Throughput vs DRAM and Flash Specs

Why Throughput Is More Important Than DRAM and Flash Specs

When it comes to designing and managing networks, there is an endless challenge of trying to improve performance. Inevitably, that includes upgrading and replacing hardware. When you research hardware for upgrades and replacements, you’ll see a lot of different tech specs listed in the descriptions.

Two common terms that come up are throughput and bandwidth. You’ll see bandwidth more in advertising spaces, and you’ll see throughput more when you look up real-world tests of the hardware. Why is throughput so important? Why don’t you see more about flash specs?

What Is Throughput?

It might help to clearly define throughput. In networking, this is a term used in the measurement of how much data is transferred through a network node at any given moment. Throughput measurements count the number of packets that are successfully received, and this is most easily expressed in bits (or Megabits in the case of most networking hardware).

Throughput makes a little more sense when you compare it to bandwidth. Bandwidth is the theoretical limit of how much data can pass through a network node at a given moment.

What Does It Matter for Networking?

Thinking in both of these terms, you can reframe the ideas. Bandwidth is a metric that helps you analyze network design. What is the network capable of doing? Bandwidth numbers will tell you that.

Meanwhile, throughput is a measure of actual performance. How well is the network doing its job? For that, throughput gives you a clear number that you can use for analysis. It can help you find disconnects between design elements in your network and their real-time performance. This can improve future network designs and help you find where applications disconnect from design theory.

Why Is It More Important Than Other Specs?

Perhaps the most useful application of throughput is using it to find network problems and bottlenecks. If the network is performing well below its theoretical limits, then something is creating problems with the network.

So, you can look at throughput measurements at each node to find the holdup. This advances network troubleshooting and helps you optimize an existing network without the need for redesign. Instead, you can repair, replace, or otherwise treat the issues that are holding a network back.

While all of this is pretty straightforward, why wouldn’t you also need additional specs to design your network? We can look at two specific cases to paint a clear picture: RAM and CPU.

RAM

RAM, memory, or DRAM, is a component in a device that handles data access for simultaneous processing. Many people call it the computer’s short-term memory, and that’s a fair analogy.

RAM is measured in bytes (which are bigger than bits), and typical RAM modules hold capacities in the Gigabytes.

So, your RAM capacity is roughly indicating how much information the device can process simultaneously. This is a theoretical or design measurement. When you compare the data capacity of RAM to other elements of a network, it’s clear why you don’t have to worry about RAM specs.

If a single RAM module can handle more than 1 Gigabyte of data, then it’s not going to be the source of a bottleneck for a network with a bandwidth measured in Megabits per second. RAM just isn’t the limiting factor.

CPU

CPU paints a similar picture. There are several ways to measure CPU performance. One of the most common is with clock speed. This is another theoretical measurement that describes how quickly a CPU can process information. Modern CPUs provide clock speeds of multiple gigahertz. On top of that, CPUs can normally handle multiple processes at the same time, with each process running at these speeds.

It’s another case where the metric in question is never going to be the limiting factor in a network.

Bottomline

Throughput is a valuable measurement because it’s giving you active data about networking performance, and it’s showing you exactly where to focus your attention if you want to improve networking performance. Theoretical measurements and metrics that are never the source of bottlenecks just aren’t as helpful.

Additional Learning Center Resources