We are in the process of migrating this forum. A new space will be available soon. We are sorry for the inconvenience.

Hosting Update


Thomas
02-15-2013, 04:02 PM
Hello,

It's been almost one year since the shared hosting that we propose is not at its best. We noted that in the all cases it is a CMS or a Prestashop, which have a lot of requests to view a page. In average problems begin with 1000 requests needed to display
a page.

Technically requests are not a problem, but it takes time.We worked on reducing time of each request. In our case we have reduced all time except the time between "web server / php + the file server. "

For this, we have added many many resources, optimised settings, transferred sites between file servers, etc. Despite performance increase were not meeting you.

Since early January we prepare therefore changing the type of load balancing:

- Currently a site is hosted in shared hosting distributed on all servers in the cluster. in parallel it is treated by 1000 HG type physical servers . It is good for large sites who benefit from lots and lots of computing power. But it is not good for sites that have few visitors per day. For each visitor statistically speaking falls on a server different from the previous visitor. It will not benefit from caching the visitor above has established and therefore will take again several seconds to generate a page.

- The new load balancing site redirects on all servers
in parallel, but each site is redirected to the same server. Ie a site is generated by a given server, and for all visitors. The consequence is that we must see accelerating sites, except for the first visitor. Visitors will enjoy the following cache generated by the first visitor (and following). The consequence is also that very, very large sites will no longer be housed on Hosting. It is not worse because we're talking about sites that are several million visits per day. Why it becomes impossible? Because one server can not keep a
such site. So we will have a natural limit to a site related to the real power of a server.
Warning: we're talking about very, very large site requiring the equivalent 3-5 servers HG
300/ month room .. For these customers then we have created a Special small-cluster contain them within each cluster. Currently talking 10-12 sites per cluster, a small 100 large sites that used 40% of the total resources .. They were redirected to the sub-cluster which had to resulting in reduced load for all and better performance last night.

http://travaux.ovh.net/?do=details&id=8039

The work is not finished but will in the right direction. And that is the purpose of the email.

We continue the development of the hosting platform. We have quite a delay on
news because we have all resources on upgrading from shared hosting.
We are behind on mutu 2013 increasing resources SQL. We test currently techno "geo cache" on 1000gp: CDN close this technology will be included default and allow the mutus improve the performance of your sites and accelerate the exchange visitor
in Europe, USA and Canada. We have full other projects in the pipeline that will
begin to reopen soon ..

I know we have tested your patience and I thank you for your confidence.
The platform currently hosts 1.1M login and almost 2M sites. We have 7000 servers
HG and turn the difficulty is in is to propose a general platform performing any type of site. This is that there has not been good for 1 year and we think it is long way to sharpen our Hosting any kind of website.

Outside Hosting, we have GAMMA VPS 2013 with a v6 API. We work
the manager v6 that will replace the v3. And we propose the mutu with private
resources guarantees BETA end of March.

Regards
Octave