Browsed by
Tag: 10GbE

Introduction to 10GbE/25GbE/40GbE/100GbE Fiber Optic Cabling

Introduction to 10GbE/25GbE/40GbE/100GbE Fiber Optic Cabling

Technology is changing rapidly. Just when you got used to Gigabit Ethernet speeds being a fast & reliable system, someone unveiled 10GbE, 25GbE, 40GbE or even 100GbE systems a few years later. The newer and higher performing iterations are indeed the great breakthrough for telecommunication industry, but also pose difficulty in choosing network migration path—10G to 40G to 100G, or to 25G to 50G to 100G. We have described 10G, 25G, 40G and 100G Ethernet technology before, now in this blog, we’d like to introduce the four fiber optic cabling, and compare two 100G migration paths.

Cost-effective 10GbE Fiber Optic Cabling

10 Gigabit Ethernet technology defined by IEEE 802.3ae-2002 standard, is matured nowadays. Just like the “old” Gigabit Ethernet, 10Gb network can be terminated with either copper or fiber cabling. 1000BASE-T standard usually uses the Cat5e cables as the transmission media, while 10GbE bandwidth requires high grade copper cables like Cat6/Cat6a/Cat7 cables to support 10Gbps data rate. For instance, 10G SFP+ 10GBASE-T transceiver modules utilize Ethernet copper cables (Cat6a/Cat7) for a link length of 30m. SFP+ direct attach cables (DAC) and active optical cables (AOC) are also regarded as the cost-effective solutions for 10G short-reach applications. Besides 10G copper cables, there are single-mode (OS2) and multimode fiber patch cables (OM3/OM4) applied to different 10GbE IEEE standards. For the detailed information about the 10G cabling options, please see the following table.

10G fiber optic cable

As to the 10G fiber optic transceivers, there are a series of optical form factors including the XENPAK, X2, XFP, SFP+. The former three 10GbE optical transceivers were released earlier than smaller 10G form factor—SFP+ module. However, owing to their larger footprint, they are not successful on the 10G hardware market. Furthermore, SFP+ optics, compliant with several IEEE standards (SR, LR, LRM, ER, ZR and 10GBASE-T…) wins the heart of 10G end-users.

Singe-lane Design Makes 25GbE Shine

When 25G Ethernet was developed to support a single-lane 25Gbps standard in 2014, it was treated as the “new” 10GbE technology but delivers 2.5 times more data. Compared to 40GbE that was based on 10GbE, 25GbE with one lane obviously improves the port density and cost requirement. 25GbE network can support both copper and fiber optic cables, seen in the below table.Similar to 10GbE networks, 25G Ethernet physical interface specification supports several 25Gbps capable form factors, including CFP/CFP2/CFP4, SFP28 (1×25 Gbps) and QSFP28 (4×25 Gbps), which is also used for 100GbE. SFP28 25GBASE-SR and 25GBASE-LR SFP28 are two popular 25GbE optical transceiver modules available on the market, the former supports up to 100m link length while the latter allows a maximum transmission distance of 10 km.

25G optical modules

The available optical switches of the market do not support direct 25GbE connections using an SFP28 direct attach copper (DAC) cable. It is recommended to use a breakout cable that allows four 25GbE ports to connect to a 100GbE QFSFP28 switch port. FS.COM SFP28 DAC cable lengths are limited to four meters (1m, 2m, 3m, 5m) for 25GbE. And if you prefer a longer length, the 25GbE active optic cable (AOC) solutions are good recommendations.

25G Optics SFP28 Type
Media/Reach
All 25G SFP28 Ports 25GBASE-SR 50µm MMF / 70m
25GBASE-LR 9µm SMF / 10km
25GBASE-AOC Pre-terminated in 3, 5, 7, 10, 15, 20, 25, 30m lengths
OM4 MMF MTP/MPO 150m
25G Copper SFP28 Type Media/Reach
All 25G SFP28 Ports 25GBASE-CR Twinax / ‘Direct Attach’ Pre-terminated in 1m, 2m, 3m, 5m lengths

 

Fast & Reliable 40GbE Fiber Optic Cabling

Like the 10GbE fiber optic cabling, there are several IEEE standards of 40GbE transceiver in the whole evolution. 40G QSFP+ optical transceivers are the most commonly used optics for 40G network. So how to choose fiber optic cables for 40G optical transceivers? The following table will help you out.

40G modules

Besides the QSFP+ fiber transceivers and fiber optic cables, 40G DAC cables available in QSFP+ DAC cables and AOC cables enable short-reach options. For 40G cabling, QSFP+ to QSFP+ (40G to 40G) and QSFP+ to 4SFP+ (40G to 10G) breakout cables satisfy customers for various fiber types and reach requirements.

100GbE Fiber Optic Cabling For Future Proofing

With the price of 100G optics cutting down in 2017, 100GbE network is no longer out of customers’ reach. Telecom giants like Cisco, Arista, HPE launches series of 100G optical switches to meet the market demand. And for other 100G components like 100G optical transceivers, fiber patch cables, racks & enclosures, etc, those are ubiquitous on the market.

100G optical transceivers including the CFP, CFP2, CFP4, CXP and the most popular 100G QSFP28 optics in IEEE standards provide a great selection to the overall users.For 100G inter-rack connections, QSFP28 to QSF28 Direct Attach Copper (DAC) Cables and Active Optical Cables (AOC) as well as the QSFP28 to SFP28 breakout cables are the cost-effective solutions.

Path 1: 10G to 40G to 100G

Many of the largest data centers has already moved to 40GbE, which are constructed out of 4 parallel SerDes 10Gb/s links between the Ethernet chip and the QSFP pluggable. The short-reach QSFP interfaces (QSFP+ SR4 modules) use 4 pair of fiber between them, and the copper Direct Attach Cable (DAC) equivalent carry the same on several copper cables inside the big cable. Longer-reach QSFP interfaces (QSFP+ LR4 optics) put the 4 10Gb/t streams onto separate Wave Division Multiplexing (WDM) waves which can be carried over a single pair of fiber. This is part of the reason why QSFP optics are fairly expensive still, especially for longer distances.10GbE to 40GbE to 100GbESimilarly for the 100GbE interfaces that are available today, these are really constructed out of 10 parallel paths of 10Gb/t streams. 100G SR10 modules is the optical transceiver modules that support 10×10Gb/s modes. But neither the CXP SR10 modules or CFP 100G SR10 optics are not popular on the 100G hardware market owing to their larger form factors. Eventually, they need to utilize the smaller footprint 100G modules—QSFP28 transceiver, which is mentioned above as the optical transceivers that can be used in 25GbE and 100GbE.

Result: Although the migration path from 10G to 40G to 100G requires more ports and increases cost per bit, 40GbE between switches is expected to be remain and will not be affected in the near future.

Path 2: 25G to 100G—The Move From 10 Lanes to 4

The old transition path of Ethernet has increased by 10X in speed like the 10G to 40G to 100G. However, 25 Gbps over a single lane for server makes 100G migration be 4×25Gb/s mode.100-gbe-block-diagramUsing four-lane variants like QSFP28 is more economical in several ways:

  • The single-lane design makes four 25 Gbps lanes transceivers less expensive than ten 10Gbps lanes because the transceiver is simpler and less costly to manufacture.
  • The power required to run SFP28 transceivers is much less than required for a typical 10-lane transceiver, it is the same case as the cooling costs.
  • For fiber connections, moving from 10GbE to 40GbE may require an upgrade to four times the number of fibers (MPO), but a 25GbE connection does not because it is the same as 10GbE (single TX and single RX, not four each for TX and RX).
  • Moving from 10GbE to 40GbE typically requires a forklift upgrade to thicker, more expensive cables, but a 25GbE direct attach copper connection does not.

Result: There are few switches and NIC cards that directly support 25GbE. But the curve for 25GbE won’t fade away, rapid development and pre-standard 25GbE products coming soon!

Conclusion

This article introduces 10G/40G/100G fiber optic cabling, and make a clear comparison between the two paths to 100GbE. Customers prefer 4×25Gbps for the reasons: Less parallel paths, less fibers, less optics, less everything. For those who want to upgrade from 40G to 100G, appreciate the reliable performance of 40G with the potential to run across 2 parallel 25Ghz rather than 4 required today.

Media Types for 10 Gigabit Ethernet

Media Types for 10 Gigabit Ethernet

10 Gigabit Ethernet: 10g-switchesToday, the Ethernet cabling systems market is dominated by 10G links. Several physical media types are available to support 10 Gigabit Ethernet applications, including 10GBASE-CX4, SFP+ DAC, fiber optics, and 10GBASE-T (using twisted pair, RJ-45 connectors). In this post, they are introduced one by one.

10GBASE-CX4

CX4 is the first 10G copper standard published by 802.3 (as 802.3ak-2004). It offers the advantages of low power, low cost and low latency and is specified to work up to a distance of 15 meters. But it is not widely used today due to its large cable and inability to run at longer distances.

10GBASE-SFP+ Direct Attach

10GBASE-SFP+ direct attach refers to the small Form-factor Pluggable (SFP+) passive assemblies that use low power and have smaller cables than 10GBASE-CX4. They are widely used in today’s 10G network interconnection. However, from a structured cabling viewpoint, they have a limited distance (up to 10m), are not interoperable between equipment vendors, and are not as flexible or cost effective as 10GBASE-T.

Fiber Optics

Fiber optics options such as 10GBASE-SR (short range SFP+ Fiber) offers the benefit of working over longer distances and low power, as well as providing a smaller cable profile. But the high costs of fiber active equipment makes it an option that not all end users can justify for applications up to 100 meters. As an alternative to 10GBASE-SR, SFP+ AOC (Active Optical Cable) seems like a new and more cost-effective choice for users.

Note: 10G transceivers also include 10G X2, 10G XFP, and 10G XENPAK. But the most commonly used form factor for today’s 10 Gigabit Ethernet network is SFP+.

IEEE 802.3an (10GBASE-T)

10GBASE-T standard defines connections over twisted pair cable, using RJ-45 connectors. CAT 6 is can carry 10GBASE-T for shorter distances when qualified according to the guidelines in ISO TR 24750 or TIA-155-A. However, it is not recommended for new installations intended to support 10GBASE-T networks due to performance issues. CAT 6A technology can deliver 10G performance up to 500 MHz and at a distance of up to 100 meters. When comparing costs per port of equipment, maintenance, and assembly of 10 Gigabit Ethernet, it is significantly more cost-effective than other technologies. It is considered as the preferred cabling option for 10GBASE-T due to its flexibility, low cost, reach, and backwards compatibility.

The above contents introduced the 10 Gigabit Ethernet fiber optic cable and 10G Ethernet transceivers. Which do you prefer? And how about the options beyond 10G? Please stay tuned for more updates about these topics.

Buyer Guides: FS.COM offers a full range of fiber and copper products for 10 Gigabit Ethernet including SFP+ DAC, 10G SFP+ transceivers, SFP+ AOC, CAT 6A cables, and so on. If you have demands on them, you can purchase the most cost-effective products from FS.COM or contact sales@fs.com for details.
Copper Cabling Choices for 10G Ethernet

Copper Cabling Choices for 10G Ethernet

Nowadays, though fiber becomes popular with the benefit of delivering the lowest latency, many IT departments still use copper cabling for switch-to-switch or switch-to-server connections in 10G Ethernet (10 GbE). Currently, there are two major copper cabling technologies applied for 10 GbE over copper. What are they?

As the following table shown, one of the 10GbE copper cabling options is the SFP+ Direct Attach Copper Cable (DAC Cable), and the other one is the 10GBASE-T with twisted-pair Ethernet cable. In the following section, they are respectively introduced in details.

Media Copper Cable Distance (max) Averange Latency Standard
SFP+ DAC Twin-ax copper SFP+ Cu 10 m (33 ft) 0.1 μs MSA SFF-8431
10GBASE-T Twin-pair Category 6 RJ45 30 m (98 ft)-50 m(164 ft) >1.5 μs IEEE 802.3an-2006
Twin-pair Category 6A RJ45 100 m (328 ft) >1 μs
Twin-pair Category 1 GG45 100 m (328 ft) >1 μs

 

SFP+ DAC Is Convenient for Short Runs

SFP+ is the form factor for 10GbE optical transceivers. 10G SFP+ DAC cable is designed with SFP+ connectors on two ends of a twin-ax copper cable. Because of its low latency, small form factor, and reasonable cost, it has become the preferred and convenient 10GbE copper cabling option for servers and storage devices in a rack, usually within 10 meters (33 ft). 10G SFP+ DAC can help simplify rack cabling and termination. For instance, in a ToR (Top of Rack) architecture, each server and network storage device can be directly connected to the ToR switch without the need for intermediate patch panels. More over, 10G SFP+ DAC is flexible enough for vertical cabling management within the rack architecture. The only cabling outside the rack is the ToR switch uplink connection to the aggregation layer, making moving racks easy.

10G Ethernet copper cabling option: 10G SFP+ DAC

10GBASE-T With CAT 6A Cable Is Perfect for Long-Term Plan

10GBASE-T was released by IEEE 802.3an in 2006 which specifies 10Gbps data transmission over four-pair copper cabling. 10 GbE over CAT 6A cable and CAT 7 cable that use RJ45 connectors can reach up to 100 meters. Among them, CAT 6A is currently the preferred cable option for future-proofing cabling installations and widely used in 10GbE copper cabling. In addition, when using CAT 6A in 10 GbE systems, choosing the right cable type, namely unshielded or shielded (e.g. UTP or F/UTP) is also necessary. In general applications, the UTP cabling is commonly used. But for long-term consideration and noisy environment, the F/UTP may be better.

10G Ethernet copper cabling option: 10GBASE-T CAT 6A

Conclusion

10G Ethernet over copper still plays an important role in the data center switch/server interconnection. Current copper cabling choices for 10G Ethernet include SFP+ DAC and 10GBASE-T with Ethernet cables. Each of them has its own advantage. The SFP+ DAC is convenient for short runs while the 10GBASE-T solution is perfect for long-term plan in a longer distance. Only the right one is the best. So, how do you plan your copper cabling for 10 GbE?

Direct Uplink vs Fan-Out Uplink

Direct Uplink vs Fan-Out Uplink

40G-4X10GFan-out cable is considered as one of the the latest enabling technologies to help increase port densities and lower costs. Taking one (large bandwidth) physical interface and breaking it out into several (smaller bandwidth) interfaces, it is now recommended to use for network migration. In this post, I will compare the 40G direct uplink and the fan-out 4x10G uplink configurations and their inherent difference in maximum server scalability.

For a leaf/spine network architecture, the number of connections used for uplinks from each leaf switch determines the total number of spine switches in the design. Meanwhile, the number of ports on each spine switch determines the total number of leaf switches.

Now, image that we will build a network that supports 1200 10G servers in one fabric with 2.5:1 oversubscription. The network must seamlessly expand to over 5000 10G servers without increasing latency or oversubscription in the future. Next, I will compare two different approaches.

40G Leaf/Spine Fabric – With 40G QSFP Uplink

The following picture shows us a 40G fabric to us. Each of the spine switches has 32 ports of 40G. Each leaf ToR (Top of Rack) switch is connected to the spine with 4 ports of 40G using the leaf switches 40G QSFP uplink ports. There will be 40 servers per rack connected to each leaf switch. Namely, it allows a maximum is 1280 x 10G servers at 2.5:1 oversubscription. This meets the initial scale target of 1200 servers however, it cannot scale larger. Before the network can achieve the 5000 server design goal, the 40G design will have to be re-architected.

40G Fabric

10G Leaf/Spine Fabric

As mentioned above, a direct 40G uplink is not an ideal configuration for a cost-effective server scalability. So, what level of server scalability can be achieved when using fan-out the leaf switch uplink ports with 4 x 10 G each?

The picture below shows a 10G leaf/spine fabric to us. With a QSFP+ to SFP+ optical fan-out cable, each QSFP leaf port is now fanned out to 4 x 10G interfaces each, for a total of 16 x 10G uplinks. Each of the spine switches now has 128 ports of 10G. Each leaf switch is connected to the spine with 16 ports of 10G. In this case, the maximum scale is 5120 x 10G servers at 2.5:1 oversubscription. Obviously, with the same bandwidth, latency and oversubscription, this fabric is better. It can not only be built to 1200 servers for present demand but also can seamlessly scale to over 5000 servers in the future.

10G Fabric

In a word, the 10G leaf/spine fabric design offers 4X greater scalability compared to the 40G fabric design with same hardware, bandwidth, latency and oversubscription. These two configuration scenarios show us how fan-out technology may be used to scale up a data center fabric. As the 10G network is widely deployed in today’s data center, but high speed demand such 40/100G is needed. The fan-out technology not only helps to enables new levels of server scalability, but also helps save time and cost for network migration.