When comparing systemd vs SysV-init, the Slant community recommends SysV-init for most people. In the question“What are the best Linux init systems?” SysV-init is ranked 2nd while systemd is ranked 6th.
Ranked in these QuestionsQuestion Ranking
Pro Login management out of the box
Systemd contains a daemon called logind which is used for managing user logins.
Pro Cgroups processes control
Systemd groups all processes by services using Linux's cgroups. Think about memory/cpu/tasks/IO/Net limits and accounting for any service.
Pro Default for many popular distributions
Systemd is the default init system for most popular Linux distributions (Arch, Debian, Ubuntu, openSUSE, Fedora, etc.) Therefore there is an insane amount of support behind Systemd. Choosing Systemd means running with the herd, which comes with it's pros and few (or none for some people) cons.
Pro Simple to understand
Pro Very stable
Since sysvinit does only one thing (initialize the system) and one thing only, it's very stable and it's impossible for it to fail for any problems unrelated to booting the system.
Pro Better boot time and overall performance
Con UNIX-like isn't the same as the UNIX philosophy
One of the main argument that people who are against using systemd is that it does not follow on of UNIX'S core philosophies. 'Do one thing and do it well', instead systemd represents a collection of dozens of tightly coupled libraries. With responsibilities that exceed those of a simple init system because it also tries to handle things like device management, power management, mount points, cron, encryption, syslog, network configuration etc...
Con Makes dependent products difficult to port
Software dependent on systemd. Becomes difficult to port to systems that lack systemd.
Con Too monolithic
It tries to do too much. I don't think most people who use systemd are even aware of most of the features as they don't really use them. It makes it really complicated to deal with sometimes, and it's possible that in a few years this project will be a nightmare to maintain, and with that the users of it will start to feel the fallout.
Con Binary Logging
A binary log structure means that any tools you want to use to parse it will have to be aware of its format, and know how to deal with the ways it can become corrupted. Otherwise, it's not really possible in the easy sense.
Con It's all or nothing
While it is technically possible to use software without SystemD, it really is true that it is "almost impossible" to use software without it, given that all the hard porting work to other init systems has not been done already for you, or given the fact that you are trying to install unported packages directly from the authors (either from binary, like a DEB file, or from source.) Consider the following:
Every major Linux distribution runs SystemD as an only option for init systems.
Around %95 of GNU/Linux users use SystemD, give or take.
SystemD makes things easier for lazy developers (at a cost, however.)
Therefore, most software packages that depend on an init system are developed with inherent and sole SystemD support, in favor of the status quo. While distributions such as Artix-, Gentoo-, and Void Linux have been able to correct packages that depend on SystemD, it is not the easiest to do so. Developers and users wanting to package their own software or build software from source may give up trying to work without SystemD since the software may need to be edited for extra compatibility.
Con Runs only on Linux
Con Connects to Google DNS servers by default
Con Need glibc
Con Kills background processes after the user logs out by default
Con Overkill as an init system, resulting to unsafe init system
1.2 million code lines...
How this monster is controlled with such a huge size, for security leaks and other aspects?
How about if, the homed new service fails and your home folder may become unreadable or useless?
Con Lead developer shows near complete lack of care for standards of quality needed for developing a part of an OS as integral as the init system is
Con Duplicated implementation for every service
Every init script needs to reinvent the wheel for every script: argument processing, start/stop/restart/reload/status/whatever processing, finding/clearing/creating PID files, sourcing defaults, building and setting configuration options, so on and so forth.
Con Launches a bunch of processes to launch a process
Every init script spawns at least sh/dash/bash, and probably also additional processes such as cat, echo, start-stop-daemon, etc, just to start a single daemon that may not even be needed at the time of boot. This massive overhead results in poor performance, and is a killer for embedded systems.
Con Can lead to slow boot
Since init starts tasks serially, it has to wait for a certain task to finish in order to start the next one. But when startup processes end up I/O blocked, this leads to considerable delays during boot.
Con Initscripts is not portable
It is virtually impossible to write portable sysvinit-scripts
Con Hard to write sysv-initscripts
As one needs programming skills (opposed to a declarative style).