Comments on: The Next Wave of Hosting: Grid vs. The Cloud vs. What Works http://utropicmedia.net/blog/the-next-wave-of-hosting-grid-vs-the-cloud-vs-what-works SaaS, Managed Hosting, Disaster Recovery, Colocation News and Information Thu, 02 Sep 2010 03:14:29 +0000 hourly 1 http://wordpress.org/?v=abc By: karl http://utropicmedia.net/blog/the-next-wave-of-hosting-grid-vs-the-cloud-vs-what-works/comment-page-1#comment-11745 karl Thu, 30 Oct 2008 15:14:29 +0000 http://www.utropicmedia.net/blog/?p=4468#comment-11745 @Bert - Thanks for the insight. My conversations with Vlad specifically talking about Layered Tech's initial implementation(TGL) revolved around customer feedback on major issues with the lack of failover. This very well may have changed; in the article I tried to focus on how important it is to not have failover as a bolt-on application. VMWare is guilty of this to a varying degrees, as is every Cloud platform I am aware of at this time. SMP APIs for the past few decades have addressed this problem, though it seems Cloud infrastructures have not heeded the lessons. I'm only a single voice, but I would like to see a Cloud platform that at its core is not meant to scale vertically, but horizontally. @Eric - Thanks for taking the time to post. I agree there are a number of problems with getting non-grid/cloud applications to utilize the underlying services, but I think that is a litmus test of the grid/cloud platform. Most any client/server application should be able to work on a Cloud without modification and reap all the benefits the infrastructure provides, including distributed filesystems, which apparently is still an issue. @Bert -

Thanks for the insight. My conversations with Vlad specifically talking about Layered Tech’s initial implementation(TGL) revolved around customer feedback on major issues with the lack of failover. This very well may have changed; in the article I tried to focus on how important it is to not have failover as a bolt-on application. VMWare is guilty of this to a varying degrees, as is every Cloud platform I am aware of at this time. SMP APIs for the past few decades have addressed this problem, though it seems Cloud infrastructures have not heeded the lessons. I’m only a single voice, but I would like to see a Cloud platform that at its core is not meant to scale vertically, but horizontally.

@Eric -

Thanks for taking the time to post. I agree there are a number of problems with getting non-grid/cloud applications to utilize the underlying services, but I think that is a litmus test of the grid/cloud platform. Most any client/server application should be able to work on a Cloud without modification and reap all the benefits the infrastructure provides, including distributed filesystems, which apparently is still an issue.

]]> By: Eric Novikoff http://utropicmedia.net/blog/the-next-wave-of-hosting-grid-vs-the-cloud-vs-what-works/comment-page-1#comment-11744 Eric Novikoff Thu, 30 Oct 2008 00:41:10 +0000 http://www.utropicmedia.net/blog/?p=4468#comment-11744 I will have to take issue with your article. (I can't quite read Bert's reply - it seems to have been mangled by FireFox.) We are a managed cloud hosting provider that have been using AppLogic to serve our customers for 18 months. AppLogic does indeed have an abstracted failover model: any applications running on servers that fail are automatically restarted on standby application servers. This is the identical model to VMWare's failover model. The time to accomplish this depends on various system parameters as well as the memory image size of the application. Where you can run into problems is if you use software which cannot tolerate unexpected downtime (such as I saw today with MySQL on a MyISAM table.) Even AppLogic's data redundancy wasn't able to solve the fact that the restarted MySQL failed to come up due to data corruption. However, this would be true under VMWare as well. What is missing is carefully designed applications that take advantage of the failover capabilities of tools like AppLogic. When we deploy our customers' code to an AppLogic Grid, we work with them to produce restartable deployments, which turn out to survive failures very successfully. I will have to take issue with your article. (I can’t quite read Bert’s reply – it seems to have been mangled by FireFox.) We are a managed cloud hosting provider that have been using AppLogic to serve our customers for 18 months. AppLogic does indeed have an abstracted failover model: any applications running on servers that fail are automatically restarted on standby application servers. This is the identical model to VMWare’s failover model. The time to accomplish this depends on various system parameters as well as the memory image size of the application.

Where you can run into problems is if you use software which cannot tolerate unexpected downtime (such as I saw today with MySQL on a MyISAM table.) Even AppLogic’s data redundancy wasn’t able to solve the fact that the restarted MySQL failed to come up due to data corruption. However, this would be true under VMWare as well. What is missing is carefully designed applications that take advantage of the failover capabilities of tools like AppLogic. When we deploy our customers’ code to an AppLogic Grid, we work with them to produce restartable deployments, which turn out to survive failures very successfully.

]]> By: Bert Armijo http://utropicmedia.net/blog/the-next-wave-of-hosting-grid-vs-the-cloud-vs-what-works/comment-page-1#comment-11743 Bert Armijo Wed, 29 Oct 2008 19:05:06 +0000 http://www.utropicmedia.net/blog/?p=4468#comment-11743 AppLogic absolutely DOES provide failure detection and correction, and has since release 1.0. Data redundancy is included as well, and our disaster recovery suite is making it easy for users to get active-active redundancy with data replication between sites. Your post revolves around providing servers on-demand via virtualization, but that only scratches the surface. AppLogic is unique in that it provides a new abstraction layer, the application, allowing users to package existing application code, data and infrastructure into portable, scalable, instantiable entities. Want 20 copies of a 10 server app for QA - it's one command. Want to backup that 30 server ERP system - it's one command. Need that 100 server web service moved accross the country to a new data center - it's still one command. No human intervention. No code modification. No rebuilds. No risk. With all the hype around lately, it's sometimes tough to cut through the marketing blitz, but there is real technology in the cloud and it is effecting substantial change for users. Bert Armijo 3tera AppLogic absolutely DOES provide failure detection and correction, and has since release 1.0. Data redundancy is included as well, and our disaster recovery suite is making it easy for users to get active-active redundancy with data replication between sites.

Your post revolves around providing servers on-demand via virtualization, but that only scratches the surface.

AppLogic is unique in that it provides a new abstraction layer, the application, allowing users to package existing application code, data and infrastructure into portable, scalable, instantiable entities. Want 20 copies of a 10 server app for QA – it’s one command. Want to backup that 30 server ERP system – it’s one command. Need that 100 server web service moved accross the country to a new data center – it’s still one command. No human intervention. No code modification. No rebuilds. No risk.

With all the hype around lately, it’s sometimes tough to cut through the marketing blitz, but there is real technology in the cloud and it is effecting substantial change for users.

Bert Armijo
3tera

]]>