For the past 15+ years the majority of PON (Passive Optical Network) equipment sold worldwide has been based on the GPON (Gigabit-capable PON) standard. This is the story of how GPON was created and why it became universally adopted. Even though this is nearly 20 years old, this story is relevant today because it offers a lesson on the importance of industry standards and the importance of the specific standards bodies that create them.
GPON was created by the ITU-T (International Telecommunications Union), which is an agency of the United Nations based in Geneva, Switzerland. Responsibility for drafting Fiber-to-the-Home (FTTH) standards for the ITU is delegated to the Full Service Access Network (FSAN) group. FSAN has a unique structure that has made it very successful. It is composed of two types of members: telecom operators and PON equipment vendors. But these two types of members are not equal. The vendors provide technical expertise and offer advice. But all decisions are made by the operators.
To appreciate how GPON became dominant and why this FSAN structure was important it’s necessary to dig into a few important details that are inherent only to passive optical networks. One involves upstream transmissions (from the ONU’s to the OLT). The ONU laser must be completely off when it is not its turn and must quickly turn back when it is its turn. We call this “bursting”. There is an additional cost for both the ONU, to burst on and off, and the OLT, to reliably detect the bursting data. How much it costs depends on the upstream bit rate and what the state-of-the-are is at any given time. When we were writing the GPON standard, the state of the art was for 1.24 Gbps downstream and 622 Mbps upstream. But we were confident that in 3 to 5 years we could do 2.4 Gbps downstream and 1.2 Gbps upstream. We presented this information to the FSAN operators. While they were reticent to approve two pairs of D/S and U/S speeds, they agreed to include both in the standard. [As an aside, we presented this same information to the IEEE (see below) and they basically said they only work w/ bit rates that are powers of 10.]
GPON was created by the ITU-T (International Telecommunications Union), which is an agency of the United Nations based in Geneva, Switzerland. Responsibility for drafting Fiber-to-the-Home (FTTH) standards for the ITU is delegated to the Full Service Access Network (FSAN) group. FSAN has a unique structure that has made it very successful. It is composed of two types of members: telecom operators and PON equipment vendors. But these two types of members are not equal. The vendors provide technical expertise and offer advice. But all decisions are made by the operators.
To appreciate how GPON became dominant and why this FSAN structure was important it’s necessary to dig into a few important details that are inherent only to passive optical networks. One involves upstream transmissions (from the ONU’s to the OLT). The ONU laser must be completely off when it is not its turn and must quickly turn back when it is its turn. We call this “bursting”. There is an additional cost for both the ONU, to burst on and off, and the OLT, to reliably detect the bursting data. How much it costs depends on the upstream bit rate and what the state-of-the-are is at any given time. When we were writing the GPON standard, the state of the art was for 1.24 Gbps downstream and 622 Mbps upstream. But we were confident that in 3 to 5 years we could do 2.4 Gbps downstream and 1.2 Gbps upstream. We presented this information to the FSAN operators. While they were reticent to approve two pairs of D/S and U/S speeds, they agreed to include both in the standard. [As an aside, we presented this same information to the IEEE (see below) and they basically said they only work w/ bit rates that are powers of 10.]
Another property of PON is that it’s inherently a point-to-multipoint (P2MP) network. This type of network is fundamentally different from the point-to-point or multipoint-to-multipoint networks that we’re used to with Ethernet networks. A P2MP is asymmetric – when the OLT transmits a packet it is received by all the ONUs; but when an ONU transmits a packet it is only received by the OLT. This places certain requirements on all PON protocols. For example, an ONU cannot just transmit upstream anytime it wants because it doesn’t know if another ONU is currently transmitting. Consequently, a PON must use a controller/client model where the OLT controls precisely when and for how long a given ONU can transmit its data. In addition, a PON protocol must encapsulate all the data transported on the PON. By “encapsulate” I simply mean there must be a header on each packet that identifies which ONU is the recipien (in the downstream), or which ONU is the sender (in the upstream).
The predecessor to GPON was called BPON. It used an encapsulation method based on ATM (Asynchronous Transfer Mode). An ATM cell is 53 bytes long and consists of a 5-byte header and a 48-byte payload. This format was ideal for transporting TDM services like voice and T1’s, which fit nicely into 48-byte payloads. But ATM was very inefficient at transporting Ethernet frames because it had to break these long (1500+ Byte) frames into many 48-byte segments.
This inefficiency was hotly debated when FSAN began working on GPON in the Fall of 2001. Some wanted to continue using ATM cells for encapsulating Ethernet frames. It was a well understood algorithm and was the only standardized method that was published at the time. But there were others of us who wanted a new method that could encapsulate Ethernet frames without the segmentation that ATM requires. That would be much more efficient but it meant figuring out a new and untested method. Keep in mind that we knew the IEEE was also working on EPON and if we encountered a long delay figuring out a new encapsulation method it would delay and possibly doom GPON.
Fortunately, there was an ITU group working on this very problem. The method was called Generic Framing Protocol (GFP). It wasn’t a ratified standard yet and it needed some modification for GPON, but it allowed whole Ethernet frames to be encapsulated as is, without segmentation. Over many months the vendors presented our arguments for and against to the FSAN operators. But they delayed making any decision. This gave us enough time to figure out and document how GFP would be adapted to the specific needs of GPON. Ultimately, our proposal was accepted by the operators. We named the new method GPON Encapsulation Mode or GEM and it became a core part of GPON.