Scaleway Cloud Launches x86-64 C2 Servers: Dedicated Avoton C2550 and C2750

[fm7]

Scaleway dedicated C2 S  (hourly rate billing)


12€/month 4 cores  8GB RAM 50GB LSSD 300Mbit unmetered


1€/month each additional 50GB LSSD


1€/month each additional IPv4

BYTE UNIX Benchmarks (Version 5.1.3)


   CPU 0: Intel(R) Atom(TM) CPU C2550 @ 2.40GHz (4787.8 bogomips)
   CPU 1: Intel(R) Atom(TM) CPU C2550 @ 2.40GHz (4787.8 bogomips)
   CPU 2: Intel(R) Atom(TM) CPU C2550 @ 2.40GHz (4787.8 bogomips)
   CPU 3: Intel(R) Atom(TM) CPU C2550 @ 2.40GHz (4787.8 bogomips)


Benchmark Run: Sun Mar 13 2016 21:41:29 - 22:02:09
4 CPUs in system; running 1 parallel copy of tests
System Benchmarks Index Score (Partial Only) 804.3


Benchmark Run: Sun Mar 13 2016 22:02:09 - 22:22:55
4 CPUs in system; running 4 parallel copies of tests
System Benchmarks Index Score (Partial Only) 2337.2


:~# 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, 5.47228 s, 196 MB/s

 

[DomainBop]

OVH VPS 2016 SSD 1 (flat-rate)


3€ VPS KVM OpenStack 1 vCPU 2GB RAM 10GB SSD 100Mbit unmetered


           BYTE UNIX Benchmarks (Version 5.1.3)


           CPU 0: Intel Xeon E312xx (Sandy Bridge) (4788.9 bogomips)


        Benchmark Run: Thu Feb 18 2016 17:23:52 - 17:44:44
        1 CPU in system; running 1 parallel copy of tests
        System Benchmarks Index Score (Partial Only)  529.6


:~# 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, 6.06531 s, 177 MB/s

These are the results for a 2GB Public Cloud VPS-SSD 1 instance in Gravelines GRA1 (€2.99 monthly or €0.008 /hour)

BYTE UNIX Benchmarks (Version 5.1.3)


   System: bordeaux.pig.bz: GNU/Linux
   OS: GNU/Linux -- 4.3.0-0.bpo.1-amd64 -- #1 SMP Debian 4.3.5-1~bpo8+1 (2016-02-23)
   Machine: x86_64 (unknown)
   Language: en_US.utf8 (charmap="UTF-8", collate="UTF-8")
   CPU 0: Intel Xeon E312xx (Sandy Bridge) (4788.9 bogomips)
          x86-64, MMX, Physical Address Ext, SYSENTER/SYSEXIT, SYSCALL/SYSRET, Intel virtualization
   20:57:44 up 10:28,  1 user,  load average: 0.00, 0.03, 0.05; runlevel 5


------------------------------------------------------------------------
Benchmark Run: Sat Mar 05 2016 20:57:44 - 21:25:57
1 CPU in system; running 1 parallel copy of tests


Dhrystone 2 using register variables       28816487.9 lps   (10.0 s, 7 samples)
Double-Precision Whetstone                     3698.5 MWIPS (9.8 s, 7 samples)
Execl Throughput                               4860.8 lps   (29.9 s, 2 samples)
File Copy 1024 bufsize 2000 maxblocks       1239251.7 KBps  (30.0 s, 2 samples)
File Copy 256 bufsize 500 maxblocks          343182.8 KBps  (30.0 s, 2 samples)
File Copy 4096 bufsize 8000 maxblocks       2681223.4 KBps  (30.0 s, 2 samples)
Pipe Throughput                             2622817.2 lps   (10.0 s, 7 samples)
Pipe-based Context Switching                 407802.5 lps   (10.0 s, 7 samples)
Process Creation                              12052.8 lps   (30.0 s, 2 samples)
Shell Scripts (1 concurrent)                   6296.5 lpm   (60.0 s, 2 samples)
Shell Scripts (8 concurrent)                    843.9 lpm   (60.0 s, 2 samples)
System Call Overhead                        3975227.2 lps   (10.0 s, 7 samples)


System Benchmarks Index Values               BASELINE       RESULT    INDEX
Dhrystone 2 using register variables         116700.0   28816487.9   2469.3
Double-Precision Whetstone                       55.0       3698.5    672.5
Execl Throughput                                 43.0       4860.8   1130.4
File Copy 1024 bufsize 2000 maxblocks          3960.0    1239251.7   3129.4
File Copy 256 bufsize 500 maxblocks            1655.0     343182.8   2073.6
File Copy 4096 bufsize 8000 maxblocks          5800.0    2681223.4   4622.8
Pipe Throughput                               12440.0    2622817.2   2108.4
Pipe-based Context Switching                   4000.0     407802.5   1019.5
Process Creation                                126.0      12052.8    956.6
Shell Scripts (1 concurrent)                     42.4       6296.5   1485.0
Shell Scripts (8 concurrent)                      6.0        843.9   1406.5
System Call Overhead                          15000.0    3975227.2   2650.2
                                                                   ========
System Benchmarks Index Score                                        1713.6

dd VPS:
 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, 2.49473 s, 430 MB/s


dd Additional Disk (High Speed Volume, triple replication, €0.08 /month/GB of storage):
 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, 2.86435 s, 375 MB/s

 

[fm7]

Oh I got confused and thought LSSD was local SSD.

LSSD (which stands for local SSD) is local but not directly attached (except C2 L -- 250GB SSD included).


BTW LSSD space is "temporary", Each time you stop/start a VM the content of each attached volume is copied to/from storage located elsewhere. Depending on how many attached disks you have and the space used it may take an awful lot of time to stop and start a VM.


From the FAQ


Each server has access to a pool of local drives. These drives are exported to servers via the NBD protocol which effectively makes them network drives. However, network between these drives and Cx nodes are dedicated PCB tracks which ensures minimal latency and avoids network congestion.


There is no redundancy on these volumes, you need to handle redundancy on your side! They are archieved to a permanent storage when you start and stop your server.

 

[fm7]

These are the results for a 2GB Public Cloud VPS-SSD 1 instance in Gravelines GRA1 (€2.99 monthly or €0.008 /hour)

Thanks!


The results I did post before refer to OVH-BHS (Montreal).


BTW that VPS recently had ~9h20m downtime caused by power outage (short-circuit, 4h:40m to repair the electrical damage) followed by network outage/issues. No redundancy at all, no way comparable to Scaleway's Iliad DC-3 2N facility.


Same data center (OVH-BHS), another instance, few months ago:


BYTE UNIX Benchmarks (Version 5.1.3)


Benchmark Run: Mon Nov 23 2015 17:57:02 - 18:17:56
1 CPU in system; running 1 parallel copy of tests
System Benchmarks Index Score (Partial Only) 1545.8


:~# 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, 2.71918 s, 395 MB/s


Strasbourg SBG1:


BYTE UNIX Benchmarks (Version 5.1.3)


CPU 0: AMD Opteron(tm) Processor 6386 SE 


Benchmark Run: Fri Jun 05 2015 06:06:56 - 06:28:23
1 CPU in system; running 1 parallel copy of tests
System Benchmarks Index Score (Partial Only)   338.7


root@vps175585:/# 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, 5.56097 s, 193 MB/s

 

[fm7]

100% designed by Scaleway R&D teams
​

​

The C2 chassis has 18 servers per 3U
​

 
​

Scaleway C2 @ Online.net / Iliad DC-3

 

[DomainBop]

Scaleway C2 offers


€11.99month
Avoton C2550
4 Dedicated x64 Cores
8GB Memory
50GB SSD Disk 


€17.99month
Avoton C2750
8 Dedicated x64 Cores
16GB Memory
50GB SSD Disk 

Online.net released their new Personal Range dedicateds today.  €20.00 setup fee again and:


1. €8.99 Dedibox SC Avoton C2350 (2 cores) 4GB RAM, 500GB HDD or 120GB SSD (replaces the 5.99 2GB RAM Nano U2250)


2.€15.99 Dedibox XC Avoton C2750 (8 cores) 16GB RAM, 1TB HDD or 250GB SSD (double the RAM of the old offering and SSD increased from 120 to 250GB)

 

[willie]

That's interesting, comparing the new online.net personal range to the new Scaleways.  The monthly charge is a little lower than the C2S/C2M and you get direct disk, but there's the setup fee.  The 750GB ftp backup for 5eur/mo available with those servers is also nice.

 

[graeme]

Could anyone run networking benchmarks and/or ioping on scaleway and/or online.net personal and/or OVH cloud?

 

[fm7]

Scaleway dashing.io (Twitter post, March 25, 2016)

 

[willie]

Could anyone run networking benchmarks and/or ioping on scaleway and/or online.net personal and/or OVH cloud?

Scaleway C1 (ARM) ioping:


10 requests completed in 9.02 s, 900 iops, 3.52 MiB/s
min/avg/max/mdev = 684 us / 1.11 ms / 1.50 ms / 285 us


 dd if=/dev/zero of=test bs=64k count=16k conv=fdatasync; unlink test
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 10.2818 s, 104 MB/s


Do you have a specific network benchmark you want me to run?

 

[DomainBop]

Could anyone run networking benchmarks and/or ioping on scaleway and/or online.net personal and/or OVH cloud?

Disk:
Scaleway C1 (Armv7)
10 requests completed in 9.0 s, 875 iops, 3.4 MiB/s
min/avg/max/mdev = 886 us / 1.1 ms / 1.4 ms / 186 us


 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, 9.654 s, 111 MB/s


Online.net Dedibox SC (Avoton C2350)
--- /tmp (ext4 /dev/sda2) ioping statistics ---
10 requests completed in 9.0 s, 7.4 k iops, 28.9 MiB/s
min/avg/max/mdev = 114 us / 135 us / 153 us / 14 us


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, 12.1061 s, 88.7 MB/s


Online.net Dedibox SC (VIA Nano U2250)
--- /tmp (ext4 /dev/sda2) ioping statistics ---
10 requests completed in 9.0 s, 2.2 k iops, 8.7 MiB/s
min/avg/max/mdev = 347 us / 449 us / 545 us / 60 us


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, 14.7896 s, 72.6 MB/s


OVH VPS-SSD  (Public Cloud OpenStack version):
--- /tmp (ext4 /dev/vda1) ioping statistics ---
10 requests completed in 9.0 s, 2.0 k iops, 7.9 MiB/s
min/avg/max/mdev = 355 us / 492 us / 570 us / 70 us


 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, 2.21723 s, 484 MB/s


Additional Disk (OVH Public Cloud):
--- /opt (ext4 /dev/vdb) ioping statistics ---
10 requests completed in 9.0 s, 686 iops, 2.7 MiB/s
min/avg/max/mdev = 914 us / 1.5 ms / 2.9 ms / 494 us


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, 3.07906 s, 349 MB/s


================================================


Network:


OVH VPS-SSD  (Public Cloud OpenStack version):


Location        Provider    Speed
CDN            Cachefly    11.3MB/s


Atlanta, GA, US        Coloat        1.31MB/s 
Dallas, TX, US        Softlayer    7.93MB/s 
Seattle, WA, US        Softlayer    8.41MB/s 
San Jose, CA, US    Softlayer    7.38MB/s 
Washington, DC, US    Softlayer     7.44MB/s 


Tokyo, Japan        Linode        5.76MB/s 
Singapore         Softlayer    5.66MB/s 


Rotterdam, Netherlands    id3.net        11.8MB/s
Haarlem, Netherlands    Leaseweb    11.9MB/s 


Hosted by Orange (Paris) [1.71 km]: 9.007 ms
Testing download speed........................................
Download: 100.06 Mbit/s
Testing upload speed..................................................
Upload: 95.62 Mbit/s
 


Scaleway C1 (Armv7)


Location        Provider    Speed
CDN            Cachefly    99.4MB/s


Atlanta, GA, US        Coloat        1.20MB/s 
Dallas, TX, US        Softlayer    10.1MB/s 
Seattle, WA, US        Softlayer    7.49MB/s 
San Jose, CA, US    Softlayer    7.62MB/s 
Washington, DC, US    Softlayer     11.9MB/s 


Tokyo, Japan        Linode        6.39MB/s 
Singapore         Softlayer    2.42MB/s 


Rotterdam, Netherlands    id3.net        45.8MB/s
Haarlem, Netherlands    Leaseweb    49.9MB/s 


Hosted by NEOTELECOMS (Paris) [1.59 km]: 4.18 ms
Testing download speed........................................
Download: 918.68 Mbit/s
Testing upload speed..................................................
Upload: 177.68 Mbit/s
 


Online.net DC2 (Dual Xeon E5620) 


Location        Provider    Speed
CDN            Cachefly    83.9MB/s


Atlanta, GA, US        Coloat        16.1MB/s 
Dallas, TX, US        Softlayer    11.5MB/s 
Seattle, WA, US        Softlayer    8.03MB/s 
San Jose, CA, US    Softlayer    9.29MB/s 
Washington, DC, US    Softlayer     17.4MB/s 


Tokyo, Japan        Linode        10.3MB/s 
Singapore         Softlayer    6.56MB/s 


Rotterdam, Netherlands    id3.net        62.1MB/s
Haarlem, Netherlands    Leaseweb    101MB/s 


Hosted by NEOTELECOMS (Paris) [1.59 km]: 1.882 ms
Testing download speed........................................
Download: 904.85 Mbit/s
Testing upload speed..................................................
Upload:338.54 Mbit/s
 


Online.net DC3 (Xeon E3-1220)


Location        Provider    Speed
CDN            Cachefly    98.5MB/s


Atlanta, GA, US        Coloat        15.5MB/s 
Dallas, TX, US        Softlayer    15.2MB/s 
Seattle, WA, US        Softlayer    10.9MB/s 
San Jose, CA, US    Softlayer    11.7MB/s 
Washington, DC, US    Softlayer     19.1MB/s 


Tokyo, Japan        Linode        7.92MB/s 


Singapore         Softlayer    8.10MB/s 


Rotterdam, Netherlands    id3.net        63.1MB/s
Haarlem, Netherlands    Leaseweb    58.8MB/s


Hosted by NEOTELECOMS (Paris) [1.30 km]: 1.796 ms
Testing download speed........................................
Download: 891.31 Mbit/s
Testing upload speed..................................................
Upload: 408.21 Mbit/s
 

 

[willie]

Scaleway C2M ioping:


10 requests completed in 9.00 s, 2.98 k iops, 11.6 MiB/s
min/avg/max/mdev = 259 us / 335 us / 515 us / 73 us


I tried to get a C2L but they were out of stock.  I've been running a few other benchmarks on the C2M but want to release it tonight or tomorrow, so let me know if there's any particular tests anyone wants.

 

[fm7]

 

[graeme]

Thanks @willie and @DomainBop for running those tests.


I have been testing two other providers (Bytemark and Upcloud) recently for a project of mine, and everyone seems to provide bandwidth reasonable close to limit within Europe. They are all a lot slower to Asia or the US. Upcloud has very fast storage.


I did some download testing on an OVH dedi I have access to, and OVH is easily the slowest to the US (quite good to Asia though). I also got a rather unimpressive reply to a question - although that maybe because they have an issue at the moment and are handling a lot of enquiries. Among other things they said the Ceph storage is mounted locally, which confuses me.

One thing I did find out is that you cannot seamlessly move from a cloud VPS to a public cloud instance.


On the whole I am not that keen on using OVH for anything that matters. How good is online.net customer service and does anyone have any idea how reliable are those cheap dedis are likely to be?

 

[willie]

I don't think Scaleway is that solid right now.  It's a cool product but it feels like something in beta test.  There will probably also always be hardware shortages.  That said I'm getting more and more paranoid about how to do HA, so would want to spread critical services across multiple providers, meaning Scaleway can be one of them.  I've gotten reasonable replies from their support though.  I don't know if having a developer account (3 euro/month) mattered for that.

 

[drmike]

Just wanted to say thanks to everyone on this thread.  Nice to see numbers and experiences.   I like what online.net is doing, even though some rough spots to it and bumps along the way.


Saw someone with one of the Avotons pushing a ton of data - like 30TB in 24 hours.  Pretty impressed that they let someone push packets like that

 

[DomainBop]

I don't think Scaleway is that solid right now.  It's a cool product but it feels like something in beta test. 

The Online compiled Debian ARM kernels have had some bugs .  There have been a couple of bugs on various kernel upgrades that caused iptables to spew out "can't initialize iptables table `filter': Table does not exist (do you need to insmod?" error messages.  I last encountered this problem today after changing to the 4.5 docker kernel on Debian Jessie (today's problem was easily solved with a quick "mv /lib/modules /lib/modules.old" and "bash -x /usr/local/sbin/oc-sync-kernel-modules".   IPtables errors on other kernel versions have also been easily solved by adding things like module-init-tools that were missing from an earlier version of Online's Debian templates, etc...note the missing module-init-tools bug was fixed in September)

Saw someone with one of the Avotons pushing a ton of data - like 30TB in 24 hours.  Pretty impressed that they let someone push packets like that

Online will tolerate short term bursts of data usage like that for a few days but if someone does that every day  they're going to run into problems like this guy did http://www.webhostingtalk.com/showthread.php?t=1481943

does anyone have any idea how reliable are those cheap dedis are likely to be?

The hardware and network are very reliable on those sub $10 dedis at Online and Kimsufi and you rarely, if ever, will need to open a ticket (especially at Kimsufi/OVH because their automated monitoring system is excellent and automatically dispatches technicians when it detects a problem).

 

[willie]

One of the ingredients of Scaleway was supposed to be an S3-like object store at .02/GB/month, triple replication, and no bandwidth charges.  It got oversubscribed instantly (people seemed to be using it as a CDN) and so it's been unavailable to new customers for the past 5 months or so while they scale it out.  There are some indications that they might reopen it soon, but who knows.  Its unavailability is one of the reasons Scaleway doesn't seem completely together yet.

 

[fm7]

IMO "mission critical" is incompatible with non-ECC RAM, non-redundant local storage,  shared chassis, ...


Online.net offers low-priced (as low as 20€/month) A/B powered enterprise-grade HP/Dell servers with IPMI, RAID HW, SAS HDs, ECC RAM, plus serious SLAs and inexpensive true SAN-HA storage.


Scaleway's baremetal server is just Arnaud's "cloud alternative" as the Online.net's CEO/CTO thinks the regular VPS/VM brings the worst of worlds: what is bad in dedicated servers plus what is bad in shared hosting.


The thing is ... this site is about VPS, public cloud and I think it is delusional to talk "mission critical" applications running on VPS.

 

[fm7]

One of the ingredients of Scaleway was supposed to be an S3-like object store at .02/GB/month, triple replication, and no bandwidth charges.  It got oversubscribed instantly (people seemed to be using it as a CDN) and so it's been unavailable to new customers for the past 5 months or so while they scale it out.  There are some indications that they might reopen it soon, but who knows.  Its unavailability is one of the reasons Scaleway doesn't seem completely together yet.

I guess part of problem is because the content of all volumes attached to a Scaleway server is copied to the permanent storage each time the server is stopped.


Recently Online.net's CEO posted on his Twitter account a short video showing a row of racks filled with storage hardware but didn't comment the intended usage. At the time he was asked about the "buckets":


HADJEDJ Vincent ‏@VincentHadjedj Mar 12


@online_fr bientôt le retour des Buckets chez scaleway ?


Translated from French by Bing


@online_fr soon the return of the Buckets in scaleway?


Online.net - Arnaud ‏@online_fr Mar 12


@VincentHadjedj oui !