Not to me; if you have a problem which NodeBalancer can solve, then it's likely that you already have multiple Linodes and an additional $19.95 will be a relative drop in the bucket, especially given what you get from it.
If your stack requires a load balancer, you would have to pay $19.95 for a dedicated 512 node that you would need to then configure and manage your own load balancer. This new service is also managed, distributed, and has an API. $19.95/month seems like a steal to me.
True. I was intending to draw a comparison to what kind of load balancing you can get on Linode for $19.95 if you were to do it yourself. I highly doubt it is possible to roll a superior load balancer service on your own for less than NodeBalancer.
I was recently trying to figure out how I was going to handle this on Linode. I am very glad to see this new service. If NodeBalancer has the ability to fail over to other data centers in the instance of a data center wide outage then I'll be set.
Twenty bucks a month is nothing. It's pretty hard to find any hardware load balancers for less than $1500 each. At my last job I wrote a hardware proposal that called for redundant $80K load balancers. Though I'm sure that the NodeBalancer can't match the feature set yet of the high end machines I sure like this trend.
Most likely they would be using Nginx. I find the offering interesting because most of the companies in need of a load balancer would be able to configure Nginx load balancer quite easily. Not knowing underlying technology and not being able to go deep into modifying the algorithms and other aspects of load balancer makes me slightly uncomfortable, so why should we go with it? The only positive aspect of this offering is automatic health check but Nginx does that already.
I think this offering is nothing more than a GUI for generating nginx configuration files.
Most people able to launch and configure Amazon AMIs could configure Nginx too. Yet Amazon offers Elastic Load Balancers. There's a huge market for stuff like this -- just because you can muddle your way through configuring an instance doesn't mean you want to be in charge of configuring and running all the moving parts.
It's a custom implementation. They were vague on giving any details, but if I recall correctly it was written in Python. So probably a Twisted, Gevent, or Eventlet implementation.
I've set up high availability load balancers on linode before. Basically two 512mb vm's running heartbeat between them and both running the same config of haproxy. This seems like a spiffy interface to something like that minus the heartbeat.
What is going on here under the hood? Anybody know? Haven't seen any deep details yet. Then again, I haven't logged into their irc either.
I just happened to have been in my linode control panel a lot the past few days and noticed the new tab. I don't think I could have clicked the link any faster. I had to do a double take when I saw 19.95. I almost thought they meant per day.
What's the big deal? I'd really like to know too. I'm not just asking that to be snarky. :-)
AWS and a few other hosts[1] have had this for a while. Is Linode doing anything differently, or is this just an instance of another (big name) hosting company joining the fray?
This isn't a big deal in terms of being revolutionary. I doubt they're doing anything differently than anyone else...load-balancing web servers is a pretty well-solved problem at this point. What's a big deal is that Linode already has many fans around here and this makes their product better for a significant portion of their users without adding a lot of cost.
Joe, if you look at the Howto it shows when you setup the balancer you assign it to a data center (like NJ) and I assume it balances nodes only in that data center otherwise what would be the point of assigning the BalancerNode itself to a data center?
25 comments
[ 5.4 ms ] story [ 93.8 ms ] threadI was recently trying to figure out how I was going to handle this on Linode. I am very glad to see this new service. If NodeBalancer has the ability to fail over to other data centers in the instance of a data center wide outage then I'll be set.
So, I'd say linode's price seems quite reasonable. :)
1. working out how to do it yourself
2. ironing out any kinks in your homegrown solution
3. worrying about it 2am.
Totally worthwhile.
(edit: formatting)
=]
I think this offering is nothing more than a GUI for generating nginx configuration files.
Possibly they use Nginx for the HTTP balancing only.
What is going on here under the hood? Anybody know? Haven't seen any deep details yet. Then again, I haven't logged into their irc either.
One less thing for linode sysadmins to worry about. This is a good move, up there with offering bullet proof DNS services in my opinion.
Edit: shame that the API doesn't look like it supports alarms/thresholds.
AWS and a few other hosts[1] have had this for a while. Is Linode doing anything differently, or is this just an instance of another (big name) hosting company joining the fray?
[1] like contegix: http://www.contegix.com/infrastructure/load-balancing.php
At least that is my thinking.