3 ms·

Death of the Pure IaaS Cloud: Part 2-

25 points by miha123 16y ago

moe 16y ago
First they ignore you. Then they ridicule you...
hapless 16y ago
I'm tired of hearing about the "death" of IaaS. This guy is paranoid. The reason that "PaaS" platforms are multiplying is that they are easy to put together. By deliberately restricting what the customer can do, you can offer cheaper and higher-performing services. Vendor lock-in is a nice plus for the vendor, but that alone wouldn't drive these product offerings. The truth is that we haven't worked out the kinks in infrastructure as a service. Only trivial designs can be implemented in today's public clouds. Judging from vCloud Service Director (or whatever Redwood is called this week), that's not likely to change in the immediate future. Get back to me when "IaaS" products permit the kind of extensive network design customization that your managed service colo host will typically support. If those types of products are supplanted by vendors pushing PaaS, then I'll start to worry.
- cloudsigma 16y ago
  When you say: "extensive network design customization that your managed service colo host will typically support" could you give a couple of examples of the sort of thing you are talking about?
  - hapless 16y ago
    * Customer-managed firewalls * Redundant switching/interior routing * Private layer 2 networks * Bring your own IP space * Redundant exterior routing (e.g. I advertise my prefix through a new provider during an extended downtime) * Out-of-band access
    - cloudsigma 16y ago
      In terms of our platform: * Customer-managed firewalls Yes * Redundant switching/interior routing Yes * Private layer 2 networks Yes we have private VLAN functionality. We'll be expanding this to up to 8 per server shortly. * Bring your own IP space Yes, this needs to be manually added to our BGP sessions but isn't a problem and we do this regularly for customers. * Redundant exterior routing (e.g. I advertise my prefix through a new provider during an extended downtime) We have multiple redundant carriers and they failover automatically. Actually the failover happens instantaneously, even if you are actively pinging you don't even get one dropped packet :-) If you have your own private infrastructure space elsewhere we also are able to choose the preferred carrier too. * Out-of-band access Not sure in what context you are talking about here. If you extrapolate I can give feedback.

25 points by miha123 16y ago

moe 16y ago

First they ignore you. Then they ridicule you...

I'm tired of hearing about the "death" of IaaS. This guy is paranoid. The reason that "PaaS" platforms are multiplying is that they are easy to put together. By deliberately restricting what the customer can do, you can offer cheaper and higher-performing services. Vendor lock-in is a nice plus for the vendor, but that alone wouldn't drive these product offerings. The truth is that we haven't worked out the kinks in infrastructure as a service. Only trivial designs can be implemented in today's public clouds. Judging from vCloud Service Director (or whatever Redwood is called this week), that's not likely to change in the immediate future. Get back to me when "IaaS" products permit the kind of extensive network design customization that your managed service colo host will typically support. If those types of products are supplanted by vendors pushing PaaS, then I'll start to worry.

cloudsigma 16y ago
When you say: "extensive network design customization that your managed service colo host will typically support" could you give a couple of examples of the sort of thing you are talking about?
- hapless 16y ago
  * Customer-managed firewalls * Redundant switching/interior routing * Private layer 2 networks * Bring your own IP space * Redundant exterior routing (e.g. I advertise my prefix through a new provider during an extended downtime) * Out-of-band access
  - cloudsigma 16y ago
    In terms of our platform: * Customer-managed firewalls Yes * Redundant switching/interior routing Yes * Private layer 2 networks Yes we have private VLAN functionality. We'll be expanding this to up to 8 per server shortly. * Bring your own IP space Yes, this needs to be manually added to our BGP sessions but isn't a problem and we do this regularly for customers. * Redundant exterior routing (e.g. I advertise my prefix through a new provider during an extended downtime) We have multiple redundant carriers and they failover automatically. Actually the failover happens instantaneously, even if you are actively pinging you don't even get one dropped packet :-) If you have your own private infrastructure space elsewhere we also are able to choose the preferred carrier too. * Out-of-band access Not sure in what context you are talking about here. If you extrapolate I can give feedback.