What feature is missing in the current Platform as a Service? [closed]

Closed. This question is opinion-based. It is not currently accepting answers.

---

Want to improve this question? Update the question so it can be answered with facts and citations by editing this post.

Closed 9 years ago.

Improve this question

I have a project which is to make a "Platform as a Service" environment like Heroku, GAE or dotCloud to name a few.

One of the recurring questions I ask is "What feature is missing in the current platforms ?" Currently most platforms allow developers to deploy their applications (PHP / Python / RoR / JAVA / ...) and manage them by a SDK, 开发者_开发问答a console or an IDE plugin. A few allow some features like : - online IDE, - custom domains, - management of addons, - logs system, - cron tasks, - collaborative aspect, - ...

According to you what is your favorite feature in the current PaaS or, if it does't exist, what would you see in this type of service ?

Thank you for your explanations and your help.

Nico (French developper - hence my spelling mistakes ^^)

---

One of the main problems still not solved in cloud is the security. Every application should have data associated with that. With cloud where to store data? is it secure? can the owners of the system prepare it when some problem happens? ..

Another part is auto scaling. Can the users deploy their application and platform takes care of auto scaling load balancing etc ..

Versioning can the platform support different versions at once.

---

I think the main feature that is missing in all PaaS platforms is that they are scaling by duplication rather than parallelization. In order to scale, common platforms duplicate a worker, a service or an application and then re-aggregate this by deploying a (virtual) load balancer in front of it. Thus, the units of scale are relatively inelastic blocks (a block could also be a VM).

The ideal way of scaling should be smaller units though e.g.. threads, processes. This way scaling would leave this legacy approach towards a real elastic paradigm.

By the way: this way also the multicore-challenge could be addressed. When you have a system with 100s of cores, then probably the sole way for an application to scale is by thread or process that can be distributed over the available cores. Not by duplicating and deploying legacy load balancers.