Unless otherwise indicated herein, the approaches described in this section are not prior art to the claims in this application and are not admitted to be prior art by inclusion in this section.
Data centers perform countless computing jobs for businesses and individual users. A modern data center, for example, may enable tens of thousands of individuals to browse the Internet or perform operations using extensive computational resources. To perform these duties, data centers often rely on communications between servers in the data center. Currently, interfaces responsible for communications between servers often store data of a packet when received and then, once all data of that data packet is stored, forwards the data packet. This technique, however, can be slow because of the latency inherent in waiting to forward a packet until the packet has been fully stored.
Another conventional technique may instead be used by the interfaces. This other technique begins to forward a packet prior to all data of that packet being stored. By so doing it can be faster than the store-and-forward technique noted above due to having a very low or zero latency. This other technique, however, is often slow as well because the technique cannot be used to forward packets received at lower-speed transmission rates at higher-speed transmission rates, as this causes an under-run condition due to the speed mismatch. When such a situation exists, the higher-speed transmission rates may go unused, causing slow communications between servers, or the transmitted packets may be corrupted.
This summary is provided to introduce subject matter that is further described below in the Detailed Description and Drawings. Accordingly, this Summary should not be considered to describe essential features nor used to limit the scope of the claimed subject matter.
In one embodiment, a method is described comprising receiving packet streams at a lower-speed transmission rate, marking packets of the packet streams effective to indicate associations between the packets and the packet streams, and transmitting the marked packets of the packet streams in a cut-through mode, interleaved, and at a higher-speed transmission rate than the lower-speed transmission rate.
In another embodiment, a method is described comprising receiving, at a higher-speed transmission rate, marked packets, wherein the marked packets are interleaved, and wherein the marked packets associated with different packet streams having a lower-speed transmission rate, determining, based on a marking on each of the marked packets, to which of the different packet streams each marked packet is associated, and recreating each of the different packet streams based on which of the different streams each marked packets is associated with.
In still another embodiment, an apparatus is described having a buffer configured to buffer packets associated with packet streams having a lower-speed transmission rate, a controller configured to pull data from the buffer as each packet associated with each packet stream is buffered and mark each of the packets as said packet is buffered, the marking effective to indicate to which of the packet streams said packet is associated, and a media access controller configured to receive the marked packets and transmit the marked packets at a higher-speed transmission rate than the lower-speed transmission rate.
The detailed description is described with reference to the accompanying figures. In the figures, the left-most digit of a reference number identifies the figure in which the reference number first appears. The use of the same reference numbers in different instances in the description and the figures indicate similar or identical items.
As noted in the Background above, communication interfaces using conventional store-and-forward or cut-through communication techniques can be slow due to latency or an overrun condition, respectively. This disclosure describes techniques and apparatuses enabling low-to-high speed cut-through communication without creating an overrun condition. By so doing, the techniques and/or apparatuses enable communication interfaces to communicate at higher speed, such as by avoiding store-to-forward latency.
This discussion proceeds to describe an example operating environment in which the techniques may operate, methods performable by the techniques, and an example apparatus below.
Operating Environment
Communication interface 104 includes a buffer 112, an interface controller 114, a direct memory access module (DMA) 116, and a media access controller (MAC) 118. As will be described in greater detail below, communication interface 104 buffers data from packet streams 108 received a lower transmission rates, generally all at once. Controller 114, using DMA 116, accesses the data from buffer 112, marks the data such that the marked data can later be associated with its respective packet stream 108, and transmits the data at higher-speed transmission rates using MAC 118. Interface controller 114 can transmit the data from these lower-speed packet streams 108 in a cut-through mode, beginning to transmit data when packets of data are partially received. Note that the lower-speed transmission rates may or may not be equal or, when combined, sufficient to use all of the higher-speed transmission rate (e.g., pipes 102 may not equal pipe 106 or one or more of pipes 102 may not include a data stream).
By way of example, consider example environment 100 in the context of an Ethernet communication network. In this context, low-speed pipes 102 can be 10 gigabits/second pipes and high-speed pipe 106 a 40 or 100 gigabits/second pipe, while communication interface 104 is an Ethernet-capable interface.
Techniques for Low-to-High Speed Cut-Through Communication
The following section describes various techniques for low-to-high speed cut-through communication. Aspects of these techniques may be implemented in hardware, firmware, software, or a combination thereof. These techniques are illustrated in part using example methods, which are shown as acts that specify operations performed by one or more entities. These acts are not necessarily limited to the order shown or to entities performing them.
At 204, packets of the packet streams are marked effective to indicate associations between packets and packet streams. This marking of the packets can be performed when each packet is partially received (and/or buffered) or at an end of a packet and after transmission has begun, or, if a received packet is to be transmitted in segments, at a beginning or end of each segment. In all of these cases, however, the markings enable an upstream entity to determine to which packet stream a packet is associated without prohibiting cut-through transmission of the packets.
Packets and portions of packets (if a received packet is segmented prior to transmission) can be marked in various ways and according to various communication protocols. In some cases, packets are segmented and identified with a context and these segments sent as a marked, segmented packet (either separately or combined) and following a particular protocol, such as an Institute of Electrical and Electronics Engineers (IEEE)-standard and complete packet. In such a case, multiple IEEE standard packets may be used to transmit the packet received at block 204, and in some cases other packets received as part of other packet streams 108. Segments of these IEEE-standard packets can later be re-assembled based on their markings. Manners in which an upstream entity may handle marked packets are covered hereinafter.
Continuing the ongoing example, communication interface 104 receives packets from low-speed packet streams 108 and through low-speed pipes 102 and buffers each packet into a dedicated compartment of buffer 112, after which DMA 116 pulls data from buffer 112 and controller 114 marks each packet with appropriate context sufficient to reassemble the packet stream from the marked packets.
At 206, the marked packets of the packets streams are transmitted in a cut-through mode, interleaved, and at a higher-speed transmission rate than the lower-speed transmission rate at which they were received. This is visually represented in
For some Ethernet packet streams, for example, method 200 segments a set of Ethernet packets from low-speed interfaces and interleaves these segments within a normal, outgoing Ethernet packet on a higher-speed port. Each segment of the Ethernet packets being sent out at the higher-speed port includes internal markings identifying the source of the segment. Thus, at 204, the markings mark each segment and indicate from which packet stream each segment originates. At 206, the marked segments are interleaved into an Ethernet packet and transmitted.
In some cases a number of packet streams received is insufficient to make full use of higher-speed transmission rates. Consider, for example, a case where three data streams 108 received, all at equal (10-gigabits/second) transmission rates but that the higher-speed transmission rate is 40 gigabits/second. In such a case controller 114 creates blank, marked packets and transmits these interleaved with marked packets of the three data streams 108. These marked, blank packets are marked sufficient to indicate that they are place holders rather than data for a data stream.
By way of example, consider
Upstream interface 120 can be dissimilar, similar, or identical to interface 104 of
In this example, assume that marked packet 502-1 is received by upstream interface 120 at 40 gigabits/second and marked with a tag in a header to associate marked packet 502-1 with low-speed packet stream 108-1 (assume low-speed is 10 gigabits/second). Marked packet 502-1 is received first, followed immediately by marked packet 502-2, which is marked with a tag in a header to associate marked packet 502-2 with low-speed packet stream 108-2. Marked packet 502-1 is followed immediately by blank, marked packet 504, which is marked with a tag in a header to associate blank, marked packet 504 with no packet stream (thus, low-speed packet stream 108-3 is empty/does not exist). Following reception of blank, marked packet 504, marked packet 502-4 is immediately received, tagged in its header to associate marked packet 502 with low-speed packet stream 108-4.
Note that in some embodiments marked packets (e.g., 502-1, 502-2, and 502-4 and blank, marked packet 504) can be transmitted as a single Ethernet packet. In such a case, controller 114 marks segments of the data streams 108, after which MAC 118 transmits these segments within a single IEEE standard packet. Thus, the MAC 118 can act in a standard manner and without knowledge or interaction with the segments or their markings. In such a manner Ethernet MAC compliance can be maintained. This single Ethernet packet is received at block 402 having segments from multiple different packet streams 108. Each of these segments is marked sufficient to recreate these different packet streams 108.
At 404, it is determined to which of the different packet streams each marked packet (or its segment) is associated based on a marking on each of the marked packets. This can be determined by reading a tag in a header, a tag in some other portion of a marked packet, and/or other marking that provides appropriate context.
Continuing the ongoing example, assume that upstream interface 120 receives the packets at block 402 and buffers each marked packet in upstream buffer 122. After, before, or commensurate with buffering each market packet, upstream controller 510 reads the tag of each marked packet and determines to which low-speed packet stream 108 the packet belongs (if any). Thus, in this example upstream controller 510 determines that marked packet 502-1 is associated with low-speed packet stream 108-1, marked packet 502-2 is associated with low-speed packet stream 108-2, and marked packet 502-4 is associated with low-speed packet stream 108-4. Upstream controller 510 also determines that blank, marked packet 504 is not associated with a packet stream.
At 406, the different packet streams are recreated with data received in the marked packets. The techniques can recreate the different data streams in various ways, such as to read each marked packet as the marked packet is being received and/or buffered, stripping off the tag or other marking, and buffering either the data of the packet, the packet without the marking, or the marked packet as-is. The buffering can be made into a dedicated portion or section of a buffer or otherwise allocated to the appropriate lower-speed packet stream.
Continuing the ongoing example, assume that controller 510 of upstream interface 120 receives marked packets 502 and 504 as noted above, determines to which packet stream each belongs, if any, and in the case of marked packets 502, strips the tags from each header and buffers (using DMA 512) the now-unmarked packets in upstream buffer 508 in portions dedicated to the respective low-speed packet streams 108. In the case of blank, marked packet 504, upstream controller 510 discards the packet. This buffering is illustrated in
Method 400 may continue at 408, at which packets of data streams are retransmitted, such as in a cut-through mode, interleaved, and at a highest-speed transmission rate higher than the higher-speed transmission rate. In some communication networks an interface, such as an Ethernet interface in an Ethernet communication network, transmits data at varying transmission speeds and through a hierarchy. Such an example network may include, for example, 10, 40, and 100 gigabits/second pipes.
Continuing the ongoing example, assume that upstream interface 120 retransmits the three lower-speed packet streams shown in
Alternatively, upstream interface 120 may transmit buffered packets through low-speed pipes, such as back to various servers or server-accessible memory. This is shown in
System-on-Chip Example
SoC 700 can be integrated with electronic circuitry, a microprocessor, memory, input-output (I/O) logic control, communication interfaces and components, other hardware, firmware, and/or software needed to run an entire device. SoC 700 can also include an integrated data bus (not shown) that couples the various components of the SoC for data communication between the components. A device that includes SoC 700 can also be implemented with many combinations of differing components.
In this example, SoC 700 includes various components such as an input-output (I/O) logic control 702 (e.g., to include electronic circuitry) and a microprocessor 704 (e.g., any of a microcontroller or digital signal processor). SoC 700 also includes a memory 706, which can be any type of random access memory (RAM), a low-latency nonvolatile memory (e.g., flash memory), read only memory (ROM), and/or other suitable electronic data storage. SoC 700 can also include various firmware and/or software, such as an operating system 708, which can be computer-executable instructions maintained by memory 706 and executed by microprocessor 704. SoC 700 can also include other various communication interfaces and components, wireless LAN (WLAN) or PAN (WPAN) components, other hardware, firmware, and/or software.
SoC 700 may include controller 114 of
Controller 114 in SoC 700, either independently or in combination with other entities, can be implemented as computer-executable instructions maintained by memory 706 and executed by microprocessor 704 to implement various embodiments and/or features described herein. Controller 114 may also be provided integral with other entities of the SoC, such as integrated with DMA 116. Alternatively or additionally, controller 114 and the other components can be implemented as hardware, firmware, fixed logic circuitry, or any combination thereof that is implemented in connection with the I/O logic control 702 and/or other signal processing and control circuits of SoC 700.
Although the subject matter has been described in language specific to structural features and/or methodological techniques and/or acts, it is to be understood that the subject matter defined in the appended claims is not necessarily limited to the specific features, techniques, or acts described above, including orders in which they are performed.
This application claims priority to U.S. Provisional Patent Application Ser. No. 61/298,837 filed Jan. 2, 2010, the disclosure of which is incorporated by reference herein in its entirety.
Number | Name | Date | Kind |
---|---|---|---|
5307345 | Lozowick | Apr 1994 | A |
5784559 | Frazier | Jul 1998 | A |
6076062 | Van Steenbrugge | Jun 2000 | A |
6128666 | Muller et al. | Oct 2000 | A |
6317440 | Sung | Nov 2001 | B1 |
6529957 | Joergensen | Mar 2003 | B1 |
6542581 | Suonsivu et al. | Apr 2003 | B2 |
6791942 | Jin | Sep 2004 | B2 |
6920132 | Lo | Jul 2005 | B1 |
6938097 | Vincent | Aug 2005 | B1 |
6975637 | Lenell | Dec 2005 | B1 |
7058823 | Lapidus | Jun 2006 | B2 |
7127521 | Hsu et al. | Oct 2006 | B2 |
7379422 | Nation | May 2008 | B2 |
7450509 | Hao | Nov 2008 | B2 |
7519370 | Febvre et al. | Apr 2009 | B2 |
7558874 | Kodukula et al. | Jul 2009 | B1 |
7664972 | Diab et al. | Feb 2010 | B2 |
7760723 | Daines et al. | Jul 2010 | B1 |
RE41494 | Ahern et al. | Aug 2010 | E |
7826360 | Hutchison | Nov 2010 | B1 |
7839152 | Lee et al. | Nov 2010 | B2 |
7885546 | Liu et al. | Feb 2011 | B2 |
8009574 | Schenk | Aug 2011 | B2 |
8401014 | Capurso et al. | Mar 2013 | B2 |
8670335 | Barkan et al. | Mar 2014 | B2 |
20020087716 | Mustafa | Jul 2002 | A1 |
20040001508 | Zheng | Jan 2004 | A1 |
20040095935 | Connor | May 2004 | A1 |
20050163117 | Rhim | Jul 2005 | A1 |
20060023735 | Sasson | Feb 2006 | A1 |
20060063498 | Chan | Mar 2006 | A1 |
20060268733 | Rhee et al. | Nov 2006 | A1 |
20070248118 | Bishara et al. | Oct 2007 | A1 |
20080089433 | Cho et al. | Apr 2008 | A1 |
20080123555 | Qi et al. | May 2008 | A1 |
20080129118 | Diab | Jun 2008 | A1 |
20080148081 | Diab et al. | Jun 2008 | A1 |
20080172564 | Diab et al. | Jul 2008 | A1 |
20080192631 | Ardhanari et al. | Aug 2008 | A1 |
20080212485 | Schenk | Sep 2008 | A1 |
20080219289 | Harrison et al. | Sep 2008 | A1 |
20080225881 | Powell | Sep 2008 | A1 |
20080310304 | Edwards, III et al. | Dec 2008 | A1 |
20090154465 | Diab et al. | Jun 2009 | A1 |
20090210725 | Kim et al. | Aug 2009 | A1 |
20090252054 | Barkan | Oct 2009 | A1 |
20100077239 | Diab et al. | Mar 2010 | A1 |
20100262850 | Lin et al. | Oct 2010 | A1 |
20110051620 | Taich et al. | Mar 2011 | A1 |
Number | Date | Country |
---|---|---|
1246398 | Oct 2002 | EP |
WO-2006052360 | May 2006 | WO |
Entry |
---|
“Amendment: Physical Layer and Management Parameters for 10 GB/s Operation—Type 10GBASE-T”, (2004), LAN MAN Standards Committee of the IEEE Computer Society, (2004) 145 pages. |
“Non-Final Office Action”, U.S. Appl. No. 12/193,519, (Mar. 29, 2010), 21 pages. |
“Non-Final Office Action”, U.S. Appl. No. 12/415,705, (Apr. 11, 2011), 27 pages. |
“Notice of Allowance”, U.S. Appl. No. 12/193,519, (Jun. 22, 2010), 6 pages. |
“Part 3: Carrier Sense Multiple Access with Collision Detection (CSMA/CD) Access Method and Physical Layer Specifications”, IEEE Standards, (Mar. 8, 2002), 10 pages. |
“PCT Search Report”, Application No. PCT/US2009/038989, (Jul. 3, 2009), 14 pages. |
“U.S. Appl. No. 61/041,744”, filed Apr. 2, 2008, 10 pages. |
“Final Office Action”, U.S. Appl. No. 12/415,705, (Nov. 22, 2011), 36 pages. |
“Non-Final Office Action”, U.S. Appl. No. 12/415,705, Feb. 20, 2013, 26 pages. |
“Final Office Action”, U.S. Appl. No. 12/415,705, (Jun. 25, 2013), 28 pages. |
“Advisory Action”, U.S. Appl. No. 12/415,705, Sep. 20, 2013, 2 pages. |
“Notice of Allowance”, U.S. Appl. No. 12/415,705, Oct. 15, 2013, 5 pages. |
“Foreign Office Action”, TW Application No. 98110958, Sep. 23, 2014, 13 Pages. |
“Foreign Office Action”, TW Application No. 98110958, Aug. 20, 2015, 8 Pages. |
Number | Date | Country | |
---|---|---|---|
61298837 | Jan 2010 | US |