Computing units don't have to be confusing

By

Overview

Often programmers will use units in a non standard way and usually you can work out what they meant by context.  Many developers never realise there is inconsistencies is technical material as the difference is small to the amount of error and the difficult of reproducing benchmarks reported.

There are times when the difference is so large it does matter and the writer has confused themselves by a lack of understanding which makes your job of trying to work out what they done harder.

An array of measures


  • b = bit.
  • B = byte or 8 bits or 8b
  • g = gram
  • kb = kilobit or 1000 bits. kilo being the scientific prefix for 1000 like kilometer.
  • kB = kilobyte or 1000 bytes, sometimes used for storage
  • kg = kilogram
  • Kb = Kibit = kibibit or 1024 bytes.
  • KB = KiB = kibibyte or 1024 bytes, sometimes used for memory
  • Mb = megabit = 1000^2 bits or 125 kB. Mb/s is used for network bandwidth.
  • MB = megabyte = 1000^2 bytes, MB/s is used for disk bandwidth.
  • MiB = mebibyte = 1024^2 bytes, often referred to as MB or Megabyte for memory.
  • mb = millibits = 0.001 bits - never use correctly.
  • mB = millibytes = 0.008 bits - could be used for compression, but isn't
  • Gb = gigabit = 1000^3 bits or 125 MB, Gb/s is used for network bandwidth.
  • GB = gigabyte = 1000^3 bytes, used for disk space, GB/s is used for memory bandwidth
  • GiB = gibibyte = 1024^3 bytes, often referred to as GB or Gigabyte for memory.
  • gb = gram-bit (also means to feel itchy)
  • gB = gram-byte (someone's Xbox id)
  • TB = Terabyte = 1000^4 bytes, used for disk space.
  • TiB = Tebibyte = 1024^4 bytes, used for memory but usually written as TB.
  • PB = Petabyte = 1000^5 bytes, used for large storage solutions.

Why does it matter?

Where the confusion often arises is when network bandwidth, which is usually in bits/second is confused with disk or memory bandwidth which is usually in bytes/second. To add to the confusion, disk and memory use different prefixes.  

Q: You might see a HDD which can write at 40 MB/s and a network which is 100 Mb/s.  Which is faster? 

A: The HDD is much faster as it can write 320 Mb/s and the network can only write up to 12.5 MB/s (most likely only gets 11 MB/s with overhead)

The use of 'b' and 'B' is a factor of eight difference.

Also confusing is

... memory and disk sizes.  Memory is measured in Gibibytes (1024^3 bytes) as memory has to be a power of 2 in sizes but most people call it GB or gigabytes.  Disk space is measured in gigabytes (1000^3 bytes) as disk space doesn't have to be a power of two, but mostly because it makes the drives appear almost 8% larger.  This is not such a serious problem and it is small enough difference that most people ignore it.

I remember when 4.0 GB drives were replaced by 4.3 GB drives and was amazed to discover they were exactly the same model, but had just been re-labelled.

Memory bandwidth measures are inconsistent and I suspect most of the time it is GB/s which is reported.  (as it is simpler for  human to understand, but more importantly higher) This means if you can read 1 GB (GiB) of memory in one second, the bandwidth is reported as 1.08 GB/s.

Conclusion

Be aware that marketing material and even other developers will mix their units.  You want to have at least a clear idea in your own mind what is really going on. If you can educate them along the way, that would be an improvement.


Comments

Post a Comment

Popular posts from this blog

Java is Very Fast, If You Don’t Create Many Objects

By
Image
  You still have to watch how many objects you create. This article looks at a benchmark passing events over TCP/IP at 4 billion events per minute using the net.openhft.chronicle.wire.channel package in Chronicle Wire and why we still avoid object allocations..  One of the key optimisations is creating almost no garbage. Allocation is a very cheap operation and collection of very short-lived objects is also very cheap. Does this really make a difference? What difference does one small object per event (44 bytes) make to the performance in a throughput test where GC pauses are amortised? While allocation is as efficient as possible, it doesn’t avoid the memory pressure on the L1/L2 caches of your CPUs and when many cores are busy, they are contending for memory in the shared L3 cache.  Results Benchmark on a Ryzen 5950X with Ubuntu 22.10. JVM Vendor, Version No objects Throughput, Average Latency* One object per event Throughput, Average Latency* Azul Zulu 1.8.0_322 60.6 M event/s, 528
Read more

System wide unique nanosecond timestamps

By
A Unique Identifier can be very useful for tracing. Those ids are even more useful when they contain a high-resolution timestamp.  Not only do they record the time of an event, but if unique can help trace events as they pass through the system. Such unique timestamps however can be expensive depending on how they are implemented.   This post explores a lightweight means of producing a unique, monotonically increasing system-wide nano-second resolution timestamp available in our open-source library. Uses for Unique Identifiers Unique identifiers can be useful to associate with a piece of information so that information can be referred to later unambiguously. This could be an event, a request, an order id, or a customer id. They can naturally be used as a primary key in a database or key/value store to retrieve that information later. One of the challenges of generating these identifiers is avoiding creating duplicates while not having an increasing cost.  You could record every identi
Read more

Comparing Approaches to Durability in Low Latency Messaging Queues

By
Image
  A significant feature of Chronicle Queue Enterprise is support for TCP replication across multiple servers to ensure the high availability of application infrastructure. I generally believe that replicating data to a secondary system is faster than syncing to disk, assuming the round trip network delay wasn’t high due to quality networks and co-located redundant servers. This is the first time I have benchmarked it with a realistic example. Little’s Law and Why Latency Matters In many cases, the assumption is that the latency won't be a problem as long as throughput is high enough. However, latency is often a key factor in why the throughput isn’t high enough. Little’s law states, “ the long-term average number  L  of customers in a  stationary  system is equal to the long-term average effective arrival rate  λ  multiplied by the average time  W  that a customer spends in the system”. In computer terminology, the level of concurrency or parallelism a system has to support must be
Read more