New on LowEndTalk? Please Register and read our Community Rules.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
All new Registrations are manually reviewed and approved, so a short delay after registration may occur before your account becomes active.
Comments
So what was the problem? And could you do a new download test on your new droplet?
They didn't disclose that. I assume that there was some traffic shaping configuration left in their routers for the IPs they were handing out to me. That seems to be cleared up now.
Sure. DigitalOcean resuses IP addresses so the new droplet is on the same IP:
DL test from OVH (OVH server connected with 100Mbit/sec):
xxx@yyy:~$ wget -O /dev/null 198.211.100.5/test100.zip
100%[======================================>] 104,857,600 9.52M/s in 12s
2013-03-01 16:46:55 (8.14 MB/s) - `/dev/null' saved [104857600/104857600]
DL test from BuyVM:
xxx@bvm:~$ wget -O /dev/null 198.211.100.5/test100.zip
100%[======================================>] 104,857,600 2.79M/s in 38s
2013-03-01 07:53:44 (2.60 MB/s) - `/dev/null' saved [104857600/104857600]
Perhaps I should check, my download performenace is not so good. Ticket time!
Hey guys,
Ben here, one of the co-founders of DigitalOcean.
We do not oversubscribe network bandwidth. Each physical server is connected via a 1Gbps uplink to our core network. Our core network is connected via multiple redundant 10G providers to the net, Cogent and Level3 are currently deployed for New York.
By default we use the e1000 network driver for KVM virtualization and from our early analysis that is responsible for the degraded network performance that you guys have been experiencing. We're working on implementing VirtIO as the default network driver that has shown consistently to have better performance.
For verified users you already have the option of selecting VirtIO network driver when you create a new droplet, so you can test to see if it alleviates the network speed problems you're encountering.
I just want to reiterate that we do not oversubscribe our bandwidth or play favorites with different providers.
We're more than happy to work with you and see if VirtIO is the solution - so feel free to open a support ticket and mention LET post and debugging VirtIO network speeds. We'll dig into it and make sure that we can deliver full speed without any limits.
Regards!
@zagi my earlier claim I thought you oversold the network was wrong. I opened a ticket and now the performance is basically 100Mbit which I'm fine with.
But yeah, nice company.
Heres my DO VM. I got tired of them trafic shaping me down to 2Mbps so a support request seems to have solved it.
--2013-03-03 02:59:49-- http://speedtest.wdc01.softlayer.com/downloads/test500.zip
Resolving speedtest.wdc01.softlayer.com... 208.43.102.250
Connecting to speedtest.wdc01.softlayer.com|208.43.102.250|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 524288000 (500M) [application/zip]
Saving to: `/dev/null'
100%[======================================================================>] 524,288,000 109M/s in 4.9s
2013-03-03 02:59:53 (103 MB/s) - `/dev/null' saved [524288000/524288000]
--2013-03-03 03:00:07-- http://speedtest.hou02.softlayer.com/downloads/test500.zip
Resolving speedtest.hou02.softlayer.com... 173.193.70.98
Connecting to speedtest.hou02.softlayer.com|173.193.70.98|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 524288000 (500M) [application/zip]
Saving to: `/dev/null'
100%[======================================================================>] 524,288,000 88.4M/s in 6.3s
2013-03-03 03:00:14 (79.6 MB/s) - `/dev/null' saved [524288000/524288000]
--2013-03-03 03:00:32-- http://speedtest.ams01.softlayer.com/downloads/test500.zip
Resolving speedtest.ams01.softlayer.com... 159.253.132.18
Connecting to speedtest.ams01.softlayer.com|159.253.132.18|:80... connected.
HTTP request sent, awaiting response... 200 OK
Length: 524288000 (500M) [application/zip]
Saving to: `/dev/null'
100%[======================================================================>] 524,288,000 32.6M/s in 17s
2013-03-03 03:00:49 (28.7 MB/s) - `/dev/null' saved [524288000/524288000]
Thx, Would be nice if this piece of Info was actually available on your website. As i have thought the (relatively) poor network performance was because of DO relying solely on Cogent.
@zagi I was always using the VirtIO option when creating droplets.
In any case I have not been able to reproduce the problem at all for a while now using the same configuration as when I started the thread (even getting the same IP address). So all is well now for me.
How do you explain Etel's response to one of the folks in this thread? He basically told him that under 100KB/s was perfectly normal and to go fly a kite.
@luma Etel mentioned bandwidth delay product as a consideration in determining network throughput, but I don't see her saying that 100KB/s is normal. Additionally I think Benjamin Franklin is the only one who found flying kites as a productive task
We do our best to help our users overcome problems with the service, sometimes it's hard to determine the exact cause and so esoteric edge cases need to be considered.
There certainly are cases where we've been able to replicate the slow network speeds and have attributed it to the e1000 network driver as in my OP. Going forward we're focusing on VirtIO as a solution to this problem. Please spin up a new droplet with VirtIO selected during the create process and let's see if we can reproduce the issue when the new driver is used.
they are awesome oversellers, thats why i megre from them to another providers, i say it, because it's based on my experience, at night, i have good internet speed, + fast node, at day - same to topic starter.
For the price you pay it's not that bad at all.
monthly
rx | tx | total | avg. rate
------------------------+-------------+-------------+---------------
Feb '13 5.70 TiB | 5.87 TiB | 11.57 TiB | 41.09 Mbit/s
Mar '13 772.04 GiB | 793.95 GiB | 1.53 TiB | 51.94 Mbit/s
------------------------+-------------+-------------+---------------
estimated 7.98 TiB | 8.21 TiB | 16.20 TiB |
daily
rx | tx | total | avg. rate
------------------------+-------------+-------------+---------------
yesterday 289.55 GiB | 297.41 GiB | 586.96 GiB | 56.99 Mbit/s
today 248.48 GiB | 255.52 GiB | 504.00 GiB | 52.78 Mbit/s
------------------------+-------------+-------------+---------------
estimated 268.02 GiB | 275.61 GiB | 543.64 GiB |
] 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, 10.5443 s, 102 MB/s
@neqste we certainly do not oversell the network or servers. The reason why you saw more degraded performance during peak hours is because of the e1000 network bug I mentioned. As more users are interacting with the network the larger the performance degradation that occurs. Therefore it stands to reason that at night as the load goes down, the network is able to perform better.
The symptoms would look identical to oversubscription but if you look over the last few comments from other users they are able to achieve amazing speeds using VirtIO network drivers or running on nodes that are not having the e1000 network issue.
I ask that you give our service another chance, especially if you can ask our support team to verify your account and ultimately allow you to use VirtIO for your droplet.
Thanks for this topic. I have been waiting for a LEB competitive provider to pair with my existing Hostigation accounts that has reasonable price, terms, and control panel, free of WHMCS and Solus panels. I went ahead and signed up.
Going forward, I am curious if DigitalOcean handles a client's innocent exposure to DDOS attacks more responsibly than the other provider that would have otherwise met my criteria (albeit at a higher price). AFAIK, none of our sites have ever experienced this, but if it did, I would want my provider to help, not simply shut us off.
Update: Quoting from https://www.digitalocean.com/droplets/new:
A secure page that decides to "e-mail" my password. Disappointing...
@erichi we have a very novice user base and we used to provide absolutely free trials without the need for any credit card information. Just an email and a password would allow you to test the service. This method allowed us to verify the email address without any question, and made it easy for beginners to start up.
Any serious user can either:
Add SSH keys to their accounts and then there is no password set at all making the server extremely secure.
Once they receive the emailed password they can login and change the security profile to meet their exact requirements.
In all fairness we have received numerous complaints about the method that we have chosen to distribute your first password, and we'll review it now that our platform has matured significantly over the past few months.
Thanks for the reply. I am just a client/user (don't really know your business), but I would think that a novice user base is exactly the group that should be given a "best practices" treatment. However, I recognize that LEB providers don't seem to agree with me.
@zagi Is there a way to turn on VirtIO for running instance?
I have number of instances with you, and I can't recreate all those instance.
I have number of instances with you, and I can't recreate all those instance.
Take snapshot/backup then create new instance from backup/snapshot?
I just put up a droplet in the US and wow, the performance difference between it and my one in the Netherlands is fairly shocking. It's very sluggish, tried destroying it and deploying a new one but no, could not work with that kind of sluggish performance.
From Amsterdam with VirtIO. Not that great but better than before. Lets try it again on different times. (It is 20:30PM now)
I am using Virtio and it's improved quit a lot, enough that I am happy with it.
NYC:
root@us1 [~]# wget freevps.us/downloads/bench.sh -O - -o /dev/null|bash
CPU model : QEMU Virtual CPU version 1.0
Number of cores : 1
CPU frequency : 2299.986 MHz
Total amount of ram : 996 MB
Total amount of swap : 511 MB
System uptime : 2:06,
Download speed from CacheFly: 57.3MB/s
Download speed from Coloat, Atlanta GA: 23.8MB/s
Download speed from Softlayer, Dallas, TX: 54.8MB/s
Download speed from Linode, Tokyo, JP: 1.64MB/s
Download speed from Linode, London, UK: 7.66MB/s
Download speed from Leaseweb, Haarlem, NL: 24.8MB/s
Download speed from Softlayer, Singapore: 4.57MB/s
Download speed from Softlayer, Seattle, WA: 16.8MB/s
Download speed from Softlayer, San Jose, CA: 16.2MB/s
Download speed from Softlayer, Washington, DC: 73.8MB/s
I/O speed : 637 MB/s
Netherlands:
root@nl1 [~]# wget freevps.us/downloads/bench.sh -O - -o /dev/null|bash
CPU model : QEMU Virtual CPU version 1.0
Number of cores : 1
CPU frequency : 1999.999 MHz
Total amount of ram : 996 MB
Total amount of swap : 511 MB
System uptime : 1:39,
Download speed from CacheFly: 33.1MB/s
Download speed from Coloat, Atlanta GA: 12.0MB/s
Download speed from Softlayer, Dallas, TX: 18.6MB/s
Download speed from Linode, Tokyo, JP: 335KB/s
Download speed from Linode, London, UK: 36.0MB/s
Download speed from Leaseweb, Haarlem, NL: 44.7MB/s
Download speed from Softlayer, Singapore: 7.87MB/s
Download speed from Softlayer, Seattle, WA: 16.4MB/s
Download speed from Softlayer, San Jose, CA: 13.6MB/s
Download speed from Softlayer, Washington, DC: 23.9MB/s
I/O speed : 584 MB/s
Amsterdam:
root@server:~# wget freevps.us/downloads/bench.sh -O - -o /dev/null|bash
CPU model : QEMU Virtual CPU version 1.0
Number of cores : 1
CPU frequency : 1999.999 MHz
Total amount of ram : 502 MB
Total amount of swap : 0 MB
System uptime : 47 days, 22:27,
Download speed from CacheFly: 25.8MB/s
Download speed from Coloat, Atlanta GA: 2.60MB/s
Download speed from Softlayer, Dallas, TX: 1.29MB/s
Download speed from Linode, Tokyo, JP: 702KB/s
Download speed from Linode, London, UK: 5.73MB/s
Download speed from Leaseweb, Haarlem, NL: 24.5MB/s
Download speed from Softlayer, Singapore: 993KB/s
Download speed from Softlayer, Seattle, WA: 574KB/s
Download speed from Softlayer, San Jose, CA: 448KB/s
Download speed from Softlayer, Washington, DC: 1.65MB/s
I/O speed : 288 MB/s
That's bad compared to others here...
Are you using Virtio? It seems to make a significant difference. What you are getting is about the same I was seeing until I switched to Virtio.
To be honest, with Virtio it also isn't awesome. But maybe I'm just used to ramnodes gigabit port.
Installing stuff via debian's US mirror now, bandwidth jumps between 1xx kb/s and 3Mbyte/s, I'm not very impressed
Mehh, even CVPS who we all know oversells has better performance than this...
You think? Same test, it was painful to watch them processing.
CVPS - Chicago
chicago ~]# wget freevps.us/downloads/bench.sh -O - -o /dev/null|bash
CPU model : Intel(R) Xeon(R) CPU E5-2620 0 @ 2.00GHz
Number of cores : 4
CPU frequency : 2000.205 MHz
Total amount of ram : 2048 MB
Total amount of swap : 0 MB
System uptime : 12 days, 4:38,
Download speed from CacheFly: 16.6MB/s
Download speed from Coloat, Atlanta GA: 1.45MB/s
Download speed from Softlayer, Dallas, TX: 1.94MB/s
Download speed from Linode, Tokyo, JP: 387KB/s
Download speed from Linode, London, UK: 670KB/s
Download speed from Leaseweb, Haarlem, NL: 626KB/s
Download speed from Softlayer, Singapore: 271KB/s
Download speed from Softlayer, Seattle, WA: 1.18MB/s
Download speed from Softlayer, San Jose, CA: 1.10MB/s
Download speed from Softlayer, Washington, DC: 1.96MB/s
I/O speed : 80.6 MB/s
CVPS - Bufallo
buffalo ~]# wget freevps.us/downloads/bench.sh -O - -o /dev/null|bash
CPU model : Intel(R) Xeon(R) CPU E3-1270 V2 @ 3.50GHz
Number of cores : 4
CPU frequency : 3501.000 MHz
Total amount of ram : 2048 MB
Total amount of swap : 0 MB
System uptime : 12 days, 4:39,
Download speed from CacheFly: 3.51MB/s
Download speed from Coloat, Atlanta GA: 2.38MB/s
Download speed from Softlayer, Dallas, TX: 1.24MB/s
Download speed from Linode, Tokyo, JP: 316KB/s
Download speed from Linode, London, UK: 692KB/s
Download speed from Leaseweb, Haarlem, NL: 591KB/s
Download speed from Softlayer, Singapore: 246KB/s
Download speed from Softlayer, Seattle, WA: 582KB/s
Download speed from Softlayer, San Jose, CA: 652KB/s
Download speed from Softlayer, Washington, DC: 3.32MB/s
I/O speed : 86.7 MB/s
Well at least CVPS doesn't cap at 1MB/s, which I'm getting now for some random reason on DigitalOcean. Bandwidth is never constant for me (1Mb/s -> 60Kb/s -> 400Kb/s in one minute yay)
Well that's what was happening to me until I used virtio, since then it's been fine. Well much better than it was and quite acceptable for me at least. Does not see to work for everyone so not sure what they are doing.
I am not able to create a VM with virtio.
The labs feature is grey for me
Same for me