Linode currently offers far more bandwidth for your money and pools each allocation amongst all of your servers.
In my situation Linode would be offering me 48TB of pooled bandwidth for the same money DO is offering me 5TB x 6 servers.
This also makes running a load-balancer at DO completely unreasonable.
If you guys actually expect me to use the bandwidth, it shouldn't matter which server gets the heaviest load.
As far as I am aware, and based on reading other forums etc DO will / do pool bandwidth between all droplets. Admins, is this correct?
Attachments Open full size
@woblit I submitted a support ticket regarding this and was told that each droplet will have it's own bandwidth (no pooling)
Attachments Open full size
@Bogdan - I think they might find that a bit difficult to do when they have already stated (at least one of their employees has stated) that bandwidth will be pooled. For more information see:
https://www.digitalocean.com/community/questions/bandwidth-pooling
I hope they stick to this. Would be bad form not to, as this is the basis that I have been working on.
Attachments Open full size
I will not build a platform on Digital Ocean only to have the rug pulled out from under me if they decide not to offer pooled bandwidth when they start tracking it.
I've asked support this before creating this suggestion:
Hello,
We do not currently offer pooled bandwidth plans. We often build up our features based on user requests. You can submit and vote for features you would like to here:
http://digitalocean.uservoice.com/forums/136585-digital-ocean
Thanks,
Support
Attachments Open full size
Definitely voting for his one, I have a fair amount of infra already, and collectively have 74TB bandwidth available, pooling this would primarily give me piece of mind that what i'm paying for i'll be able to have full access to. - I'm progressively adding more and more infrastructure, so obviously this bandwidth allocation will increase significantly.
Attachments Open full size
I've noticed that there are a few questions and other messages out there saying that it will be offered, or it is offered, or it isn't offered, and it seems like there's no definitive authoritative answer, if they could just add the true answer to the FAQ then it would either validate, or deprecate this request.
Attachments Open full size
One of the things that is definitely required before i can set up a load balanced cluster at DO. Any comments from the DO staff?
Attachments Open full size
While I don't think DO is like many BS shared hosts that claim you get so much, but if you use it they kick you off, I assume DO is betting that many users won't use all the bandwidth included in a package. If people used ALL their bandwidth always, I would expect the amount they give you per a package to lower, or the price to go up.
What I'm getting at is, with pricing so low, I do think they are betting that most users won't be using their allotted bandwidth (I assume from statistics, this is a very safe assumption. I know I've never used more than 10% of my bandwidth on any VPS I've run).
Just guessing though, but I think DO is trying their best to offer a competitive price, as well as bandwidth. I'm assuming pooling would be hard to do at this price point. And they are likely reluctant to do so since they don't want to raise the price, or lower the initial allotted bandwidth.
But, maybe there is a happy medium. Such as, have half (or some other proportion of) the bandwidth to be pool-able.
Just throwing the idea out there. Its not fair to compare with linode when linode's prices are so much higher. And DO's quality is still up there with linode's... the price difference doesn't come out of thin air.
Attachments Open full size
The question was on pooled bandwidth, not overage costs.
Attachments Open full size
You haven't answered the question. It's not about individual droplet bandwidth, it's about pooling the bandwidth we've purchased between droplets. For example, I'd love to use your private network to connect several back end servers with a web tier, and then use small front-end droplets to load balance between the web tier. I'll have bought considerable bandwidth from you - which I can't use, because the heavy-lifting is being done in the back end tier, and I can't route that bandwidth through the front end without paying overage.
It seems like the answer to Mr. Cremin's question is "we actually don't expect you to and furthermore know you can't use the bandwidth, and we're counting on it". I'm sorry to hear that. As has been pointed out by others, it dramatically increases the footprint of any application with a load-balanced front end / private back end architecture. Sadly, that means I won't be considering DO for most of the applications I had hoped to deploy here. I respect that this is DO's business decision, but it is disappointing.
Attachments Open full size
If what you say is true (i.e. people pushing bandwidth), you should have no problem crediting people's accounts for the bandwidth that's unused on each droplet, which they can then in turn allocate to their load balancer.
The pooling is the issue here, not the price!
Attachments Open full size
Oh the feature talks about sharing the bandwidth not the coast of overuse! Please consider this feature, Thanks!
Attachments Open full size
I interacted with someone at DO over Twitter last night:
https://twitter.com/jedgar/status/441314391301296128
> I just talked to our CEO, it will be pooled and it looks like billing for it is fairly high on the roadmap.
Attachments Open full size
I too would very much appreciate pooled bandwidth billing once you do start charging for bandwidth. Without it, I'm a afraid you may lose a good amount of business. The whole point of a cloud infrastructure (at least to me) is to make it easy to deploy (you've nailed it here) and scale your applications. Without pooled bandwidth on your service, it doesn't make sense to do any type of proxy based load-balancing. Yes you could do it at the DNS level, but honestly its not as reliable as it should be (not all isps will honor a short ttl, and a huge influx win a short period could easily overload a node). setting up an ha proxy in front of a few servers along with round-robin / geo-dns load-balancing is a much better option.
Also maybe you could even implement your own load-balancing into the cloud?
Attachments Open full size
Hi,
I agree with all the comments here, How could be set a load balancer ? All the droplets' bandwidth would be "lost"....
Attachments Open full size
I would very much like to see pooled bandwidth added - this would make running infrastructure on DO feasible vs. individual servers.
Attachments Open full size
This just sounds demanding, it's really garbage, and 6 servers 30TB is loads; you guys need to make bank as well as provide services; maybe you'll make more money telling clients like this to jog on, and focusing on providing awesome offers for those customers that don't need 30TB of data (most web-hosts offer 50-100GB, or make it unlimited but slow, so I'd rather have 30TB over 6 servers that are actually responsive)
Attachments Open full size
It does look like it will be pooled and start to be billed soon. Hang tight guys. It would be nice to have this by the end of July!
Attachments Open full size
I do not know if this will be available by the end of July. It is not in Planned yet.
Attachments Open full size
Guys... "Gathering feedback" for 2 years. Serious-fucking-ly?
Attachments Open full size