Digital Ocean Secretly Wants You to Switch to Larger Droplets?

Rackspace Cloud Servers do the same, their instructions for enabling swap are here: http://www.rackspace.com/knowledge_center/article/create-a-l...

Their explanation for the change is [1]:

Why remove swap?

In a multi-tenant cloud environment certain resources are shared amongst customers, and in the case of swap the key resource affected is “disk IOPS” (IOPS stands for “Input/Output Operations per Second”; literally the number of read/write operations that can be performed on the disk per second). Disk IOPS are consumed whenever an application performs any sort of read or write to the physical hard disks.

If Cloud Servers running on the same physical host are running more processes than their allotted RAM, they will begin to heavily utilize swap. This has the side effect of consuming a large portion of the available disk IOPS pool, thereby causing what we refer to as the “noisy neighbor” effect. Put plainly, this means that other virtual machines can monopolize the disk and affect your performance, much like a “noisy neighbor” in the real world can disrupt your quiet dinner plans.

Prior to the change described above, a separate partition, solely dedicated to swap memory, would be given to each virtual machine by default. In order to provide the best service and consistent performance, as well as aligning to industry standard practice, we have removed this default swap partition. We believe this will lead to better customer experiences across the board.

I've also just added a note about it in a blog post here: http://blog.human-friendly.com/a-week-with-docker

I think for beginners having some swap enabled would be a good thing although there are probably production scenarios where there are risks of massive performance degradation once you start swapping and you may prefer to know that you have hit the wall.

On the SSD backed cloud servers I would hope that swap would perform relatively well compared to spinning rust and if that is how you want to use the i/o that they allocate you it should be fine.

[1] http://www.rackspace.com/knowledge_center/article/changes-to...

Not enabling swap is extremely common in the hosting business. Even in EC2, a lot of the AMIs have no swap. This is to prevent users from thrashing your server hard drives. It's often better to have the customer run out of memory and realize they have gotten too small of a server then for them to start thrashing your HD and then just think your hosting is slow.

I prefer having swap disabled by default and hitting the wall hard when out of memory. Some swap is nice on desktop systems where memory usage can vary a lot. On server, RAM usage should be relatively stable. Tune your stack, pick appropriate plan that leaves some RAM headroom, and be done with it. Ah, and add alerts for when memory usage goes over some threshold.

Like cuu508 said server memory usage should be very stable, if you're running into swap it means you don't have enough memory and you need a bigger droplet.

Except every other provider enables swap by default.

I understand another reason why they don't do it though... SSD has a very low amount of writes available, and if low mem machines are swapping all the time, then the SSD will wear out much more quickly.

Another annoying thing is that they disabled resizing the SSD space used. So now, when people upgrade from say 20GB to 30GB there is no easy way to do it. They did used to do this, but they disabled the easy interface for it. So lots of people upgrading are not getting what they paid for, or expected (easy upgrades of VMs).