Now that we’ve talked about twisted pair, let’s move into something a little more advanced: transceivers.
Transceivers are modules that plug into slots on switches to allow for different forms of connectivity. These are colloquially termed Small Form Pluggable devices (SFPs) and SFP slots, though that name isn’t one hundred percent accurate because there are several versions of this standard. Typically, almost all modern transceivers are implemented using the physical SFP/QSFP standard, which is the successor to the old Gigabit Interface Converter (GBIC) and does essentially the same thing, just in a different form factor. Some people still refer to these module-based transceivers as GBIC as well, even though GBICs haven’t been on new equipment for years. SFPs were created to be a universal slot for various different needs, and as such can be harnessed for a plethora of different speeds and types of connections (with limitations).
When selecting transceivers, it’s important to know that not all slots are created equal. A slot that looks like a traditional SFP slot can be any one of four different technologies, running at any one of ten different speeds. It’s therefore very important to know what kind of slot the switch has before you try to buy SFPs for it. To start out with, let’s look at the main two types of physical SFP connectors: SFP and QSFP.
SFPs & QSFPs
SFP slots are a square connector with a latch on the bottom to hold the transceiver in place, as seen to the left below. QSFP are Quad Small Form-factor Pluggable and are essentially the equivalency of four SFPs in a single slot. They appear below on the right, and you can see they just look like a stretched out SFP slot.
The corresponding transceivers that go in these slots are shown below, a QSFP on the left and SFP on the right.
While that seems simple, it gets very complicated from here on. The SFP specification has had several additions over the years and the different specifications can run at different speeds, so let’s jump into these.
SFP, SFP+, SFP28
These transceivers are the most common in modern networking environments, but there are several other spin-offs, such as XFP, so it’s still worth double checking your ports. Within this one physical form factor, there are three main sub-categories, SFP, SFP+ and SFP28.
SFP is rated up to 1.25 Gbits, though it’s implemented as 1 Gbit for ethernet. It has a plethora of different connectivity options for different types of fiber of different sizes and different lengths. The size is measured in nanometers, measuring the diameter of fiber, which needs to match between the SFP and the cable.
- SFP cannot, by specification, use Direct Attached Cables (DACs), which we’ll talk about in a later post.
- Multimode (850 nanometers) is abbreviated as SX and is used for most campus and datacenter connections under 550 meters. It normally has a black extraction lever.
- Single-mode is used for longer distances and normally has a blue extraction lever. There are several specific iterations of single mode connectivity on a couple different sized fibers and for different uses. The most common is the LX standard which is 1310 nanometers up to 10 kilometers. The EX standard can go up to 40 kilometers.
- There are other single-mode standards for further distances and other fancy configurations like dense wave division multiplexing (DWDM) but normally those are relegated to internet service providers.
- SFP can also be converted to 1 Gbit BASE-T (RJ45) with a transceiver that looks something like this:
Next up is SFP+, which is rated up to 16 Gbit. SFP+ in most implementations can run at either 1 Gbit or 10 Gbit in speed.
- It can be converted to 1 Gbit BASE-T (RJ45) with the above transceiver.
- It CANNOT, by specification, run at less than 1 (one) Gbit.
- It CANNOT, by specification, be converted to 10 Gbit BASE-T. The specification doesn’t allow this and there are not transceivers for it. The main reason for this is that running 10 Gbit on BASE-T requires a significant amount of power. More power than normal SFP slots are designed for.
- SFP+ also introduces us to support for DACs.
- Multimode is still the most common fiber type, and is implemented with the standard called SR. The distance limitation of this is very dependent on the type of fiber plugged into it.
- Single-mode is common for longer distances and the LR standard can reach up to 10 kilometers. The less common ER standard can reach up to 40 kilometers.
- Like normal SFPs, there are various other standards for DWDM and longer distances, but they are, again, normally relegated to internet service providers.
- SFP+ can also run Fibre Channel at speeds of 2, 4, 8, and 16 Gbits. The reason those speeds follow Base 2 instead of Base 10 like most networking speeds has to do with how data is encoded on them. You can read more here.
Finally we reach SFP28 which, as the name implies, is rated up to 28 Gbits. It is implemented as 25 Gbit in ethernet and is quickly taking over for 10 Gbit, mostly due to the more logical way that it scales up with multiple ports. We’ll discuss that later with QSFP. 25-Gbit ethernet can also run at 10 Gbit in most implementations.
Let’s address the elephant in the room first: why in the world is it 28 Gbit instead of 25 Gbit? The desired outcome for these SFPs was to be able to push 25 Gbits of data. Due to encoding overhead, the link was designed larger (28 Gbit) to be able to support 25 Gbits of data. This is similar to the way that SFP+ is actually capable of up to 16 Gbits, but is implemented for ethernet as 10 Gbit, or SFP is capable of 1.25 but is rated for 1 Gbit.
- It cannot be converted to 1 Gbit BASE-T (RJ45) as it cannot run at 1 Gbit (and remember, there aren’t 10-Gbit BASE-T transceivers).
- SFP28 also has its own DACs rated for its speed.
- Multimode is still the most common fiber type and is implemented again with the SR standard for up to 100 meters.
- Single mode is again using the LR standard for up to 10 kilometers or the ER standard for up to 40 kilometers.
QSFP, QSFP+, QSFP28
The Q stands for quad. The simple way to look at these is that in the slightly larger form factor there are now four times of whatever form factor follows the Q. For example, QSFP+ is four lanes of SFP+, which are each 10 Gbit, making it 40 Gbit total.
QSFPs didn’t really see much use in the real world. At the time most people simply manually bound multiple 1 Gbit connections together and weren’t worried about a neat form factor that did it for them. You probably won’t ever see one of these, though the format did exist.
This was the first time that the QSFP physical form factor really started showing up. 10-Gbit ports proved to be too expensive to simply aggregate a lot together like 1 Gbit. QSFP+ ports started showing up on switches for uplinks and then for generic server traffic.
- QSFP+ is normally a single 40-Gbit port.
- QSFP+ can have a special cable that takes the single port and splits it into four separate ports, giving four 10-Gbit ports on a single QSFP port.
QSFP+ ports were causing some engineers anguish because they bucked the trend of the last several decades of everything being Base 10. Having a 40-Gbit port was weird. It also made it hard to increment to the next logical speed of 100 Gbit, because it would either be 80 or 120 Gbit depending on if two or three links went together.
QSFP28 is four links of SFP28, or 112 Gbits. This is implemented in ethernet as 100 Gbit ethernet. The SFP28 standard actually came backwards from the 100 Gbit standard, which was attempting to figure out ways of getting 100 Gbit in a single port. As mentioned above, that was proving frustrating with the QSFP+ standard.
- QSFP28 is normally a single 100-Gbit port
- QSFP+ can have a special cable that takes a single port and splits it into four separate ports, giving four 25-Gbit ports on a single QSFP port.
- You cannot break a single QSFP28 port into ten 10-Gbit ports. Remember, it’s four SFP28, not just a generic 100 Gbit.
SFP-DD is a new(er) standard for SFPs and QSFPs. It is the same physical form factor as SFP and QSFP, but it has an additional line of contacts. This port is made to be backwards compatible with normal SFP28/QSFP28 and allows for double the bandwidth. In other words, a normal SFP28-DD port can do 50 Gbits and a QSFP28-DD can do 200 Gbits.
Whew, we made it through all the slots. That was a lot of effort. Now we’re done, right? Well, not quite. All those SFPs need something to plug into them, that’s where we get into fiber cables and things called DACs. Check back with us next week where we’ll break down what you need to know about Direct Attached Cables!