InfiniBand

Template:Short description Template:Infobox organization Template:Redirects here InfiniBand (IB) is a computer networking communications standard used in high-performance computing that features very high throughput and very low latency. It is used for data interconnect both among and within computers. InfiniBand is also used as either a direct or switched interconnect between servers and storage systems, as well as an interconnect between storage systems. It is designed to be scalable and uses a switched fabric network topology. Between 2014 and June 2016,<ref name="down">{{#invoke:citation/CS1|citation |CitationClass=web }}</ref> it was the most commonly used interconnect in the TOP500 list of supercomputers.

Mellanox (acquired by Nvidia) manufactures InfiniBand host bus adapters and network switches, which are used by large computer system and database vendors in their product lines.<ref name="oracle">{{#invoke:citation/CS1|citation |CitationClass=web }}</ref>

As a computer cluster interconnect, IB competes with Ethernet, Fibre Channel, and Intel Omni-Path. The technology is promoted by the InfiniBand Trade Association.

HistoryEdit

InfiniBand originated in 1999 from the merger of two competing designs: Future I/O and Next Generation I/O (NGIO). NGIO was led by Intel, with a specification released in 1998,<ref>Template:Cite news</ref> and joined by Sun Microsystems and Dell. Future I/O was backed by Compaq, IBM, and Hewlett-Packard.<ref>Template:Cite news</ref> This led to the formation of the InfiniBand Trade Association (IBTA), which included both sets of hardware vendors as well as software vendors such as Microsoft. At the time it was thought some of the more powerful computers were approaching the interconnect bottleneck of the PCI bus, in spite of upgrades like PCI-X.<ref name=pentakalos>{{#invoke:citation/CS1|citation |CitationClass=web }}</ref> Version 1.0 of the InfiniBand Architecture Specification was released in 2000. Initially the IBTA vision for IB was simultaneously a replacement for PCI in I/O, Ethernet in the machine room, cluster interconnect and Fibre Channel. IBTA also envisaged decomposing server hardware on an IB fabric.

Mellanox had been founded in 1999 to develop NGIO technology, but by 2001 shipped an InfiniBand product line called InfiniBridge at 10 Gbit/second speeds.<ref name=timeline>{{#invoke:citation/CS1|citation |CitationClass=web }}</ref> Following the burst of the dot-com bubble there was hesitation in the industry to invest in such a far-reaching technology jump.<ref name=kim>{{#invoke:citation/CS1|citation |CitationClass=web }}</ref> By 2002, Intel announced that instead of shipping IB integrated circuits ("chips"), it would focus on developing PCI Express, and Microsoft discontinued IB development in favor of extending Ethernet. Sun Microsystems and Hitachi continued to support IB.<ref>{{#invoke:citation/CS1|citation |CitationClass=web }}</ref>

In 2003, the System X supercomputer built at Virginia Tech used InfiniBand in what was estimated to be the third largest computer in the world at the time.<ref>Template:Cite news</ref> The OpenIB Alliance (later renamed OpenFabrics Alliance) was founded in 2004 to develop an open set of software for the Linux kernel. By February, 2005, the support was accepted into the 2.6.11 Linux kernel.<ref>Template:Cite news</ref><ref>Template:Cite news</ref> In November 2005 storage devices finally were released using InfiniBand from vendors such as Engenio.<ref name="comeback">Template:Citation</ref> Cisco, desiring to keep technology superior to Ethernet off the market, adopted a "buy to kill" strategy. Cisco successfully killed InfiniBand switching companies such as Topspin via acquisition.<ref>{{#invoke:citation/CS1|citation |CitationClass=web }}</ref> Template:Citation needed

Of the top 500 supercomputers in 2009, Gigabit Ethernet was the internal interconnect technology in 259 installations, compared with 181 using InfiniBand.<ref>{{#invoke:citation/CS1|citation |CitationClass=web }}</ref> In 2010, market leaders Mellanox and Voltaire merged, leaving just one other IB vendor, QLogic, primarily a Fibre Channel vendor.<ref>{{#invoke:citation/CS1|citation |CitationClass=web }}</ref> At the 2011 International Supercomputing Conference, links running at about 56 gigabits per second (known as FDR, see below), were announced and demonstrated by connecting booths in the trade show.<ref>Template:Cite news</ref> In 2012, Intel acquired QLogic's InfiniBand technology, leaving only one independent supplier.<ref>Template:Cite news</ref>

By 2014, InfiniBand was the most popular internal connection technology for supercomputers, although within two years, 10 Gigabit Ethernet started displacing it.<ref name="down"/>

In 2016, it was reported that Oracle Corporation (an investor in Mellanox) might engineer its own InfiniBand hardware.<ref name="oracle"/>

In 2019 Nvidia acquired Mellanox, the last independent supplier of InfiniBand products.<ref>Template:Cite news</ref>

SpecificationEdit

Specifications are published by the InfiniBand trade association.

PerformanceEdit

Original names for speeds were single-data rate (SDR), double-data rate (DDR) and quad-data rate (QDR) as given below.<ref name="comeback" /> Subsequently, other three-letter acronyms were added for even higher data rates.<ref name="fdr_fact_sheet">{{#invoke:citation/CS1|citation |CitationClass=web }}</ref>

InfiniBand unidirectional data rates
		1x	Line code		Signaling rate (Gbit/s)	4x	8x	12x
	Year<ref name="ccgrid11-ib-hse-23">{{#invoke:citation/CS1\|citation	CitationClass=web }}</ref>	Line code		Signaling rate (Gbit/s)	Throughput (Gbit/s)<ref name="ib_over">{{#invoke:citation/CS1\|citation				CitationClass=web }}</ref>	Adapter latency (μs)<ref>http://www.hpcadvisorycouncil.com/events/2014/swiss-workshop/presos/Day_1/1_Mellanox.pdf // Mellanox</ref>
Template:Abbr	2001, 2003	NRZ	8b/10b<ref>{{#invoke:citation/CS1\|citation	CitationClass=web }}</ref>	2.5	2	8	16	24	5
Template:Abbr	2005			5	4	16	32	48	2.5
Template:Abbr	2007			10	8	32	64	96	1.3
Template:Abbr	2011		64b/66b	citation	CitationClass=web }}</ref>	10	40	80	120	0.7
Template:Abbr	2011			citation	CitationClass=web }}</ref><ref name="fdr_fact_sheet" />	13.64	54.54	109.08	163.64	0.7
Template:Abbr	citation			CitationClass=web }}</ref>	25.78125	25	100	200	300	0.5
Template:Abbr	2018<ref name="ib_roadmap" />	PAM4		citation	CitationClass=web }}</ref>	50	200	400	600	<0.6<ref>https://www.mellanox.com/files/doc-2020/pb-connectx-6-vpi-card.pdf Template:Bare URL PDF</ref>
Template:Abbr	2022<ref name="ib_roadmap" />		256b/257bTemplate:Efn-lr	citation	CitationClass=web }}</ref>	100	400	800	1200	Template:Dunno
Template:Abbr	citation		CitationClass=web }}</ref>	Template:TBD	200	200	800	1600	2400	Template:TBD
Template:Abbr	Template:TBA	Template:TBD	400	Template:TBD	400	1600	3200	4800		Template:TBD

Notes

Template:Notelist-lr

Each link is duplex. Links can be aggregated: most systems use a 4 link/lane connector (QSFP). HDR often makes use of 2x links (aka HDR100, 100 Gb link using 2 lanes of HDR, while still using a QSFP connector). 8x is called for with NDR switch ports using OSFP (Octal Small Form Factor Pluggable) connectors {{#invoke:citation/CS1|citation |CitationClass=web }}

InfiniBand provides remote direct memory access (RDMA) capabilities for low CPU overhead.

TopologyEdit

InfiniBand uses a switched fabric topology, as opposed to early shared medium Ethernet. All transmissions begin or end at a channel adapter. Each processor contains a host channel adapter (HCA) and each peripheral has a target channel adapter (TCA). These adapters can also exchange information for security or quality of service (QoS).

MessagesEdit

InfiniBand transmits data in packets of up to 4 KB that are taken together to form a message. A message can be:

a remote direct memory access read or write
a channel send or receive
a transaction-based operation (that can be reversed)
a multicast transmission
an atomic operation

Physical interconnectionEdit

File:Infinibandport.jpg

InfiniBand switch with CX4/SFF-8470 connectors

In addition to a board form factor connection, it can use both active and passive copper (up to 10 meters) and optical fiber cable (up to 10 km).<ref name=faq>{{#invoke:citation/CS1|citation |CitationClass=web }}</ref> QSFP connectors are used.

The InfiniBand Association also specified the CXP connector system for speeds up to 120 Gbit/s over copper, active optical cables, and optical transceivers using parallel multi-mode fiber cables with 24-fiber MPO connectors.Template:Citation needed

Software interfacesEdit

Mellanox operating system support is available for Solaris, FreeBSD,<ref>{{#invoke:citation/CS1|citation |CitationClass=web }}</ref><ref>{{#invoke:citation/CS1|citation |CitationClass=web }}</ref> Red Hat Enterprise Linux, SUSE Linux Enterprise Server (SLES), Windows, HP-UX, VMware ESX,<ref>{{#invoke:citation/CS1|citation |CitationClass=web }}</ref> and AIX.<ref>{{#invoke:citation/CS1|citation |CitationClass=web }}</ref>

InfiniBand has no specific standard application programming interface (API). The standard only lists a set of verbs such as ibv_open_device or ibv_post_send, which are abstract representations of functions or methods that must exist. The syntax of these functions is left to the vendors. Sometimes for reference this is called the verbs API. The de facto standard software is developed by OpenFabrics Alliance and called the Open Fabrics Enterprise Distribution (OFED). It is released under two licenses GPL2 or BSD license for Linux and FreeBSD, and as Mellanox OFED for Windows (product names: WinOF / WinOF-2; attributed as host controller driver for matching specific ConnectX 3 to 5 devices)<ref>Mellanox OFED for Windows - WinOF / WinOF-2</ref> under a choice of BSD license for Windows. It has been adopted by most of the InfiniBand vendors, for Linux, FreeBSD, and Microsoft Windows. IBM refers to a software library called libibverbs, for its AIX operating system, as well as "AIX InfiniBand verbs".<ref>{{#invoke:citation/CS1|citation |CitationClass=web }}</ref> The Linux kernel support was integrated in 2005 into the kernel version 2.6.11.<ref>{{#invoke:citation/CS1|citation |CitationClass=web }}</ref>

Ethernet over InfiniBandEdit

Ethernet over InfiniBand, abbreviated to EoIB, is an Ethernet implementation over the InfiniBand protocol and connector technology. EoIB enables multiple Ethernet bandwidths varying on the InfiniBand (IB) version.<ref>{{#invoke:citation/CS1|citation |CitationClass=web }}</ref> Ethernet's implementation of the Internet Protocol Suite, usually referred to as TCP/IP, is different in some details compared to the direct InfiniBand protocol in IP over IB (IPoIB).

Ethernet over InfiniBand performance
Type	Lanes	Bandwidth (Gbit/s)	Compatible Ethernet type(s)	Compatible Ethernet quantity
SDR	Template:01	Template:02.5	GbE to 2.5 GbE	Template:02 × GbE to 1 × Template:02.5 GbE
	Template:04	Template:010	GbE to 10 GbE	10 × GbE to 1 × 10 GbE
	Template:08	Template:020	GbE to 10 GbE	20 × GbE to 2 × 10 GbE
	Template:012	Template:030	GbE to 25 GbE	30 × GbE to 1 × 25 GbE + 1 × Template:05 GbE
DDR	Template:01	Template:05	GbE to 5 GbE	Template:05 × GbE to 1 × Template:05 GbE
	Template:04	Template:020	GbE to 10 GbE	20 × GbE to 2 × 10 GbE
	Template:08	Template:040	GbE to 40 GbE	40 × GbE to 1 × 40 GbE
	Template:012	Template:060	GbE to 50 GbE	60 × GbE to 1 × 50 GbE + 1 × 10 GbE
QDR	Template:01	Template:010	GbE to 10 GbE	10 × GbE to 1 × 10 GbE
QDR	Template:04	Template:040	GbE to 40 GbE	40 × GbE to 1 × 40 GbE

ReferencesEdit

Template:Reflist

External linksEdit

Template:Computer-bus Template:Authority control