| Teraoctets per second (To/s) | Kilooctets per second (ko/s) |
|---|---|
| 1 Teraoctet per second | 1000000000 ko/s |
| 2 Teraoctets per second | 2000000000 ko/s |
| 3 Teraoctets per second | 3000000000 ko/s |
| 4 Teraoctets per second | 4000000000 ko/s |
| 5 Teraoctets per second | 5000000000 ko/s |
| 10 Teraoctets per second | 10000000000 ko/s |
| 20 Teraoctets per second | 20000000000 ko/s |
| 25 Teraoctets per second | 25000000000 ko/s |
| 50 Teraoctets per second | 50000000000 ko/s |
| 100 Teraoctets per second | 100000000000 ko/s |
| Reference | Teraoctets per second (To/s) | Kilooctets per second (ko/s) |
|---|---|---|
| A dial-up modem | 0.000000007 To/s | 7 ko/s |
| Typical home broadband | 0.0000125 To/s | 12500 ko/s |
| Gigabit Ethernet | 0.000125 To/s | 125000 ko/s |
| Streaming a 4K film | 0.000003125 To/s | 3125 ko/s |
The teraoctet per second is a unit of data transfer rate equal to a thousand gigaoctets per second, or eight terabits per second. Its symbol is To/s. It describes memory bandwidth inside the fastest processors and the aggregate throughput of large computing systems, rather than any link between separate machines.
Graphics and accelerator chips are the clearest example. A modern accelerator uses stacked memory placed alongside the processor on the same package, and the bandwidth between them reaches several teraoctets per second. That figure is what allows thousands of arithmetic units to be kept busy at once, and it is now the property that most often determines how fast a machine learning workload runs.
The reason bandwidth rather than arithmetic has become the limiting factor is worth stating. Processors have grown far faster at computing than memory has at supplying data, so a modern chip can perform tens of operations for every octet it reads. Any calculation that touches memory more often than that is limited by the memory, and most real calculations are.
Supercomputer interconnects reach this range in aggregate. The network joining thousands of nodes carries teraoctets per second across the whole machine, though no single link does. The design problem is to arrange the topology so that any node can reach any other quickly, which is why these networks are built as multi-dimensional meshes and trees rather than as simple stars.
To make the number concrete, one teraoctet per second would fill a large consumer hard drive in twenty seconds and transfer the entire text of every book in a national library within a minute. Nothing in ordinary use approaches it, and no external cable of any kind carries it.
The unit also appears in descriptions of storage arrays. A large parallel filesystem, spread across thousands of drives, can deliver a few teraoctets per second in aggregate to a supercomputer, and that figure is quoted as a headline specification because it determines how quickly a simulation can be written out and read back.
One teraoctet per second equals 1,000 gigaoctets per second, 8 terabits per second, or about 0.9095 tebioctets per second.
The kilooctet per second is a unit of data transfer rate equal to one thousand octets per second, and therefore to eight kilobits per second. Its symbol is ko/s. It is the unit in which file transfers were reported through the whole of the dial-up era, and it still appears whenever a transfer is slow enough to need it.
The conversion from advertised connection speeds is the reason it matters. A 56-kilobit modem delivered about 7 kilooctets per second in practice, and users learned the relationship by watching progress bars: a one-megaoctet file took about two and a half minutes. The habit of dividing the advertised number by eight and then subtracting a bit for overhead dates from that period.
Storage devices of the era were similar. A floppy disc drive read at roughly 30 to 60 kilooctets per second, a single-speed compact-disc drive at 150, and an early hard drive at a few hundred. Loading a program from any of them was a matter of seconds to minutes, and software was written with that expectation in mind.
The unit still appears in serial communication. The classic serial port ran at rates up to 115,200 bits per second, which is 14.4 kilooctets per second, and equivalent rates are still used to talk to microcontrollers, scientific instruments and industrial equipment. A protocol designed for such a link cannot assume that a large message will arrive quickly.
It also appears at the bottom end of modern networking. A congested mobile connection, a distant satellite link or a heavily shared wireless network can fall to a few tens of kilooctets per second, and at that rate an ordinary web page — which now runs to a few megaoctets — takes a minute or more to load. The unit is a reminder of what the network assumes about its users.
For scale, one kilooctet per second moves a thousand characters of plain text each second, so a short letter transfers in a second and a novel in about ten minutes. Nothing about text has ever needed more than this; every increase in transfer rate since has been consumed by images, sound and video.
One kilooctet per second equals 1,000 octets per second, 8 kilobits per second, or about 0.9766 kibioctets per second.