DelimiterVPS 12 Hour Atom Review
Today I saw Delimiter's $10 Atom server deal and since I was in the market for one, I bought one. Here's my 12 hours review, and you'll find out why it won't be longer than that.
I started out by paying the $10, I emailed Steve a few "pre sales" questions and he said he'd give me a 24 hour "refund" period if the server didn't live up to my expectations. He was really nice about the whole thing and even had my server set up within 30 minutes.
I initially installed CentOS 6.4 as thats my "go to" OS. Once I had logged into the server and gotten cozy I decided to do a "sudo yum update", it froze half way through and wouldn't even finish. I thought maybe it was a fluke, these things happen. So I went ahead reinstalled the server with Ubutnu 12.04, it took about an hour in total for it to reinstall. Once it had reinstalled, I logged in and ran "sudo apt-get update" and it finally worked. I installed a few piece of software and ran some benchmarks:
FreeVPS.us, the test froze after Harlem. I re ran it twice and gave it an hour to maybe unfreeze. Here's as far as it went:
[email protected]:/home/minecraft# wget freevps.us/downloads/bench.sh -O - -o
CPU model : Intel(R) Atom(TM) CPU 230 @ 1.60GHz
Number of cores : 2
CPU frequency : 1597.018 MHz
Total amount of ram : 868 MB
Total amount of swap : 89 MB
System uptime : 11:15,
Download speed from CacheFly: 47.6MB/s
Download speed from Coloat, Atlanta GA: 12.6MB/s
Download speed from Softlayer, Dallas, TX: 10.3MB/s
Download speed from Linode, Tokyo, JP: 1.66MB/s
Download speed from i3d.net, NL:
Download speed from Leaseweb, Haarlem, NL: 3.48MB/s
Black's Speedtest completed though:
[email protected]:/home/minecraft# wget dl.getipaddr.net/speedtest.sh 2>/dev/null -O- |
Testing US locations
Speedtest from Los Angeles, CA, USA [ generously donated by http://maximumvps.net ]
on a shared 1 Gbps port
Download Speed: 3.36 MB/sec Upload speed: 3.89 MB/sec
Speedtest from Los Angeles, CA, USA [ generously donated by TeraFire, LLC ] on a shared 1
Download Speed: 2.75 MB/sec Upload speed: 4.20 MB/sec
Speedtest from South Bend, Indiana, USA [ generously donated by NodeBytes ] on a shared
100 Mbps port
Download Speed: 0 MB/sec Upload speed: 0 MB/sec
Speedtest from Chicago, IL, USA [ generously donated by http://vortexservers.com ] on a
shared 1Gbps port
Download Speed: 4.83 MB/sec Upload speed: 8.95 MB/sec
Speedtest from Buffalo, NY, USA on a shared 1 Gbps port
Download Speed: 5.71 MB/sec Upload speed: 10.47 MB/sec
Speedtest from Atlanta, GA, USA [ generously donated by http://hostus.us ] on a shared
Download Speed: 0 MB/sec Upload speed: 8.51 MB/sec
Speedtest from Clifton, NJ, USA [ generously donated by http://dedicatedminds.com ] on a
shared 1Gbps port
Download Speed: 7.55 MB/sec Upload speed: 15.24 MB/sec
Testing EU locations
Speedtest from Tallinn, Estonia on a shared 1 Gbps port
Download Speed: 1.24 MB/sec Upload speed: 1.99 MB/sec
CPU: 2 x Intel(R) Atom(TM) CPU 230 @ 1.60GHz
Time taken to generate PI to 5000 decimal places: 1m40.786s
As for ServerBear, the test freezes at Sydney and doesn't go any further. I've restarted the test twice. I've given up.
The server's network is just too slow for what I need, it doesn't preform to my expectations. If you were running something small or using this server as a test bed, it would be perfect. But, thats just not what I need it for.
I'll answer some of the questions everyone has right now:
"It's a $10 server, you have to high of expectations."
Maybe, but I went into this expecting a slow server and I came out with a slow server. I wasn't shocked at all. For the price, I could pay $3 more and go with someone like Online.net or Digicube and get a half decenet network (at least from my past experience).
"I told you not to go with them."
Of course, but sometimes you have to try things out for yourself to actually get the full affect. I also purchased the server so I could review it and post the specs so others can have an informed opinion before purchasing. I'm not saying they're bad, i'm just showing the facts. It's up to the consumer on how they want to interpret them.
"Look at their past"
I did, and I wasn't surprised really. Everyone has a past, some hide it better than others. No one is perfect in this industry.
Thanks for your review, being a "dedicated" even if is small, i was specting more.
but then is better a VPS on a high node.
by the way, where is Atom located?
No problem, also it's located in Atlanta.
From the looks of it, someone's spare closet
Oh, The ServerBear benchmark finally finished. It crashed the server and only took 3ish hours. But it finished. serverbear.com/benchmark/2013/11/30/zJWB0vj9ZDxDjLc4
That UnixBench looks near to mine dedicated core/thread vps http://pastebin.com/7gChHMDc
In my oppinion it is worth 10USD
Well, I care more about network, a Online.net dedicated server with the same specs and only about $3 more out preforms this server in the network section.
@skybucks100 yes. you are right 100%.
Hmm, looks like I need to remove the server from NodeBytes.
Edit - removed NodeBytes, added a Dallas, TX location.
@skybuck100 network performance is way off. What do your interface stats look like - any errors? I have personally run the Serverbear test on one of these machines and yes the CPU part was slow (as expected - its a slow old processor) but network was spot on and error free.
If you run something like mtr against Google DNS for example 220.127.116.11. Is it running error free or are there dropped packets?
It's not really packet loss. The script that's being used uses a 10 MB file for download and upload so TCP might not have enough time for it to do the "additive increase / multiplicative decrease" algorithm for congestion avoidance.
Try running it with the following command:
Edit: Screw it. I'll make the default setting to select the 100MB file instead. Hopefully the test servers don't go over their bandwidth allowance. If it does I'll just change it back.
@black if you need a VPS for your testing tool in Atlanta - let me know. Obviously pointless for Delimiter's customers because it will just be across the room but for others could be interesting
@black Can you PM me the URL of one of these downloads, I want to test the download on a server there
I'd rather pay $10/Month for a KVM with dedicated resources and Raid-10 Protection. If you need a dedicated CPU, then an Atom is for you. Just go with a kimsufi.ie reseller
@black sent you some results - definitely the 100MB file helps.
@Jeffrey - $10/month buys you the Atom, ISCSI disk hosted on NetApp. Personally I wouldn't trust a little 2.5" single disk with my data. RAID1 local disk would be nicer though but for these small systems, there is very little space to do anything like that.
@skybucks100 - if you have chance, I'd like you to retry your network tests. I've been watching traffic this afternoon coming off those Atom servers after one customer kindly tried to flood another with a full GE of garbage. I made a few tweaks to it, hopefully that will help you and others who had responsiveness issues with SSH.
Let me know how it goes.
Mine is not delivered yet. But, OK, I have to be more patient, it has been only 20 hours when I ordered the box.
I think I just laughed way to hard at that. Also, try and email [email protected]
At @MarkTurner request, updated tests:
And for some reason Tallinn, Estonia won't finish.
Are the speeds better? Of course, but here's the problem.. I live in Seattle, WA and the average speed it was giving is only 4.45MB/s. I need all the speed I can get, here are a few examples of what I mean:
The Delimiter Atom:
A $1 VPS in Germany
I don't have a Online.net dedicated server anymore so I can't run the test, but a Germany server is still quite a distance away from Seattle and yet it gets nearly twice as fast speeds. Will it always be that fast? Of course not, it's just an analogy of sorts. If the speeds Delimiters work for you, great. They just don't work for what I require.
Side note, if there are any other benchmarks/what have you that'd you like me try just ask.
Just delivered! I'll begin testing, let all you know soon!
@skybucks100 - 31ms from Seattle to Germany? How does that work?
How are you running these speed tests because it absolutely does not make sense given that unrealistic latency.
Transatlantic hop alone is going to be 70-120ms depending on network.
I have no idea quite honestly, but I can assure you the server is in Germany. Would you like the IP to test it?
@skybucks100 But that test is definitely not transatlantic. I'd be budgeting 70-120ms for the Atlantic hop. Germany to London, Paris, Amsterdam - 10-20ms; New York to Seattle (assuming that path) 50ms
So the hop is 130ms at a minimum.
How did you run this test? Which site and how did you set up the end points.
I just used https://github.com/sivel/speedtest-cli and ran
./speedtest-cli --share --server 4031on both.
That's NOT possible.
All to Server 4031
all of these areas you can get fairly cheap servers
Dear god, I have discovered a real internet worm hole... I don't think there is any other better explanation...
Hmm, but I thought Spirit said these were impossible..? This is strange...
31ms latency from Seattle to Germany? Maybe 130 ms and even this I highly doubt. Latency between central Europe and western US is usually around 180ms.
Internet Wormhole, admit it
Do the physics - (check my figures - its late here)
Speed of light = 299,792,458 metres/second = 299,792.458 metres/millisecond
Distance (as crow flies) from Seattle to Frankfurt = 5,080 miles = 8175.468km
8175.468 x 1000 = 8,175,468m
8175468/299792 = 27ms assuming that its as the crow flies and 100% optical all the way.
We all know that its not as the crow flies + we know there is a significant time when going optical -> electrical -> switching/routing -> electrical -> optical
At 31ms that is literally the speed of light which is impossible on that distance.