Systemd in regards to Linux distributions' size & performance?

reunite_pangaea · 2019-07-14 23:12:10

I've noticed that over the past decade or so that Linux distributions in general seem to require more computational resources than before. Additionally ISO image sizes seem to have gone up quite a bit as well (mainline Linux Mint was a ~950GB image in 2009, now it's almost twice that size).

I know correlation doesn't imply causation and that other components have probably gotten larger too, but I can't help but wonder if Systemd has played some role in this too.

Panopticon · 2019-07-15 13:02:33

"Moore’s law, prediction made by American engineer Gordon Moore in 1965 that the number of transistors per silicon chip doubles every year.".

https://www.britannica.com/technology/Moores-law

As computer components get ever more complex in nature, software has to keep up with it, im no expert but i believe there is probably a right way and a wrong way, currently there is possibly too many ways if you ask systemd advocates?

Last edited by Panopticon (2019-07-15 13:03:28)

Nili · 2019-07-15 15:02:26

Nowdays even the base install ISO's are getting fatter by Debian/Linux or Apps unfortunately.
However, there are still minimalistic ISO for testing purpose.

Last edited by Nili (2019-07-15 15:03:43)

reunite_pangaea · 2019-07-15 18:28:38

Panopticon wrote:

As computer components get ever more complex in nature, software has to keep up with it, im no expert but i believe there is probably a right way and a wrong way, currently there is possibly too many ways if you ask systemd advocates?

Complexity seems to be a bit vague of a word here.

The physical chip has more components on it, yes. But the x86 architecture is still the x86 architecture, I'd imagine that the more thorny details are abstracted away by chip firmware. But I wouldn't know.

Head_on_a_Stick · 2019-07-15 20:19:17

Panopticon wrote:

"Moore’s law, prediction made by American engineer Gordon Moore in 1965 that the number of transistors per silicon chip doubles every year.".

Moore's Law no longer applies, it was broken several years ago.

@OP: I've just installed Debian buster & Devuan beowulf virtually, both with the XFCE desktop and both seem to occupy roughly the same space on the drive. Devuan does use less memory but Debian's disk access is faster (no idea why) and Devuan can't shutdown or restart the system without root privileges (I presume the login session isn't fully functional yet).

It is true that the systemd binary is bigger than sysvinit but it's only a matter of a few megabytes and Debian split out some of the functions into other packages which aren't part of the default install.

MiyoLinux · 2019-07-15 21:49:48

Head_on_a_Stick wrote:

and Devuan can't shutdown or restart the system without root privileges (I presume the login session isn't fully functional yet).

All of my beowulf and ceres installs shut down and restart normally (two beowulfs: one with Openbox and one with LXQt. One ceres with Openbox). I'm using slim on the Openbox installs and sddm on the LXQt install...but then again, they probably work for me, because I'm special and know stuff...secret stuff.

Panopticon · 2019-07-16 00:28:39

Head_on_a_Stick wrote:

Panopticon wrote:
"Moore’s law, prediction made by American engineer Gordon Moore in 1965 that the number of transistors per silicon chip doubles every year.".
Moore's Law no longer applies, it was broken several years ago.
@OP: I've just installed Debian buster & Devuan beowulf virtually, both with the XFCE desktop and both seem to occupy roughly the same space on the drive. Devuan does use less memory but Debian's disk access is faster (no idea why) and Devuan can't shutdown or restart the system without root privileges (I presume the login session isn't fully functional yet).
https://i.postimg.cc/zyThf265/Screenshot-from-2019-07-15-18-33-51.png
It is true that the systemd binary is bigger than sysvinit but it's only a matter of a few megabytes and Debian split out some of the functions into other packages which aren't part of the default install.

Moores Law is dead, long live Moores law!

Its interesting to note whats next in the evolution of the transistor is though, it cant get much smaller anymore but what innovation springs forth from Moores law being broken or dieing is quite interesting.

Ill quote the whole article in below link as that web page locks up for me (must be advertising popup) and i have to use reader view. It is an interesting article worth a read though.

https://interestingengineering.com/no-m … moores-law

There have been advances in alternative models for transistors that have shown promise.
Both multigate and Tri-gate transistor models offer ways to extend Moore’s law for a time and are already being used in many electronic devices.
But all these can do is extend the effective life of Moore’s Law.

interestingengineering.com
No More Transistors: The End of Moore’s Law
John Loeffler
6-8 minutes
In 1965, Gordon Moore proposed that the number of transistors on a silicon chip would double every year. Moore’s Law, as it is now known, proved prophetic about the exponential growth of computing power that made much of the modern world possible.
Starting around 2010, however, Moore’s Law began to break down and many today are asking if our age of unprecedented growth is coming to an end. Gordon Moore is the co-founder of the Intel Corporation and one of the men largely responsible for the computer age.
His work with the silicon transistor began in 1956, when he went to work for the transistor’s inventor William Shockley and he has been inseperable from the transistor ever since. A transistor produces, amplifies, and directs an electrical signal using three leads, a source, a gate, and a drain.
When voltage is applied to the gate lead, an incoming current at the source lead will be allowed to pass through to the drain lead. Take the voltage away from the gate lead and the current cannot pass through. What this does is produce a way to compute logical values, 1 and 0 in computer terms, based on the whether there is voltage applied to the gate and the source leads. Connect the drain lead of a transistor to the source lead or the gate lead of another transistor and suddenly you can start producing incredibly complex logic systems.
Comparable to the neuron of the human brain, this network of transistors is responsible for the functioning of nearly all modern devices, from a digital alarm clock to a supercomputer. And the more transistors you can fit on a chip, the more computationally powerful this network becomes.
So when Moore was asked to submit a paper to the journal Electronics predicting the future of technology, he reviewed the data on Fairchild’s production of silicon chips.
He found that the number of transistors on a silicon chip doubled every year and proposed in his paper in Electronics that this rate of growth would continue, later revising this to a more conservative doubling every 2 years in 1975. While not a law in the mathematical sense, Moore’s Law bore out: about every 18 months, a transistor would be half the size of the current transistor. This meant more transistors could be packed into a chip, which drove the exponential growth of computing power for the next 40 years.
Why is Moore’s Law Breaking Down?
There are three major factors contributing to the slowing rate of growth in processor power, and they’re all related.
First, you have electrical leakage. For decades, as transistors got smaller, they became more energy efficient.
Now, however, they have gotten so small, as small as 10 nanometers, that the channel that carries the electrical current through the transistor cannot always contain it. This generates heat which can wear out the transistors more quickly, making them even more susceptible to leakage.
Heat isn’t just limited to one transistor though.
Billions of transistors leaking can seriously threaten the integrity of the whole chip, so the processor must reduce the amount of voltage it takes in or throttle the number of transistors in use to prevent overheating, limiting the processing power of the chip.
Finally, there is the third strike against Moore’s law: economics.
When the number of transistors doubles, so does the amount of heat they can generate. The cost of cooling large server rooms is getting more and more untenable for many businesses who are the biggest purchasers of the most advanced processing chips. As businesses try to extend the life and performance of their current equipment to save money, chipmakers responsible for fulfilling Moore’s Law bring in less revenue to devote to R&D—which itself is becoming more expensive. Without that extra revenue, it becomes much harder to overcome all of the physical impediments to shrinking the transistors even further. So, it might not be the physical challenges that bring an end to Moore’s Law, but simply the lack of demand for smaller transistors. Chipmakers and manufacturers have known about this challenge to Moore’s Law for at least a decade.
As such, they have been finding ways to continue the growth in computing power without needing to solely rely on smaller transistors every two years. There have been advances in alternative models for transistors that have shown promise. Both multigate and Tri-gate transistor models offer ways to extend Moore’s law for a time and are already being used in many electronic devices. But all these can do is extend the effective life of Moore’s Law.
One of the earliest and most effective approaches to this problem was the adoption of multiprocessor and multicore architectures. If you want more power from a chip that has come to the limits of its capacity, use two or more chips instead of one and you can continue increasing your processing power, though at greater power consumption cost. Multicore systems meanwhile use a processor design that features several execution cores in a single processor.
Each core is less powerful than the previous generation’s single core design, but several smaller chips can be used more concurrently and efficiently and give an effective increase in computing power.
Moore’s Law is Dead. Long Live Moore’s Law!
The end of Moore’s Law as we know it was always inevitable. There is a physical limit to what can fit on a silicon chip once you start working with nanometers. Go any smaller and you start dealing with subatomic particles which immediately puts you in the realm of quantum computing, which is where we’re already headed. One day though, after the transistor get stuck at three atoms and an electron, someone will notice that the computation power of newer forms of transistors are rapidly advancing. Molecular, DNA, or Spintronic transistors will appear to pick up where silicon left off and Moore’s Law will be brought out of retirement until quantum computing makes discussions about limits irrelevant.
Ultimately, this has had less to do with transistors than it has to do with us as a society. Our hope and expectations for progress won’t end with the final generation of silicon transistors because we won’t let it. We will find a way to bring Moore’s Law back for whatever else comes after simply because we want it to be true.

reunite_pangaea · 2019-07-16 22:32:41

Now that I think about it, I really only have the old Linux Mint ISO size to compare against other modern ISO sizes (it was my first introduction to Linux).

Being Ubuntu based it's probably no surprise there's been increased resource usage!

Head_on_a_Stick · 2019-07-17 18:29:55

MiyoLinux wrote:

All of my beowulf and ceres installs shut down and restart normally (two beowulfs: one with Openbox and one with LXQt. One ceres with Openbox). I'm using slim on the Openbox installs and sddm on the LXQt install...but then again, they probably work for me, because I'm special and know stuff...secret stuff.

Let me guess: polkit rules?

My comments were based on the stock XFCE desktop and beowulf hasn't been released yet, I'm sure it will all be working when it is.

MiyoLinux · 2019-07-17 21:28:34

Head_on_a_Stick wrote:

MiyoLinux wrote:
All of my beowulf and ceres installs shut down and restart normally (two beowulfs: one with Openbox and one with LXQt. One ceres with Openbox). I'm using slim on the Openbox installs and sddm on the LXQt install...but then again, they probably work for me, because I'm special and know stuff...secret stuff.
Let me guess: polkit rules?
My comments were based on the stock XFCE desktop and beowulf hasn't been released yet, I'm sure it will all be working when it is.

I'm not telling, because it's a secret.

The officially official Devuan Forum!

#1 2019-07-14 23:12:10

Systemd in regards to Linux distributions' size & performance?

#2 2019-07-15 13:02:33

Re: Systemd in regards to Linux distributions' size & performance?

#3 2019-07-15 15:02:26

Re: Systemd in regards to Linux distributions' size & performance?

#4 2019-07-15 18:28:38

Re: Systemd in regards to Linux distributions' size & performance?

#5 2019-07-15 20:19:17

Re: Systemd in regards to Linux distributions' size & performance?

#6 2019-07-15 21:49:48

Re: Systemd in regards to Linux distributions' size & performance?

#7 2019-07-16 00:28:39

Re: Systemd in regards to Linux distributions' size & performance?

#8 2019-07-16 22:32:41

Re: Systemd in regards to Linux distributions' size & performance?

#9 2019-07-17 18:29:55

Re: Systemd in regards to Linux distributions' size & performance?

#10 2019-07-17 21:28:34

Re: Systemd in regards to Linux distributions' size & performance?

Board footer