· · · · · · MR. STEPHENS:· Stop talking.· Make your
`
`·5· ·objection.· Stop talking.
`
`·6· · · · · · MR. MACK:· Stop interrupting me.
`
`·7· · · · · · MR. STEPHENS:· No.
`
`·8· · · · · · MR. MACK:· I'm making an objection.
`
`·9· · · · · · MR. STEPHENS:· Make your objection.· But
`
`10· ·don't --
`
`11· · · · · · MR. MACK:· Stop interrupting me.
`
`12· · · · · · MR. STEPHENS:· You're not entitled --
`
`13· · · · · · (The reporter requested that people not speak
`
`14· · · · · · at once.)
`
`15· · · · · · MR. MACK:· This is not in your expert's
`
`16· ·report.· It's not an instituted ground.· It's not in
`
`17· ·Dr. Almeroth's declaration.
`
`18· · · · · · MR. STEPHENS:· Absolutely, it --
`
`19· · · · · · MR. MACK:· You know this is beyond the scope.
`
`20· · · · · · (The reporter requested that people not speak
`
`21· · · · · · at once; unreportable crosstalk.)
`
`22· · · · · · MR. MACK:· You're trying to make an entire new
`
`23· ·ground that wasn't instituted.· None of this is
`
`24· ·relevant.· We're going to move to strike all of these
`
`25· ·questions.· You know none of this is coming in your
`
`INTEL EX. 1225.017
`
`
`
`·1· ·reply.· You're wasting my time; you're wasting his
`
`·2· ·time.· This is all beyond the scope.· This is
`
`·3· ·harassment.
`
`·4· · · · · · MR. STEPHENS:· Are you done?
`
`·5· · · · · · MR. MACK:· Yes.
`
`·6· · · · · · MR. STEPHENS:· Okay.
`
`·7· · · · Q.· So Dr. Almeroth, before we were interrupted by
`
`·8· ·your counsel, or Alacritech's counsel -- let's see
`
`·9· ·here.· Okay.
`
`10· · · · · · So when it says, "Furthermore, since TCP must
`
`11· ·acknowledge all of this incoming data," that
`
`12· ·establishes that this sentence is talking about the
`
`13· ·side of the SMB transaction that is receiving the 44
`
`14· ·TCP segments.· Would you agree with that?
`
`15· · · · · · MR. MACK:· Objection.· Beyond the scope.
`
`16· · · · · · THE WITNESS:· Not necessarily.
`
`17· ·BY MR. STEPHENS:
`
`18· · · · Q.· Okay.· Why not?
`
`19· · · · A.· Read the sentence.
`
`20· · · · Q.· Okay.
`
`21· · · · A.· It says, "Furthermore, since TCP must
`
`22· ·acknowledge all of this incoming data" -- it's talking
`
`23· ·about TCP generally -- those acknowledgments would be
`
`24· ·sent by the receiver.· It's possible to get another 44
`
`25· ·transmit-complete interrupts.
`
`INTEL EX. 1225.018
`
`
`
`·1· · · · · · Okay.· You're -- where do those
`
`·2· ·transmit-complete interrupts happen?· They happen when
`
`·3· ·you receive the acknowledgment.· Those are generated as
`
`·4· ·a result of the sending out of TCP acknowledgments.· So
`
`·5· ·if the receiver sends out acknowledgments, the result
`
`·6· ·is that the original sender has to process those
`
`·7· ·acknowledgments, and those generate interrupts as well.
`
`·8· · · · · · I mean, it's -- you get acknowledgments.
`
`·9· ·They're TCP packets.· They absolutely have to be
`
`10· ·processed.· They absolutely generate interrupts.
`
`11· · · · Q.· Again, I'm just trying to understand what this
`
`12· ·sentence is referring to.
`
`13· · · · · · When it says "incoming data," what does that
`
`14· ·refer to?
`
`15· · · · A.· Look.· The sentence says what it says.· It's
`
`16· ·saying that as a result of sending out the
`
`17· ·acknowledgments, that the transmit-complete
`
`18· ·interrupts -- transmission is not completed until an
`
`19· ·acknowledgment is received.
`
`20· · · · Q.· Can you just answer my question, please?
`
`21· · · · A.· I am answering your question.
`
`22· · · · Q.· Okay.· Go ahead, then.
`
`23· · · · A.· You don't get a transmit-complete until an
`
`24· ·acknowledgment is received.
`
`25· · · · Q.· I didn't ask about a transmit-complete
`
`INTEL EX. 1225.019
`
`
`
`·1· ·interrupt.
`
`·2· · · · A.· Yes, you did.· It's right there in the
`
`·3· ·sentence.
`
`·4· · · · Q.· No.· I asked about incoming data.· What does
`
`·5· ·"incoming data" refer to?
`
`·6· · · · · · MR. MACK:· Objection.· Form.· Asked and
`
`·7· ·answered.
`
`·8· · · · · · THE WITNESS:· The incoming data is what's
`
`·9· ·received at the receiver and what generates
`
`10· ·acknowledgments.
`
`11· ·BY MR. STEPHENS:
`
`12· · · · Q.· Does it refer to the 44 TCP segments?
`
`13· · · · A.· Does what refer to the 44 TCP segments?
`
`14· · · · Q.· The incoming data.
`
`15· · · · · · MR. MACK:· Objection to form.
`
`16· ·BY MR. STEPHENS:
`
`17· · · · Q.· The phrase "incoming data."· Does that refer
`
`18· ·to the 44 TCP segments?
`
`19· · · · A.· It does.
`
`20· · · · Q.· And does that mean that it is being received
`
`21· ·by the TCP that's being referred to there?
`
`22· · · · A.· If the data is send by the sender, it will be
`
`23· ·received by the receiver, assuming that there are no
`
`24· ·errors.
`
`25· · · · Q.· But when it says "since TCP must acknowledge
`
`INTEL EX. 1225.020
`
`
`
`·1· ·all the incoming data," does that mean that the TCP
`
`·2· ·that must acknowledge the incoming data is receiving
`
`·3· ·that data?
`
`·4· · · · · · MR. MACK:· Objection.· Form.· Beyond the
`
`·5· ·scope.
`
`·6· · · · · · THE WITNESS:· It says that there must be
`
`·7· ·transmissions of acknowledgments by the receiver of the
`
`·8· ·data, and those acknowledgments must be processed by
`
`·9· ·the sender of the data.
`
`10· ·BY MR. STEPHENS:
`
`11· · · · Q.· Where does it talk about the sender of the
`
`12· ·data?
`
`13· · · · A.· It's possible --
`
`14· · · · Q.· Of the incoming data --
`
`15· · · · A.· -- to get another 44 transmit-complete
`
`16· ·interrupts.
`
`17· · · · · · The only thing that gets transmit-complete
`
`18· ·interrupts is the sender.· A transmit-complete happens
`
`19· ·when an acknowledgment is received at the transmitter.
`
`20· ·And that's -- that reception of an acknowledgment at
`
`21· ·the transmitter is the result of sending TCP
`
`22· ·acknowledgments by the receiver.
`
`23· · · · Q.· I -- you've completely lost me now.· When -- I
`
`24· ·don't know what you mean by sender and receiver.· You
`
`25· ·seem to be mixing them up.
`
`INTEL EX. 1225.021
`
`
`
`·1· · · · A.· Well, so there's the sender of the data.
`
`·2· · · · Q.· Sending -- excuse me.· Sender of the incoming
`
`·3· ·data.· Right?· So that's the party that's not receiving
`
`·4· ·the incoming data --
`
`·5· · · · A.· Let's --
`
`·6· · · · Q.· -- the outgoing data.
`
`·7· · · · A.· Let's do it this way.· Host A is going to do a
`
`·8· ·write request.
`
`·9· · · · Q.· Okay.
`
`10· · · · A.· So it's going to send 64 bytes of data.· It's
`
`11· ·going to send that 64 bytes of data to Host B.
`
`12· · · · Q.· Okay.
`
`13· · · · A.· There are interrupts generated in the sending
`
`14· ·of 44 TCP segments when --
`
`15· · · · Q.· By --
`
`16· · · · A.· -- when those segments are sent --
`
`17· · · · · · (Reporter requested clarification.)
`
`18· · · · · · THE WITNESS:· When Host A sends the data.
`
`19· ·Those interrupts will be on Host A.· Host B, the
`
`20· ·receiver of those 44 TCP segments, will receive -- will
`
`21· ·have interrupts as each of those segments are received
`
`22· ·and processed.
`
`23· · · · · · As a result of receiving each of those
`
`24· ·segments, Host B will generate an acknowledgment to
`
`25· ·Host A.· In some cases, it's one acknowledgment for
`
`INTEL EX. 1225.022
`
`
`
`·1· ·every received TCP segment; in other cases, it's less
`
`·2· ·than that.
`
`·3· · · · · · The generation of the acknowledgments by
`
`·4· ·Host B will have its own set of interrupts.· Those
`
`·5· ·acknowledgments are sent to Host A.· They look like TCP
`
`·6· ·packets.· They may or may not contain data.
`
`·7· · · · · · The acknowledgments that are received back at
`
`·8· ·Host A will generate their own set of interrupts as
`
`·9· ·well, because each of the headers has to be processed,
`
`10· ·whether or not there's data that might generate
`
`11· ·additional interrupts as it's passed to the
`
`12· ·application.· But certainly there are interrupts in
`
`13· ·processing those TCP packets.
`
`14· · · · · · So there's essentially two sets.· There are
`
`15· ·four -- four groups of interrupts that happen: Sending
`
`16· ·the data on Host A, receiving the data on Host B,
`
`17· ·generating acknowledgments on Host B, and receiving the
`
`18· ·acknowledgments on Host A.
`
`19· · · · Q.· And it's your testimony that the
`
`20· ·transmit-complete interrupts refers to the interrupts
`
`21· ·on Host A that are generated in response to receiving
`
`22· ·the acknowledgment packets transmitted by Host B.· Is
`
`23· ·that right?
`
`24· · · · · · MR. MACK:· Objection.· Form.· Beyond the
`
`25· ·scope.
`
`INTEL EX. 1225.023
`
`
`
`·1· · · · · · THE WITNESS:· It's my reading that a
`
`·2· ·transmit-complete interrupt is on the reception of an
`
`·3· ·acknowledgment.· Whether it's referring to the second
`
`·4· ·set, meaning the set that's on the receiver, doesn't
`
`·5· ·really matter.· It doesn't -- it doesn't change the
`
`·6· ·fact that there are four sets of interrupts that happen
`
`·7· ·in a send and a receive.
`
`·8· ·BY MR. STEPHENS:
`
`·9· · · · Q.· Okay.· At least I understand what you mean
`
`10· ·now.· Thank you for that.
`
`11· · · · · · Then it says:
`
`12· · · · · · "While this is possible, it's not terribly
`
`13· · · · likely.· Delayed ACK timers allow us to
`
`14· · · · acknowledge more than one segment at a time
`
`15· · · · and delays in interrupt processing may mean
`
`16· · · · that we are able to process more than one
`
`17· · · · incoming network frame per interrupt."
`
`18· · · · · · Does the sentence that says "and delays in
`
`19· ·interrupt prosing may mean that we are able to process
`
`20· ·more than one incoming network frame per interrupt"
`
`21· ·refer only to acknowledgments?
`
`22· · · · · · MR. MACK:· Objection.· Form.· Beyond the
`
`23· ·scope.
`
`24· · · · · · THE WITNESS:· No.· It can refer to the
`
`25· ·reception of any frame.· So any frame would include the
`
`INTEL EX. 1225.024
`
`
`
`·1· ·original data transmissions as well as -- sorry.
`
`·2· · · · · · The reception of the original data
`
`·3· ·transmissions at Host B and the reception of
`
`·4· ·acknowledgments at Host A.
`
`·5· ·BY MR. STEPHENS:
`
`·6· · · · Q.· Okay.· So you can reduce the number of
`
`·7· ·interrupts caused by receiving ACKs through delayed ACK
`
`·8· ·timers.· Is that correct?
`
`·9· · · · · · MR. MACK:· Objects form.· Beyond the scope.
`
`10· ·BY MR. STEPHENS:
`
`11· · · · Q.· Sorry.· You can reduce the number of ACKs
`
`12· ·transmitted by using a delayed ACK timer.· Is that
`
`13· ·right?
`
`14· · · · · · MR. MACK:· Same objections.
`
`15· · · · · · THE WITNESS:· You can.· And consequently, if
`
`16· ·you're sending fewer ACKs, you're also receiving fewer
`
`17· ·ACKs.
`
`18· ·BY MR. STEPHENS:
`
`19· · · · Q.· Okay.· And then -- and using delays in
`
`20· ·interrupt processing, you may be able to process more
`
`21· ·than one incoming network frame per interrupt of any
`
`22· ·kind, whether it's ACKs or data or both.
`
`23· · · · A.· Exactly.
`
`24· · · · Q.· Got it.· Thank you.· Okay.
`
`25· · · · · · So then it goes on to say:
`
`INTEL EX. 1225.025
`
`
`
`·1· · · · · · "Nevertheless, even if we assume four
`
`·2· · · · incoming frames per input and acknowledgment
`
`·3· · · · for every 2 segments (as is typical for" --
`
`·4· · · · "typical per the ACK-every-other-segment
`
`·5· · · · property of TCP), we are still left with 33
`
`·6· · · · interrupts per 64K SMB request."
`
`·7· · · · · · Now, is there anything inaccurate about that
`
`·8· ·paragraph?
`
`·9· · · · A.· Yeah.
`
`10· · · · · · MR. MACK:· Objection.· Form.
`
`11· · · · · · THE WITNESS:· Well, I don't know if it's
`
`12· ·inaccurate, but I think it's kind of an anecdote and
`
`13· ·represents a general sort of calculation.
`
`14· · · · · · I think, depending on the implementation, you
`
`15· ·could certainly have more interrupts, or you could
`
`16· ·potentially have fewer interrupts.· I mean, depending
`
`17· ·on how large blocks of data are processed.
`
`18· · · · · · I think what is clear, and as I testified to
`
`19· ·yesterday, the concept that there are too many
`
`20· ·interrupts is accurate.· The concept of what happens in
`
`21· ·a transmission, kind of the four portions between
`
`22· ·Host A and Host B that I described and that this
`
`23· ·section is describing, where interrupts can happen in
`
`24· ·that process, is accurate.
`
`25· · · · · · But I don't think that every single
`
`INTEL EX. 1225.026
`
`
`
`·1· ·implementation in that exact scenario would necessarily
`
`·2· ·generate exactly 33 interrupts per 64 kilobyte SMB
`
`·3· ·request.
`
`·4· · · · Q.· Fair enough.
`
`·5· · · · A.· Sorry.· And I don't think a person of skill in
`
`·6· ·the art reading this would understand that that's what
`
`·7· ·the inventors intended.
`
`·8· · · · Q.· Okay.· So in some circumstances, there may be
`
`·9· ·fewer interrupts for 44 TCP segments, and in some cases
`
`10· ·there might be more.· Is that what you're saying?
`
`11· · · · A.· Yes.· That's what I'm saying.
`
`12· · · · Q.· Okay.· What circumstances would result in
`
`13· ·more?
`
`14· · · · · · MR. MACK:· Objection.· Form.· Beyond the
`
`15· ·scope.
`
`16· · · · · · THE WITNESS:· Situations where there's lost
`
`17· ·data and retransmissions, out-of-order delivery, time
`
`18· ·expiration, congestion constraints.· I mean, there