All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Spare server, what kind of vps?
We replaced one of our old r1soft backup server. It's smaller than our cluster servers so I was thinking about using as standalone server only for small vps. (the alternative should be to use for second tier backups, but the disks are a bit small)
Here are some details:
4 core xeon X3360 @ 2.83GHz
8 GB ram
4x512GB SATA disks with software RAID 10:
cat /proc/mdstat
Personalities : [raid1] [raid10]
md1 : active raid10 sda2[0] sdd2[3] sdc2[2] sdb2[1]
976510848 blocks 64K chunks 2 near-copies [4/4] [UUUU]
md2 : active raid1 sda1[0] sdb1[1]
128384 blocks [2/2] [UU]
dd if=/dev/zero of=iotest bs=64k count=16k conv=fdatasync
16384+0 records in
16384+0 records out
1073741824 bytes (1.1 GB) copied, 8.46397 s, 127 MB/s
pm10:~# pvs
PV VG Fmt Attr PSize PFree
/dev/md1 bigdisk lvm2 a- 931.27g 526.46g
At present the server run debian
pm10:~# cat /etc/debian_version
6.0.3
One gigabit interface:
pm10:~# wget -O /dev/null http://cachefly.cachefly.net/100mb.test
--2012-02-02 16:58:39-- http://cachefly.cachefly.net/100mb.test
Resolving cachefly.cachefly.net... 205.234.175.175
Connecting to cachefly.cachefly.net|205.234.175.175|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 104857600 (100M) [application/octet-stream]
Saving to: “/dev/null”
100%[=====================================================================================================================================================================================================================>] 104,857,600 50.8M/s in 2.0s
2012-02-02 16:58:41 (50.8 MB/s) - “/dev/null” saved [104857600/104857600]
The questions for you guys are:
1) Xen or KVM ? (we have no openvz experience)
2) 28 VPS x 256M ram and 25GB disk (1 virtual core) or 56 x 128M ram and 15GB disk?
3) Which control panel?
4) Price?
Thank you.
P.S. at present I don't ever know if it will be eventually configured, but I'd like to get some feedbacks.
Comments
Keep in mind that you will need 28-56 IPs to host that many VPSses.
This should not be a big problem :-)
Anyway these are just my personal suggestions, and you can ask more for a VPS in Italy, I didn't see many LEB providers there.
My opinion:
Thanks.
As for the control panel I see that SolusVM is very popular and it can be integrated with whmcs (none of our actual solutions do this). Another would be HiperVM. experiences?
You mean HyperVM?
I worked a bit with it, as far as I know you can integrate it also with WHMCS. HyperVM is okay, live-migrating is a cool feature it has - but I would prefer SolusVM.
HyperVM looks a bit older, had in the last years some very bad security-bugs, has a awful design, and so on.
Debian 6.0.3 is one week old now. Don't you update your systems? ;-)
yes I do
It seems like SolusVM is a preferred choice. I see they have planned live migration for 2.0 but without a SAN, it should not be a problem.
yes we do, I just launched the upgrade just after posting here :-)
pm10:~# cat /etc/debian_version
6.0.4
4 core xeon X3360 @ 2.83GHz
8 GB ram
Never used that CPU, but I've never run a VPS node out of CPU, so I'm sure it's fine, but software raid10 you'll run out of IO with that many guests. Go look on eBay, you can pick up a used 3ware 9650SE-4LPML for $100 give or take.
I see, this would be an experiment more than a VPS production system. If I had a spare FC card I would connect it to one of our old storagetek 6130
I should say that with CDP this server had several parallels clients writing backups without problems. Anyway this is a point to consider, so it should be better to have less vps then more...
Thank you
Yes and no, it all really depends on the clients and their needs. I've seen 128mb clients use IO and bandwidth you would expect from a 2gb client, and I have seen 2gb clients idle for 6 months at a crack....
I've never seen any provider use a SAN and not find a home in the deadpool. I hear you, and that you are only trying to test/proof of concept, but this is an addiction, clients and providers alike
Just consider that we are new to the whole LEB concept, I know you see often here company/startup that never goes over a year of business, scammers and such. But I hope you all (providers and potential clients if any) will find that this isn't the case with us/me.
I'm 46 year old, started playing with computers when 1K of ram (ZX80 / ZX81) was enough :-) and started working in IT when I was 19. We started prometeus and iperweb 15 years ago and, since 2005 we have been focused on high end services. Last year, we decided to use some of the resources we have in excess to regain some clients in the bottom arena of hosting and vps market.
Thanks.
Coming back to the questions, other toughts? :-)
We used to have a lot of visibility until 2000 (our brand was iperweb.com that we incorporated in Delaware as LLC as not resident aliens in the late '90), then when big hosting player started pushing thousand of dollars in advertising (at the time banners, sponsorship and mailing lists were the main traffic drivers) we decided that it was too risky to put so much money in CPM campaigns. In the same period a few things happened:
So we decided to close IperWeb LLC and pass all the clients to Prometeus (our Italian company) so hosting business was stabilized and handled by two technician for support and no more marketing was done.
We continued to grow in Italy with managed high end services (for example we manage the web farm and all the streamings for one of the bigger Italian radio group with more than 60K daily listeners) and hosting became a fraction of what it used to be.
If you like to laugh
http://web.archive.org/web/19980428175025/http://iperweb.com/
:-)
Starting from the consideration of Tim about the possible I/O bottleneck I decided to use the server as is, with Debian, XEN and without any control panel, to conduct a test for some months. I'm going to script the installation of 25 debian squeeze with 128M and 15G of disk and give them away for free use for at least 6 months (I don't plan to sell these). No abuse will be tolerated, I need some real feedback back from end users and providers :-)
Please PM me if you are interested.
Thanks
S.
A few vm activated and this is the poor server :-)
top - 01:01:59 up 2:25, 3 users, load average: 7.39, 4.25, 2.33 Tasks: 129 total, 1 running, 128 sleeping, 0 stopped, 0 zombie Cpu(s): 0.0%us, 0.0%sy, 0.0%ni, 47.6%id, 52.4%wa, 0.0%hi, 0.0%si, 0.0%stJeez, What are you running on them containers for it to cause a 7.39 CPU Load?
X3360 isn't really old.
But a 7 load for a 4 core cpu is a lot if is constant
Look at the IO Wait at 50%, there's your load, either everyone is doing a dd test at once, or mdadm is verifying/rebuilding the raid.
@miTgiB indeed there was a mix: user activity, the script that was completing the deployment of the vm and a tail of a md check.
I posted here as soon I saw the io wait so high :-)
luckily it was a spike of about 10-15 minutes
I am interested, sending PM
Just a reminder that I am at 15 vm and there is more space for others to join (PM me). As stated before the plan is to keep this server up for 6 months, no backup will be done (so please make your duty) and support will be given with low priority only by me. No abuse or threat to others will be tolerated.
Coming back to the server, it's running a kernel 3.1.0 with xen 4.1.2 on a debian wheeze. DomU are all squeeze for now. I had one request for centos, I tried a xen-tools/rinse install but this didn't worked on the first shot, so I need to retry with a little more of spare time later.
After the high load of yesterday (I wasn't monitoring it yet) thing went well:
!(http://prometeus.net/images/pm10.png)
When you have thoughs or feedback (good or bad) you can PM or use the forum if it's an argument that might be of interest or can be opened to public discussion. I would also appreciate if you can share with me the main usage of the VM so I can make an idea of the usage patterns.
Thank you all for the help.
S.
Just an idea,
Could there be like public status page for this node, something like ramhost is having for their nodes(http://www.ramhost.us/?page=status&s=vz20.atl.rhnx.net for example).
What I'm using mine, will be smokeping server which keeps track of latency and also as a network monitoring node.
What I'm seeing with little testing, routes to Italy are pretty diverse. From Helsinki to Italy one isp takes straight Level3 path from their Helsinki router to Milan router, which gives 40ms round-trip. However, most of the isps here use linx for peering off the traffic with seabone, which gives +30ms extra latency.
What i would like to comment on your network, is that you're having truly multihomed network, always a good one. It's not the fastest MB/s wise, but I could see that you're having some kind of bandwidth throttling on place for this server? Ping times to north europe eyeball networks are generally good, which you can be proud of.
I'll work for a public status page.
As for the network throttling the server should not have limitation. Could you please give me some url I can test from the dom0 ? (Level3 and Seabone/Telecom Sparkle give us good latency also with USA).
Thank you
S.
Sure,
ftp://ftp.funet.fi/dev/100MBnull
I think they (or some intermediate carrier) are throttling down, as I get very poor results.
Please try this:
http://mirror.academica.fi/CentOS/6/isos/x86_64/CentOS-6.2-x86_64-LiveCD.iso
this is what I get from the dom0:
pm10:/etc/xen# wget -O /dev/null http://mirror.academica.fi/CentOS/6/isos/x86_64/CentOS-6.2-x86_64-LiveCD.iso
--2012-02-05 22:23:16-- http://mirror.academica.fi/CentOS/6/isos/x86_64/CentOS-6.2-x86_64-LiveCD.iso
Resolving mirror.academica.fi (mirror.academica.fi)... 87.108.37.26, 2001:a68:1:4::c
Connecting to mirror.academica.fi (mirror.academica.fi)|87.108.37.26|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 731906048 (698M) [application/octet-stream]
Saving to: `/dev/null'
100%[=======================================================================================================================================================================================>] 731,906,048 55.4M/s in 14s
2012-02-05 22:23:30 (51.5 MB/s) - `/dev/null' saved [731906048/731906048]
Really strange, with same command I get only ~2MB/s from academica. And for sure I am using right ssh connection/server, checked it twice :P
I'm trying from one domu, please wait...