[ale] Alright, it's time to move on from Linode

[Jeremy T. Bouse]

On 1/8/2016 7:34 PM, Justin Caratzas wrote:
> On 1/8/16 7:23 PM, Jeremy T. Bouse wrote:
>> On 1/8/2016 5:39 PM, James Sumners wrote:
>>> On Fri, Jan 8, 2016 at 1:13 PM, chip <chip.gwyn at gmail.com
>>> <mailto:chip.gwyn at gmail.com>> wrote:
>>>
>>>     Take a look at Vultr.com, can do it there.  They have hosting in
>>>     Atlanta too.  They're basically the economy choopa stuff.  
>>>
>>>
>>> That's looking rather nice. $5/mo for 1TB of transfer and plenty of
>>> resources for my needs.
>> Not that I have any horse in the race or anything, but as a cloud
>> service consumer here's a few of my observations...
>>
>> First off, I have/currently use LInode, AWS and DigitalOcean... Mainly
>> for one simple reason, all 3 providers have good support with SaltStack
>> so I don't actually have to log into their UI to do anything to manage
>> my servers from cradle to grave.
>>
>> I will say I did look at Vultr and they do have some nice features and
>> it does appear that Apache libcloud [1] does have support for Vultr
>> which would make a SaltStack salt-cloud driver realistically possible
>> though doesn't currently exist. I was really floored by their benchmark
>> comparisons [2] and how much it was apples and oranges. I loved how they
>> compare a 768MB/1CPU Vultr system for $5/month against a 3.75GB/2CPU AWS
>> C3.Large that will run you around $78/month on-demand or between
>> $29-54/month depending on reserved instance pricing or their 2GB/2CPU
>> Vultr system for $20/month against the 3.75GB/1CPU AWS M3.Large with run
>> costs abount $99/month on-demand and
>> $39-71/month reserved instance. Comparing against an AWS T2 instance
>> (nano 512MB/1CPU or micro 1GB/1CPU) would have seemed like better
>> candidate for comparison against the 768MB Vultr and runs closer
>> ($5/month t2.nano or $10/month t2.micro on-demand or $2-4/month t2.nano
>> or $6-7/month t2.micro reserved instance). Likewise a t2.small or
>> t2.medium would have been better comparisons for the 2GB Vultr. It
>> looked like they went out of their way to pick the most expensive option
>> to compare so their numbers looked better.  I found a blog [3] that
>> seemed to give a better comparison in fact.
> Slight disagreement, I believe the t2.* are terrible machines to
> benchmark, given the cpu bursting budget. m3/4.mediums would have been
> the better comparison, the Cs are a bit nuts w/ pricing.
Yes, the t2 instances are burstable but they are better than the older
generate t1 instances. If you're comparing cost however the t2 would be
a better comparison as the specs are closer as is the cost. When you're
comparing a $5 instance to a $78 instance your "Performance per dollar"
is obviously not going to be comparable. The C3 instances are more CPU
optimized instances, the M3 and M4 are more general purpose with
balanced CPU & memory with the M3 being SSD-based instances which is
really the only comparison against DO or Vultr with the minimum in the
series being the m3.medium which has 1 CPU and 3.75GB RAM and 4GB SSD.
> How do you like libcloud? I've been meaning to check it out.
I haven't worked with it directly myself. Many of the salt-cloud
provider drivers are written utilizing it as it provides a quick method
to do so. There are still many drivers that have libcloud support
available but still don't utilize it. In most of the cases the drivers
were written prior to libcloud support and hasn't been any real need to
re-write them yet. I'm currently working with another cloud provider
that doesn't have libcloud support so we're having to do a lot more of
the work going off API documentation from the provider as the only API
library we've been able to find for it is not fully up to the task.
>> Otherwise the pricing between DO and Vultr doesn't appear to really be
>> all that difference comparing plans either. That said I may have to
>> check out Vultr and see if I can't get the salt-cloud driver working.
>> Cost being low enough I wouldn't mind throwing some money at it to get
>> another cloud provider option made available to me. I like having the
>> ability to launch and deploy my hosts to any SaltStack supported cloud
>> provider for a DR/BC perspective and keeps me from being locked into any
>> one provider. Then again I'm not worried about uploading custom ISO
>> images and if I were I'd simply build and deploy those to AWS where I
>> could easily make my own AMI offline and knowing how to work AWS to be
>> cost comparative wouldn't bother me.
>>
>> 1. http://libcloud.readthedocs.org/en/latest/compute/drivers/vultr.html
>> 2. https://www.vultr.com/benchmarks/
>> 3. http://blog.due.io/2014/linode-digitalocean-and-vultr-comparison/

-------------- next part --------------
A non-text attachment was scrubbed...
Name: smime.p7s
Type: application/pkcs7-signature
Size: 4807 bytes
Desc: S/MIME Cryptographic Signature
URL: <http://mail.ale.org/pipermail/ale/attachments/20160108/095ab6d5/attachment.p7s>