Aruba Cloud: Choosing the correct size

Discussion in 'Hosting Talk & Reviews' started by michyprima, Apr 2, 2016.

  1. michyprima

    michyprima New Member

    8
    2
    Apr 2, 2016
    Approaching to the unmanaged world is not always easy, and it's definitely hard, for a newbie, realize how many resources he will need for his setup. We have lot of variables but I'm pretty confident that, after reading this, you will have a better idea of your needs.


    Exactly like a normal PC a VPS has some RAM, bandwidth, CPUs and so on. But how much of them do I need?


    Those are the sizes offered by the company, starting from €1: arubacloud.com


    Let's start from the classical use: the web server. I actually like more nginx than apache, because it uses much less resources, and soon you'll see why. A web server, usually has three components:

    • an HTTP server (nginx in our case)
    • a dynamic pages engine (PHP)
    • the database server (MySQL)

    You can live without the last two if, obviously, you intend to only serve static pages. Yes, that means no wordpress! In this case, the size S will be more than enough because nginx, with those resources, can easily serve hundreds of people simultaneously. If you want or need php and sql, then it will be a pretty different story. You will need to choose your size on how much traffic you expect to have. You will need 200MB of RAM for MySQL, another hundred for nginx, and half hundred for each PHP process. For a website with a few thousands requests per day, the S will still be enough. From this point on, however, you should monitor your average CPU usage; if this is approaching or already is 100%, you will need to choose an L to get another core, otherwise an M will do the trick.


    Let's now talk about mail servers. A mail server usually haves:

    • an SMTP server (postfix)
    • an IMAP server (POP3? nah...)
    • a daemon scanning for spam and viruses
    • eventually, a MySQL database storing users and other data

    I can already say you will need lots of RAM for this. My tests revealed that a VPS with postfix, dovecot and amavis, has more or less an hundred MB of free RAM from a total of 1024. The virus scanner uses lots of ram and, avoiding it, you can save a lot. Optimizing your setup you could be able to also fit MySQL on an S, but I discourage doing so. If you need a webserver and a mailserver, my advice is to get two Ss, so you can split the load between the two, and keep MySQL with the webserver, which is obviously a more efficient setup. This kind of setup can tolerate 25-50 users simultaneusly.


    For both use-cases, however, the XL is a nice choice if you have a business to build up on webservices like small hosting or mail hosting or you think you will have lots of active users. It's very hard to spend all the resources this size offers, especially if your setup is correctly done.
     
  2. drmike

    drmike 100% Tier-1 Gogent

    8,573
    2,709
    May 13, 2013
    So first post here is a thinly veiled ad dressed up as a hypothetical estimator of needs?


    Not the best way to start here.
     
    fm7, VyprNetworks and Jive like this.
  3. michyprima

    michyprima New Member

    8
    2
    Apr 2, 2016
    While I may see your point, I don't think I actually did something so bad as you picture it. There still is a considerable amount of informations in the first post; someone starting from the ground obviously has no idea how much resources needs a daemon or another and the estimator I wrote is obviously valid for every provider out there.
     
    Last edited by a moderator: Apr 3, 2016
  4. drmike

    drmike 100% Tier-1 Gogent

    8,573
    2,709
    May 13, 2013
    The advertorial is a mess. Ignoring language issues, there are references to SM, L but these remain undefined variables in your programming.

    Telling someone or inferring that package S will suffice to serve hundreds of 'people' at the same time is suicide.  If we sustain say 150 users and burst to 200+ during whole of day we get:


    150 requests second x 60 seconds = 9,000 requests a minute  x 60 minutes = 540,000 requests an hour  x 24 hours = 12,960,000 requests per day


    Doable?  Sure.  Wise, not in the least bit.  


    Remember there is overhead and piling on connections.  Not everyone is moving data at full wire speed errorless.  No mention of the data size either.  Serving a million small elements has problems, serving 500k in size images has others and both require optimization.  One size shoes fit hardly anyone.


    If you are going to claim things like this, then model it and post actual test results and means to replicate it.  That said I was pushing far more than that 2 decades on very lowly hardware and on less mature server software, but it wasn't virtualized. Doing yourself a disservice and customers who believe such FUD.


    All this effort to push a 1 euro a month VPS?


    Linux



    1 Core
    Intel Xeon E5 v3



    1 GB
    RAM


    20 GB
    SSD Storage


    2 TB/month
    transfer


    logo-wmware.gif
    powered


    1,00 €/month


    What would be more compelling, and interesting here with our audience is talking about your virtualization choice, locations, certifications, etc.  We are a more mature and knowledgeable readership here.


    Also, can the rest of us not buy from Aruba?  Sounds like it:

    ^^^ Strange way of saying you offer services in those countries - as in you have servers in those countries.
     
    Last edited by a moderator: Apr 3, 2016
  5. drmike

    drmike 100% Tier-1 Gogent

    8,573
    2,709
    May 13, 2013
    ... and while I prefer Nginx, the whole idea that Apache is so bloated on RAM usage vs. Nginx is a bit outdated.


    Today Nginx is probably bloated in comparison to more hungry upstarts that stick to the basics.
     
    fm7 likes this.
  6. fm7

    fm7 Active Member

    163
    61
    Jul 26, 2014
    Someone thinking to order an unmanaged cheap vps/dedicated without knowing how much computer resources are needed ... I could anticipate that this would be just a minor problem.  Ignorance is indeed bliss :)
     
  7. DomainBop

    DomainBop Dormant VPSB Pathogen

    2,258
    2,187
    Oct 11, 2013
    I'll fill in the Aruba blanks.


    The virtualization for their budget Cloud Server Smart line is VMware.  The Cloud Server Pro line is VMware and Hyper-V.


    Company:  Aruba was founded in 1994 and is the largest hosting company in Italy, 5th largest in Europe, 10th largest in the world.  From the Aruba.it about us page: "2.1 million domains, over 7.4 million email accounts, over 4.1 million PEC accounts, 31,000 physical servers and a total of over 4.7 million customers"   Wikipedia lists the revenues as over 100 million in 2010 and employee count of 500.  Company also runs the new .cloud domain registry (ICANN chose Aruba to run .cloud over Amazon and Google who had also put in bids to run  the .cloud registry.  See http://www.thedomains.com/2014/11/18/aruba-it-formally-announces-it-acquired-cloud-beating-out-google-amazon/ ) and it sponsors a motorbike racing team with Ducati (https://www.arubaracing.com/?lang=en-GB).


    Certifications: you name it and their DC's are certified for it (see https://datacenter.aruba.it/en/certification.aspx ).  Their Aruba PEC and Actualis divisions are also accredited by the Italian government to issue digital signatures and smart cards in Italy and offer PEC (Posta Elettronica Certificata aka certified electronic email) services.

    Aruba sells to people in all countries.


      The locations for their cloud servers are United Kingdom (London), Germany (Frankfurt), Italy, France (Paris), and Czech Republic .  They own and built the two datacenters in Arezzo, Italy (IT1 is a Tier IV DC) and Ktiš , Czech Republic (run through their 100% owned Czech brand Forpsi which was founded in 1997).

     
    Last edited by a moderator: Apr 3, 2016
    drmike likes this.
  8. drmike

    drmike 100% Tier-1 Gogent

    8,573
    2,709
    May 13, 2013
    See @DomainBop exactly.  They should hire some competence and clearly have in other disciplines.   Heck you've done a far better job representing them than OP has.


    But for some reason a rep comes here to spam crud straight out of the gate.  Useless crud at that.   Math and measure that just is well, useless.


    Mind you, I was semi familiar with them, just hadn't realized their sanctioned marketing tactics.  
     
    Last edited by a moderator: Apr 30, 2017
    fm7 likes this.
  9. michyprima

    michyprima New Member

    8
    2
    Apr 2, 2016
    I'll leave here some numbers from the full review I wrote, which is in italian and obviously wouldn't have fitted here:


    linux 4.4.2 make time 149m29, command used: make -j2


    vpsbench:

    Code:
    CPU model:  Intel(R) Xeon(R) CPU E5-2650L v3 @ 1.80GHz
    Number of cores: 1
    CPU frequency:  1799.999 MHz
    Total amount of RAM: 993 MB
    Total amount of swap: 951 MB
    System uptime:   1:31,
    I/O speed:  721 MB/s
    Bzip 25MB: 8.36s
    Download 100MB file: 69.8MB/s
    Network speed tests:

    • Europe: 70MB/s
    • USA: 30MB/s
    • Asia: 10MB/s

    (replication urls are in the review)


    Latency tests:

    • 8.8.8.8: 7.774 ms
    • DC1<->DC1: 0.745 ms
    • tim.it: 14.808 ms
    • 108.61.5.91 (USA): 100.448 ms
    • 150.101.135.3 (Australia): 324.774 ms
    • 106.187.96.148 (Giappone): 285.020 ms
    • google.it: 17.327 ms
    • nvidia.com: 172.901 ms

    Let me know if you are interested in having a look in the full review, as I don't want to spam my blog if it's not needed :)


    While for someone experienced like you my write may look useless, for starters I still think it will be different. You don't need any experience to approach VPS machines. It's quite the opposite: practicing on a real VPS, facing real problems, building something satisfying for you it's actually a nice way to start. Experience will come as you learn how to get past problems so, when you learnt the basics, you can actually leave shared hosting and open yourself to a new world.
     
    Last edited by a moderator: Apr 30, 2017
  10. drmike

    drmike 100% Tier-1 Gogent

    8,573
    2,709
    May 13, 2013
    Last edited by a moderator: Apr 30, 2017
  11. michyprima

    michyprima New Member

    8
    2
    Apr 2, 2016
    I respect everybody's opinion, and as you noticed, I prefer discussions over public attacks :)


    Latency was measured from datacenter1 in Arezzo, Italy.


    Another test:

    Code:
    
        ** SIEGE 3.0.5
        ** Preparing 500 concurrent users for battle.
        The server is now under siege…
        Lifting the server siege…      done.
        Transactions:                 197955 hits
        Availability:                  99.98 %
        Elapsed time:                 599.75 secs
        Data transferred:             449.12 MB
        Response time:                  1.50 secs
        Transaction rate:             330.06 trans/sec
        Throughput:                     0.75 MB/sec
        Concurrency:                  493.83
        Successful transactions:      197955
        Failed transactions:              39
        Longest transaction:           31.54
        Shortest transaction:           0.33
    I accepted to stand behind this product because I like it and because it's actually quite good. There really aren't many VPS products with those specs capable of serve almost 500 users (we are still talking about static pages over nginx)


    I also did not "run" as I'm still here. I run a couple of those VPS myself and I can run some more (non-destructive ;)) benchmarks if you want me to.
     
    Last edited by a moderator: Apr 30, 2017
    drmike likes this.
  12. drmike

    drmike 100% Tier-1 Gogent

    8,573
    2,709
    May 13, 2013
    Well definitely interested in more benchmarks and hearing about how your cloud solution differs from run-of-the-mill VPS.  Toss the freevps script through the paces since it's commonly referred to (although flawed and incomplete as can be).


    wget freevps.us/downloads/bench.sh -O - -o /dev/null|bash



    Good, hope to see more of you here and as a good representative of Aruba.  Just being fair is all, plenty of companies pump things out and run and are not active in the communities they hope to get business from.  We frown about that :)
     
  13. michyprima

    michyprima New Member

    8
    2
    Apr 2, 2016
    No problem!

    Code:
    System Info
    -----------
    Processor       : Intel(R) Xeon(R) CPU E5-2650 v3 @ 2.30GHz
    CPU Cores       : 1
    Frequency       : 2299.998 MHz
    Memory          : 993 MB
    Swap            : 951 MB
    Uptime          : 4 days, 2:18,
    
    OS              : Ubuntu 14.04.4 LTS
    Arch            : x86_64 (64 Bit)
    Kernel          : 3.13.0-83-generic
    Hostname        : xxx
    
    
    Speedtest (IPv4 only)
    ---------------------
    Your public IPv4 is xxx
    
    Location                Provider        Speed
    CDN                     Cachefly        43.3MB/s
    
    Atlanta, GA, US         Coloat          15.8MB/s
    Dallas, TX, US          Softlayer       12.2MB/s
    Seattle, WA, US         Softlayer       9.39MB/s
    San Jose, CA, US        Softlayer       9.83MB/s
    Washington, DC, US      Softlayer       16.8MB/s
    
    Tokyo, Japan            Linode          8.56MB/s
    Singapore               Softlayer       6.02MB/s
    
    Rotterdam, Netherlands  id3.net         40.2MB/s
    Haarlem, Netherlands    Leaseweb        73.3MB/s
    
    
    Disk Speed
    ----------
    I/O (1st run)   : 739 MB/s
    I/O (2nd run)   : 1.0 GB/s
    I/O (3rd run)   : 1.1 GB/s
    Average I/O     : 247.033 MB/s
    Looking at the avg I/O I can see why you call this script flawed - 247 is definitely not the average between the three runs! :p


    Edit: looking at the bash script, I see why it's failing: it does not check if values are GB or MB, so threats 1.0 GB/s like it was 1.0 MB/s.


    I would also never alter the results of my tests. I know this is just my word and I'm no one, but hey, my ethics are pretty rigid on that :)
     
    Last edited by a moderator: Apr 3, 2016
  14. marrco

    marrco New Member

    22
    10
    May 22, 2013
    if you need more info, i'm a long time user of theirs, and in direct contact with their staff. And *NOT* accepting the free credit they offered to be an evangelist on some forums. At least not at the moment. So all the services i have are paid from my own pocket.


    That said i have 6 arubacloud vps always active since last august, and i tested all datacenter. Best vps for the price, if Europe location can do for you. Rock solid, minor packet loss when nodes went under attack. Other minor glitches sometimes, but, even if i don't love it, it's a large large with serious gear and infrastructure. And a ugly control panel.
     
  15. fm7

    fm7 Active Member

    163
    61
    Jul 26, 2014
    :~# dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync



    Frankfurt: 940 MB/s


    London: 30 MB/s


    Elapsed time deploying DE VPS: 7 minutes



    Command:


    AddVirtualMachineSmartVMWare


    Status:


    Completed


    Context:


    DE


    User:


    xxxxxxx


    Start Date And Time :


    05/04/2016 - 02:08


    Stop Date And Time :


    05/04/2016 - 02:15



    Elapsed time deploying UK VPS: 43 minutes



    Command:


    AddVirtualMachineSmartVMWare


    Status:


    Completed


    Context:


    UK


    User:


    xxxxxx


    Start Date And Time :


    05/04/2016 - 05:04


    Stop Date And Time :


    05/04/2016 - 05:47
     
    Last edited by a moderator: Apr 5, 2016
  16. marrco

    marrco New Member

    22
    10
    May 22, 2013
    it depends on the server. Some have been (wrongly?) limited. Try to open a support request. This is my UK server:


    dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync
    16384+0 records in
    16384+0 records out
    1073741824 bytes (1.1 GB) copied, 2.07431 s, 518 MB/s
     


    :~# bash bench.sh
    CPU model :  Intel(R) Xeon(R) CPU E5-2650 0 @ 2.00GHz
    Number of cores : 1
    CPU frequency :  2000.000 MHz
    Total amount of ram : 1000 MB
    Total amount of swap : 1903 MB
    System uptime :   4 min,
    Download speed from CacheFly: 108MB/s
    I/O speed :  609 MB/s
     


    :~/ioping-master# ./ioping .
    4 KiB from . (ext4 /dev/dm-0): request=1 time=143 us
    4 KiB from . (ext4 /dev/dm-0): request=2 time=5.16 ms
    4 KiB from . (ext4 /dev/dm-0): request=3 time=358 us
    4 KiB from . (ext4 /dev/dm-0): request=4 time=282 us
    4 KiB from . (ext4 /dev/dm-0): request=5 time=326 us
    4 KiB from . (ext4 /dev/dm-0): request=6 time=366 us
    4 KiB from . (ext4 /dev/dm-0): request=7 time=332 us
    4 KiB from . (ext4 /dev/dm-0): request=8 time=310 us
    4 KiB from . (ext4 /dev/dm-0): request=9 time=312 us
    4 KiB from . (ext4 /dev/dm-0): request=10 time=332 us
    ^C
    --- . (ext4 /dev/dm-0) ioping statistics ---
    10 requests completed in 7.92 ms, 40 KiB read, 1.26 k iops, 4.93 MiB/s
    min/avg/max/mdev = 143 us / 792 us / 5.16 ms / 1.46 ms


    :~/ioping-master# ./ioping -R /dev/dm-0


    --- /dev/dm-0 (block device 17.7 GiB) ioping statistics ---
    21.3 k requests completed in 2.91 s, 83.4 MiB read, 7.33 k iops, 28.6 MiB/s
    min/avg/max/mdev = 34 us / 136 us / 17.5 ms / 219 us
     
     
    Last edited by a moderator: Apr 5, 2016
  17. fm7

    fm7 Active Member

    163
    61
    Jul 26, 2014
    (CEST -> UTC+2)


    05/04/2016 15:40 - The de27 Virtual Server has been started successfully .


    05/04/2016 15:38 - The de27 Virtual Server has been shutdown successfully.


    05/04/2016 15:29 - The de27 Virtual Server has not been shutdown successfully .



    Command:


    ShutdownVirtualMachineSmartVMWare


    Status:


    Error


    Context:


    de27


    User:


    xxxxxxxx


    Start Date And Time :


    05/04/2016 - 13:01


    Stop Date And Time :


    05/04/2016 - 15:29
     
    Last edited by a moderator: Apr 5, 2016
  18. DomainBop

    DomainBop Dormant VPSB Pathogen

    2,258
    2,187
    Oct 11, 2013
    I've had my 2 VPS in IT1 since April 2015 so they're on older hardware:

    The one in London is an E5:

    @marrco said:  

    The best VPS for the price I have is an 8GB/8core/160GB disk Leaseweb VPS in Singapore that I'm paying about US $11 monthly for (S$14.97, 75% off sale when they launched the location), but the ArubaCloud VPS in both Arezzo and London has been very reliable without any outages.  If I had to grumble about something it would probably be the ordeal of trying to get VAT removed from the invoices...

    I must be the only one who doesn't mind the appearance. :p 

    I've had absolutely no contact with their staff other than opening a ticket to billing about VAT when I first signed up last year.  0 tickets needed for technical support in 12 months.
     
    fm7 likes this.
  19. michyprima

    michyprima New Member

    8
    2
    Apr 2, 2016
    That's great but even if you don't mind the current look it doesn't mean we should stop improving it ;)


    Having a technical rep directly on the forum may be a good thing in case someone needs help doing something I guess. I'm glad you haven't had any problems but I also know no system is perfect :)
     
  20. myacy1

    myacy1 New Member

    1
    0
    Nov 14, 2016
    I took a VPS service with OS Ubuntu Server 16, I installed another Zpanel, Vesta, ICPconfig and none went to the site dns add. VPS website and now I do not ns1.webdesign-bs.eu ping.