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.
Is 'cloud' worth it? What options are out there?
I've seen a few VPS providers throw around the term 'cloud', and just wanted to see if there is a general consensus about what 'cloud' actually is and are there features that make it different to a standard VPS? ( EG. Live migrations, powerdown halting billing, etc).
What do you consider to be non-amazon cloud options out there?
Comments
Some people say cloud for hourly billing, multiple locations etc.
Real cloud I would say is HA & Redundant Infrastructure.
It's literally marketing. Since there's no "set in stone" definition of Cloud.
I personally believe a cloud has the following features:
SolusVM is never a cloud, for the above reasons.
This rules out a lot of shifty "VPS" panels and providers from being Cloud.
I consider cloud a real iaas solution with pay as you go.
While ours has a lot of extra features such as load balancing, IPSec VPN for isolated networks, many bells and whistles, a minimal cloud should have HA/failover (they are not the same, must have both), hourly billing, ability to run arbitrary OSes with arbitrary storage/partitioning and arbitrary ram/cpu within reason (ours cannot run with arbitrary RAM and CPU cores yet). Multiple locations are not mandatory IMO, just nice to have.
I consider scams, even though many big providers use it, a much higher price for using more than the allocated resources (for example traffic being counted over a certain limit at 10 times the original price or more) as well as counting IOPS or cpu cycles and billing similar overages at shameless prices, clearly luring users into a trap with pre-approved drafts for billing.
The real cloud includes, generally, self-healing and redundancy. Example that I know of would be VirtuaClub.com.
A highly skewed VM, such as extremely high CPU load with minimum memory consumption or vice versa will waste the remain resources, and I doubt any provider will provide such a plan economically.
My criteria for the cloud is virtual network, available zone and API.
The load-balancing and fail over can be achieved with ha proxy and keepalived if not provided. The custom can fully control the virtual network.
Available zone guarantees that the VMs are not in the same host, no hardware single point failure.
API make the automation much easier.
The self-healing is and auto scale out is nice to have, but most likely for a stateful service is very hard: for example, how could you scale-out a db instance?
Hum, we do provide that, custom KVM:
https://my.iperweb.com/cart/customized-kvm/ With enough nodes you can balance out.
NIST and IEEE definitions basically boil down to the following negative definition (paraphrasing):
"If you've got to pin a machine in some inventory/accounting ledger, then it isn't a cloud."
Sadly, it has to be a negative definition since if we approach it by feature/tech - it creates too much overlap to existing realms.
High availability / live migrations 'features' are orthogonal to what a cloud is, as you can very well do either with a fixed infrastructure. VPS are not cloud even though they use the same hypervisors merely because you buy/lease one in much the same fashion as a dedi. Colos however can be turned into private clouds if the very use of them internally is fluid thanks to an orchestration layer like openstack.
The IEEE definitions also move up the chain into PaaS and SaaS although to get into it all is beyond the scope of this thread. http://cloudcomputing.ieee.org/standards
Depends on your needs really and the devops capacity you have. Amazon, Rackspace, HP, Azure, etc are geared towards large-spend clients that have in-house expertise to orchestrate multiple VMs. Those clients demand real uptime and performance are willing to get their hands dirty to eek out extra nines.
If you're a reader of WHT or LET, you're probably in the single-server realm so a provider running Cloudstack/OnApp on a SAN is probably closer to what you want. Uptime is okay, as is performance, especially in light of it being hands-off. Digital Ocean and other hourly on-demand providers fill a niche for developers that need to spin up fleets for testing / continuous integration hence the lack of NaaS features.