All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Dacentec $20 dedicated server: a first impressions review
MeAtExampleDotCom
Member
Story:
(see "overall" at the bottom for the tl;dr)
I picked up one of the 20$/mo machines (Opteron 1381, 8Gb RAM, 2x2TB) during one of the windows when they were available (there is at least one listed right now if you are looking). I added a third 2TB drive when ordering.
It'll be holding off-site backups and VMs for testing those backups (the backups are restored to these VMs and given cursory tests automatically too to make sure the backups are sound, they also act as pre-prepared fallbacks for the machines being backed up should they or their host fail) and a few other bits.
The machine is intended to replace a Kimsufi physical machine (not one of the ones currently listed: a box with an i3-2130, 8Gb RAM, and 1Tb drive) and a couple of VPSs.
Without the extra drive the cost is about the same as the Kimsufi box, but has four times the storage capacity before RAID gets involved. With the extra drive it still works out cheaper than the hosts it will replace.
Installation:
Free access to a KVM-over-IP for installing the base OS myself exactly to my own liking was very useful. The problems I had were pretty much my own fault as I was rushing in breaks while doing other stuff and not thinking.
Support:
I had a couple of issues due to "no brain" moments on my own part. Their response times were excellent, especially for this price level.
Machine Overall:
Everything performs as I'd expect for that spec so no complaints there.
The CPU is slower (at least per core, though it has four real cores where the i3 has two hyperthreaded ones which might make up some of the difference in some workloads as might the extra on-die cache). I was expecting some difference there as I did the research and the difference isn't nearly enough to make the machine noticeably (or at all in most cases) less suitable for the intended tasks.
Facilities:
Nothing significant is missing from the control panel: you have a simple bandwidth graph, power control, rDNS control, the ability to install standard OS images if you are less fussy than I, and so forth.
Access to an IP KVM is potentially very handy (it was for the initial setup and may well be again if/when I badly cock something up). This is not permanent, so there will be some lead-time if you need to use it beyond the initial machine setup.
Network latency:
Pretty much what I'd expect for the location to/from the sites I tested - so no unexpected problems (playing with virt-manager over SSH from the UK is a bit painful, but that isn't unexpected and I don't really know why I didn't use the command line tools anyway...).
Network Throughput (1):
Testing with a 100MByte random file using a single stream via rsync over SSH I get no worse than I get to/from the Kimsufi box.
Transfers to UK sites (home which is 76mbit/17mbit, a backupsy VPS, and a host in work's DC) were are variable: never below 1.5MByte/sec, usually between 2.5 and 3MByte/sec, sometimes getting up to (after starting slow) and maintaining 7MB.
Between the Kimsufi box and Dacentec I see rates between 7 and 11MByte/sec, the top end there presumably being limited by a 100mbps cap at the Kimsufi end.
Between Dacentec and a VM at Time4VPS (Lithuania based IIRC): T4->D was ~9.5MB/s, D->T4 ~6.0MB/s
Between Dacentec and a VM at OpenVirtuals OV->D is ~12MByte/s and D->OV ~16MByte.
Between backupsy in the US: BUUS->D was 12MByte/s, D->BUUS ~5.5MByte/s
The above are from a fairly small number of runs so not exactly scientific though the tests have been done at different times of the day to allow for variance due to that. I must get around to automating these tests at some point, to make testing a new host quicker (and to run automatic regular tests to monitor the situation).
Network Throughput (2):
Testing with a 100MByte random file transferred via HTTP in both single and multiple stream tests:
Single stream HTTP transfers give pretty much identical results so there seems to be no protocol based trafffic shaping going on. Multi-stream transfers make quite a difference to the UK hosts, for instance 8 streams at once getting a pretty reliable 8MByte/sec to home (about 80% of that line's max) and 16MByte/sec to backupsy's UK DC.
Overall:
Definitely fit for my intended purpose, so transferring real work over to it will commence shortly. Based on these initial impressions the $20 boxes are an absolute bargain unless you need more CPU umpf and/or RAM (there are of course other options if you do need those things) and apparently not "too good to be true" as I had feared it might be.
More reliable single stream throughput to the UK would be nice for my needs but I'd not expect to get better for anything like this price, Kimsufi was not significantly better, and there are workarounds if it is ever an issue for me.
Comments
+1. I have the same exact opinion about dacentec servers. I bought a 20$ with an add-on of a /29 (for just 4$). Server delivered with KVM access thru web interface letting me install whatever I wanted, even if I finally chose their big variety of OS they have in their client's panel (including proxmox, Xen Citrix, Ovirt, freeBSD and windows, with a charge of course).
Fair to very good network connections, panel with reinstalls, reboots, graphs and reverse dns, and very fast response to tickets. For the price, Dacentec is far beyond you can ask.
Your name is probably in your IP whois, right? Might be worth mentioning and that you can ask them to remove it.
@4n0nx Yes, I am aware of this. But I don't mind, even more, that's fine with me, I will use my server for some professional tasks I do. It will not host any illegal or weird stuff there.
o i just wanted to add to the review for whoever doesnt like it so they know and know they can have it removed