All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Review of Delimiter dedicated HP BL260c E5420 16GB SSD ($30 month) vs VPS SSD 3 from OVH
Hi everyone,
So I bought this dedicated server in Atlanta a few days ago and it was provisioned yesterday.
Because I always search for information here before buying I though I could leave my small contribution.
So I got the server, and tried to reinstall the OS...
I used the control panel to do so and I choose Debian 8 (sda), fill up my root password and didn't create a regular user.
After 4 times reinstalling (it was impossible to connect via ssh using root and the choosen pass) I reinstalled one last time.
This time, I created a regular used as well as the root password.
Again, impossible to login with the root user via ssh but...
I could login with the regular user and then "su root" without problems.
Their control panel says that creating the regular user is optional... so this is something to be aware as I lost quite some time until I thought that this information might have been wrong.
So I installed some packages and this are some results:
First, using Phoronix Test Suite:
Hardware:
2 x Intel Xeon E5420 @ 2.50GHz (8 Cores),
Motherboard: HP ProLiant BL260c G5,
Chipset: Intel 5100 MCH + ICH9R,
Memory: 16 Gb MB DDR2-667MHz,
Disk: 120GB Samsung SSD 850, (on a SATA II port, 3 Gbps)
Graphics: AMD ES1000 128MB,
Network: Broadcom NetXtreme BCM5715S Gigabit
NGINX Benchmark 1.0.11:
pts/nginx-1.1.0
Average: 13729.14 Requests Per Second
PHPBench 0.8.1:
pts/phpbench-1.1.0
Average: 63595 Score
x264 2015-11-02:
pts/x264-2.0.0
Average: 111.14 Frames Per Second
And a network test:
wget freevps.us/downloads/bench.sh -O - -o /dev/null|bash
Benchmark started on Sat Jan 2 12:39:41 GMT 2016
Full benchmark log: /root/bench.log
System Info
-----------
Processor : Intel(R) Xeon(R) CPU E5420 @ 2.50GHz
CPU Cores : 8
Frequency : 2500.030 MHz
Memory : 16081 MB
Swap : 4671 MB
Uptime : 18:09,
OS : Debian GNU/Linux 8
Arch : x86_64 (64 Bit)
Kernel : 3.16.0-4-amd64
Hostname : atl
Speedtest (IPv4 only)
---------------------
Your public IPv4 is 199.233.xxx.xxx
Location Provider Speed
CDN Cachefly 99.2MB/s
Atlanta, GA, US Coloat 92.4MB/s
Dallas, TX, US Softlayer 71.9MB/s
Seattle, WA, US Softlayer 29.7MB/s
San Jose, CA, US Softlayer 28.3MB/s
Washington, DC, US Softlayer 51.7MB/s
Tokyo, Japan Linode 15.1MB/s
Singapore Softlayer 8.09MB/s
Rotterdam, Netherlands id3.net 16.9MB/s
Haarlem, Netherlands Leaseweb 91.4MB/s
Disk Speed
----------
I/O (1st run) : 38.3 MB/s
I/O (2nd run) : 38.2 MB/s
I/O (3rd run) : 38.1 MB/s
Average I/O : 38.2 MB/s
Now, you could enable write back cache (hdparm -W1 /dev/sda) and reach I/O of around 200 MB/s however, this can cause data loss, specially on databases in case of a crash, power failure, etc. (Read: http://www.linux-magazine.com/Online/Features/Tune-Your-Hard-Disk-with-hdparm)
And unixbench (with write back cache enabled):
Benchmark Run: Sat Jan 02 2016 19:38:58 - 20:07:23
8 CPUs in system; running 1 parallel copy of tests
Dhrystone 2 using register variables 23883541.6 lps (10.0 s, 7 samples)
Double-Precision Whetstone 2638.9 MWIPS (9.9 s, 7 samples)
Execl Throughput 2226.4 lps (29.9 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks 686142.5 KBps (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks 198858.7 KBps (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks 976378.0 KBps (30.0 s, 2 samples)
Pipe Throughput 1371526.3 lps (10.0 s, 7 samples)
Pipe-based Context Switching 137746.2 lps (10.0 s, 7 samples)
Process Creation 4429.9 lps (30.0 s, 2 samples)
Shell Scripts (1 concurrent) 5092.4 lpm (60.0 s, 2 samples)
Shell Scripts (8 concurrent) 2203.5 lpm (60.0 s, 2 samples)
System Call Overhead 1897195.0 lps (10.0 s, 7 samples)
System Benchmarks Index Values BASELINE RESULT INDEX
Dhrystone 2 using register variables 116700.0 23883541.6 2046.6
Double-Precision Whetstone 55.0 2638.9 479.8
Execl Throughput 43.0 2226.4 517.8
File Copy 1024 bufsize 2000 maxblocks 3960.0 686142.5 1732.7
File Copy 256 bufsize 500 maxblocks 1655.0 198858.7 1201.6
File Copy 4096 bufsize 8000 maxblocks 5800.0 976378.0 1683.4
Pipe Throughput 12440.0 1371526.3 1102.5
Pipe-based Context Switching 4000.0 137746.2 344.4
Process Creation 126.0 4429.9 351.6
Shell Scripts (1 concurrent) 42.4 5092.4 1201.0
Shell Scripts (8 concurrent) 6.0 2203.5 3672.5
System Call Overhead 15000.0 1897195.0 1264.8
========
System Benchmarks Index Score 1023.9
------------------------------------------------------------------------
Benchmark Run: Sat Jan 02 2016 20:07:23 - 20:35:58
8 CPUs in system; running 8 parallel copies of tests
Dhrystone 2 using register variables 186367148.7 lps (10.0 s, 7 samples)
Double-Precision Whetstone 21101.8 MWIPS (9.9 s, 7 samples)
Execl Throughput 14247.4 lps (29.9 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks 411832.1 KBps (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks 111876.4 KBps (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks 1041605.9 KBps (30.0 s, 2 samples)
Pipe Throughput 10809656.9 lps (10.0 s, 7 samples)
Pipe-based Context Switching 2285579.7 lps (10.0 s, 7 samples)
Process Creation 29502.2 lps (30.0 s, 2 samples)
Shell Scripts (1 concurrent) 23905.4 lpm (60.0 s, 2 samples)
Shell Scripts (8 concurrent) 3310.2 lpm (60.1 s, 2 samples)
System Call Overhead 1963023.9 lps (10.0 s, 7 samples)
System Benchmarks Index Values BASELINE RESULT INDEX
Dhrystone 2 using register variables 116700.0 186367148.7 15969.8
Double-Precision Whetstone 55.0 21101.8 3836.7
Execl Throughput 43.0 14247.4 3313.3
File Copy 1024 bufsize 2000 maxblocks 3960.0 411832.1 1040.0
File Copy 256 bufsize 500 maxblocks 1655.0 111876.4 676.0
File Copy 4096 bufsize 8000 maxblocks 5800.0 1041605.9 1795.9
Pipe Throughput 12440.0 10809656.9 8689.4
Pipe-based Context Switching 4000.0 2285579.7 5713.9
Process Creation 126.0 29502.2 2341.4
Shell Scripts (1 concurrent) 42.4 23905.4 5638.1
Shell Scripts (8 concurrent) 6.0 3310.2 5517.0
System Call Overhead 15000.0 1963023.9 1308.7
========
System Benchmarks Index Score 3213.6
==========
In regard of support, I contacted them for several questions and they replied quickly (within 1 to 2 hours).
The only thing I wasn't so happy with, was when today I asked for a refund and they told me to read the TOS (no refunds, no trials, etc).
Nothing to complaint here, but be aware that there are no refunds from them if you're not happy.
My verdict:
Overall is an old server, with only SATA II and a good network attached to it.
Their network although not top of the line, is great for the value.
If you're not planning to run databases or other data loss sensitive apps there, you could enable write back cache and enjoy higher disk write speed.
Comparing this with a VPS SSD 3 that I have with OVH (11.99 eur) it's very disapointing.
While the network (1 Gbit) with Delimiter is so much better than the VPS on OVH, the 8 core dedicated server performs roughly 40% worst than my 2 core VPS, except for x264 encoding (because there are more cores to use).
However, the dedicated server costs $30 while the OVH VPS is less than $15...
Considering cost per value, I would say the OVH VPS wins big time.
If you need a regular web server, go with the OVH VPS.
If you need a seedbox (not allowed), convert x264 files or to do some network intensive tasks (backups, storage, etc) then go with the dedicated, however you could go with other european providers such as Hetzner, Onine, etc for a faster server for roughly the same price.
Below you can see the same benchmark on the VPS SSD 3 from OVH.
As you might notice, unixbench is higher on the dedicated but I care more for nginx and php benchmarks, because I need a web server and not just something with a better unixbench.
==========
First, using Phoronix Test Suite:
Hardware:
Processor: 2 x Intel Xeon E312xx (Sandy Bridge) @ 2.39GHz (2 Cores),
Motherboard: OpenStack Foundation Nova v2014.2.3,
Memory: 1 x 8000 MB RAM,
Disk: 40GB SSD with Local RAID 10
Network: 100 Mbit
NGINX Benchmark 1.0.11:
pts/nginx-1.1.0
Average: 21522.91 Requests Per Second
PHPBench 0.8.1:
pts/phpbench-1.1.0
Average: 101637 Score
x264 2015-11-02:
pts/x264-2.0.0
Average: 50.99 Frames Per Second
And a network test:
wget freevps.us/downloads/bench.sh -O - -o /dev/null|bash
Benchmark started on Sat Jan 2 13:07:53 UTC 2016
Full benchmark log: /root/bench.log
System Info
-----------
Processor : Intel Xeon E312xx (Sandy Bridge)
CPU Cores : 2
Frequency : 2394.442 MHz
Memory : 7813 MB
Swap : 0 MB
Uptime : 63 days, 23:33,
OS : Debian GNU/Linux 8
Arch : x86_64 (64 Bit)
Kernel : 3.16.0-4-amd64
Hostname : nyc
Speedtest (IPv4 only)
---------------------
Your public IPv4 is 51.254.xxx.xxx
Location Provider Speed
CDN Cachefly 11.4MB/s
Atlanta, GA, US Coloat 4.65MB/s
Dallas, TX, US Softlayer 10.1MB/s
Seattle, WA, US Softlayer 8.80MB/s
San Jose, CA, US Softlayer 3.36MB/s
Washington, DC, US Softlayer 10.5MB/s
Tokyo, Japan Linode 6.29MB/s
Singapore Softlayer 2.02MB/s
Rotterdam, Netherlands id3.net 11.7MB/s
Haarlem, Netherlands Leaseweb 11.9MB/s
Disk Speed
----------
I/O (1st run) : 253 MB/s
I/O (2nd run) : 285 MB/s
I/O (3rd run) : 271 MB/s
Average I/O : 269.667 MB/s
And unixbench:
Benchmark Run: Sat Jan 02 2016 14:43:49 - 15:11:57
2 CPUs in system; running 1 parallel copy of tests
Dhrystone 2 using register variables 28847114.0 lps (10.0 s, 7 samples)
Double-Precision Whetstone 3707.4 MWIPS (9.8 s, 7 samples)
Execl Throughput 1959.1 lps (29.7 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks 959293.8 KBps (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks 257987.4 KBps (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks 1933979.8 KBps (30.0 s, 2 samples)
Pipe Throughput 1924697.8 lps (10.0 s, 7 samples)
Pipe-based Context Switching 348671.4 lps (10.0 s, 7 samples)
Process Creation 4776.6 lps (30.0 s, 2 samples)
Shell Scripts (1 concurrent) 5834.7 lpm (60.0 s, 2 samples)
Shell Scripts (8 concurrent) 1555.7 lpm (60.0 s, 2 samples)
System Call Overhead 3546841.5 lps (10.0 s, 7 samples)
System Benchmarks Index Values BASELINE RESULT INDEX
Dhrystone 2 using register variables 116700.0 28847114.0 2471.9
Double-Precision Whetstone 55.0 3707.4 674.1
Execl Throughput 43.0 1959.1 455.6
File Copy 1024 bufsize 2000 maxblocks 3960.0 959293.8 2422.5
File Copy 256 bufsize 500 maxblocks 1655.0 257987.4 1558.8
File Copy 4096 bufsize 8000 maxblocks 5800.0 1933979.8 3334.4
Pipe Throughput 12440.0 1924697.8 1547.2
Pipe-based Context Switching 4000.0 348671.4 871.7
Process Creation 126.0 4776.6 379.1
Shell Scripts (1 concurrent) 42.4 5834.7 1376.1
Shell Scripts (8 concurrent) 6.0 1555.7 2592.8
System Call Overhead 15000.0 3546841.5 2364.6
========
System Benchmarks Index Score 1363.3
------------------------------------------------------------------------
Benchmark Run: Sat Jan 02 2016 15:11:57 - 15:40:06
2 CPUs in system; running 2 parallel copies of tests
Dhrystone 2 using register variables 57297588.4 lps (10.0 s, 7 samples)
Double-Precision Whetstone 7396.8 MWIPS (9.7 s, 7 samples)
Execl Throughput 9248.8 lps (29.7 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks 1117505.1 KBps (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks 291963.0 KBps (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks 2886097.2 KBps (30.0 s, 2 samples)
Pipe Throughput 3816311.4 lps (10.0 s, 7 samples)
Pipe-based Context Switching 677556.0 lps (10.0 s, 7 samples)
Process Creation 16892.9 lps (30.0 s, 2 samples)
Shell Scripts (1 concurrent) 11528.1 lpm (60.0 s, 2 samples)
Shell Scripts (8 concurrent) 1772.8 lpm (60.0 s, 2 samples)
System Call Overhead 5007988.9 lps (10.0 s, 7 samples)
System Benchmarks Index Values BASELINE RESULT INDEX
Dhrystone 2 using register variables 116700.0 57297588.4 4909.8
Double-Precision Whetstone 55.0 7396.8 1344.9
Execl Throughput 43.0 9248.8 2150.9
File Copy 1024 bufsize 2000 maxblocks 3960.0 1117505.1 2822.0
File Copy 256 bufsize 500 maxblocks 1655.0 291963.0 1764.1
File Copy 4096 bufsize 8000 maxblocks 5800.0 2886097.2 4976.0
Pipe Throughput 12440.0 3816311.4 3067.8
Pipe-based Context Switching 4000.0 677556.0 1693.9
Process Creation 126.0 16892.9 1340.7
Shell Scripts (1 concurrent) 42.4 11528.1 2718.9
Shell Scripts (8 concurrent) 6.0 1772.8 2954.6
System Call Overhead 15000.0 5007988.9 3338.7
========
System Benchmarks Index Score 2520.5
EDIT:
Some people commented saying that it's not fair to compare old hardware to new, that it's not fair for the dedicated Sata 2 to be compared to Sata 3 RAID 10, etc...
Well, the purpose of this review was not to compare one against the other, but rather to review this specific Delimiter server.
Later I though, why not run the same thing on my cheap vps and see how does it all stack up?
Some people might disagree, but I care not if it's a old dedicated server vs a new vps, or if one has different specs than the other.
I need a stable, average LEMP stack (Linux, Nginx, Maria DB, PHP) to run a small site with around 50.000 unique users a day (it's a jobs site).
My options are, to continue using a virtual host on my i7 3770 in europe (so far so good), or get another second server in USA for this (therefore, my Delimiter choice).
But ultimately for me, the main point is to find out if I can do more with less (money).
Should I stick to a cheap dedicated (old server) or go with a even cheaper, new hardware, stable vps?
Because both choices fit my needs, I care not about comparing old against new hardware as long as I save money and get more performance.
When I run those benchmarks and found out that the VPS is much faster than this dedicated, I thought it's a waste.
Even though it's an old server, it has 8 cores, it's a dedicated server, has a SSD... so they should be at advantage against just two cores on some cheap vps.
In the end, my opinion is more like:
Why should I pay $30 in one place, when I can pay half on another and almost double the performance?
For my workload, it makes perfect sense to compare the two because both systems get the job done.
I don't care about it being fair or not, because I care about the end results for my workload.
If the dedicated server was about the same price, then maybe I would consider keeping that dedicated... but for double the price and less performance I don't think so.
Sure, Delimiter Network of 1 Gbit is much, much better, but when am I going to use that for a normal LEMP server?
And I don't need to blast those cores to 100%, 24h a day so a VPS is still ok.
Might as well, pay less, get more and stick with a stable VPS.
In the end, i care more about finding the best deal (price/quality) for my needs.
Comments
@MarkTuner
You can't compare the two, they're apples and oranges. One is a dedicated server while the other is a virtualized instance running on a slice of resources from a dedicated server.
Well, it's a review of the dedicated server and then, I added the results for a less than half price VPS as a comparison.
They might be apples and oranges, but it matters because people buy a server or a vps to do something with it.
If I can do more with a cheaper VPS (and I don't need the features of the dedicated) that's all it matters to me.
And you can always choose a XEN or KVM VPS and still get dedicated resources on a faster machine cheaper than buying this dedicated server.
You can oversell with kvm or xen (and, in fact, ovh vps are oversold; not only the vps ssd, But the vps cloud, too)
may i request unixbench score? for both.
delimiter is dedicated and your ovh is kvm vps. you should bench it using unixbench, and see whats score of both.
yeah, i just notice that delimiter dont have refund policy on their TOS.
Debian 8 has a different security policy for SSH than older Debian versions.
If you wish to access your server over SSH when you've Debian 8 you have to create a regular user (this does not apply for OpenVZ VPSs because the templates have normal root login enabled). The security policy on SSH is the following: PermitRootLogin is set to "without-password" which means that a login as root over SSH with passwords WILL NOT WORK! You have to setup SSH public keys for root in order to login over SSH as root.
In my humble opinion this is a good thing. Beat a bit sense into peoples heads about securing SSH, especially with things like XOR around and weak passwords that are usually set by the installation systems.
You can change that in /etc/ssh/sshd_conf. Look for "PermitRootLogin" and set it to yes if you really need to login as root over SSH with passwords. I DO NOT RECOMMEND.
And please fix your formatting. Use pre tags for benchmark output and other code.
Yes.
Except at aruba, they've changed their debian 8 templates to "permitrootlogin" to "yes" at installation, iirc you can't even add secondary users.
The default Debian 8 policy applies to default installations from ISOs. Templates are always different because providers modify them or people who made them modified them. I'm not a big fan of templates on KVM/XenHVM/VMWare or hardware. Better to install from ISO.
Aruba is OpenVZ, KVM or what? Afaik if you use KVM templates you can also modify this. Probably just because their panel does not support the second user creation they enabled root login. Not even SolusVM and Virtualizor have that feature for KVM/Xen/OpenVZ. It's too tricky I guess.
VPSBit has this ability when you install/reinstall your server. Sometimes a bit buggy and you've to reinstall to get the user created.
I wish other providers would atleast give this option , without it , one needs to have KVM access to enable this feature .
What matters in the end is , if you are able get your money's worth out of it or not , for me , all OVH vps's have been nothing but a breeze to work on so far ..
Aruba uses vmware on their smart range I'm using. It's not a disaster but I've looked everywhere in their panel how I would login as there was no asking for ssh keys.
I agree that the default without-password is a lot better in deb8 than it's in ubuntu 14.04 and deb7.
delimiter seems to use the same provisioning system because it looks exactly the same with charlie as default secondary user. hostbill iirc
Why do people say this? Of course you can. You can and SHOULD compare the two. Maybe for some needs dedicated adds a couple points in favor, but this whole board is so used to massive overselling without realizing that the world runs on virtualized computing without issue.
+1
You are right (and I'm an OVH happy customer)
That IO in SSD (Samsung SSD 850) O.o
No, its NOC-PS and only handles OS reloads, very simple style bandwidth graphs, and reboot (pdu or ipmi based). NOC-PS then integrates in 2-3 options of billing systems to make it available to a customer in their portal like that.
This is the default mode of operation for Debian, just like Ubuntu 14.04 and above it requires non-password login for root by default.
Thats not correct, you have to enable the drive cache to on any system to get semi decent performance.
Default configuration:
OVH France:
Hetzner Germany:
The only reason its disabled is HP has a dirty habit of sneakily disabling the write cache on non-HP disks. Its to make the inexperienced think HP disks are better performing. If you have any other dedicated servers, you will find this is almost certainly enabled as standard.
On a properly configured Samsung 850 Evo you will get around 200MB/s on those E5420 units. Do remember E5420 came out in 2007, SATA3 standard was released two years later in 2009, so SATA2 was the latest and greatest at the time. You can't expect SATA3 when it hadn't even been ratified.
This is a silly comparison - E3-1230v3 (launch 2013) vs E5420 (launch 2007), there is 6 years difference and countless processor family revisions between them. On baremetal, E3-1230v3 leaves even E5's in the dust.
NVMe SSD RAID vs single SATA-2 SSD again not comparable.
Usually when you compare something they have to be similar items... This VPS is entirely different to the dedicated server for some key reasons... It's an E5420 CPU and is really old(but is still a good CPU to some extent just depends on your usage)
I mean if you were comparing features to them both then I'd agree but you're putting older hardware vs some of the newest hardware in which it just isn't fair for the dedicated server...
As @MarkTurner said, you're also comparing SATA2 to SATA3.
Sometimes a dedicated server isn't needed when you can trust the company enough to make sure your VPS isn't going to be slowed down by noisy neighbours.
The other thing to note is the dedicated server all resources are yours, nobody elses. You get everything quoted. With VPS that's not usually the case.
Well, I disagree because I care more about value for money for my specific workload... but because there are several people commenting the same, I edited my original review (see at the end).
True... but they are not half price this server.
Allow me to disagree. If there is a power failure and you don't have backup power, you need to disable drive cache to avoid data loss. I don't know about Delimiter server, but if they do have backup power to prevent this, the we can enable write cache. (Tha's why, i said that you could get 200 Mb/s if you want to enable it).
Que es gratis pt?
Just my username from a few years ago.
Well I am 100% certain that the host that your OVH VPS is sitting on will have it enabled
Most probably, but they do also have redundant energy backup to prevent power outages and data loss (RAID 10).
What I said is, in that situation (redundant power) you can (and should) enable it.
It's even part of MySQL optimization best practices for InnoDB...
Obviously, a $30 server is not meant for high availability and mission critical data and that's why I said that, once enabled you can reach 200 MB/s on that server (if you want).
But by default it's off on this server, so you have to enable it if you want performance.
That's also part of the review. You have to enable it manually, or you will have poor writing performance with the default settings.
My main point here is, for my needs the VPS performed much better for half price.
If it was obviously the same price or more, the dedicated would be a better choice.
Agree 100%
@gratispt you can't compare delimiter with any other host because you'll be attacked by some members here for saying truth so basically stay away from them as far as you can and enjoy your vps with OVH!
Yes! Delimiter is a scam!
I just a good deal from them and Mark was such a bastard that he was willing to help with a few questions!
OMG!
I'm outta here! I'm outta here! I'm outta here!
Take a break Alex.
And yes I like my ovh ssd vps1 with ovh!
In fact I think they are quite a OK provider (from a few days experience), except for their policy of not refunding users within first few days (no money back guarantee should be a feature).
This server however, was a disappointment. I knew it was not so great, but I was expecting it to be at least on the same level or slightly better than a 2 core cheap vps.
Sure it's better in some things, but not on what matters to me (web server).
Thats why you read the ToS before agreeing them, then you would have known that the company doesn't offer any refunds before parting with your cash!
Some ppl just dont get that there are reasons to pick a dedi over a VPS. Anway, it depends on the use. Also if you application utilizies multiple cores and what-not.
Sure the newer CPU and RAM is faster, but it all depends. I somehow like dedicated way more than virtual boxes as I dont have to care about any other users, shared ports and nobody complains about utilization or heavy IO.