Results 1 to 14 of 14
-
07-08-2012, 09:58 AM #1WHT Addict
- Join Date
- Jan 2012
- Posts
- 136
Is colocation better? If yes, how do I choose the best infrastructure?
Hello everyone. I have an upcoming company providing load balanced HA hosting services to a primary target market of medium-large enterprises and to a secondary market of individuals-small businesses.
It is an OpenStack deployment providing a wide range of selections for clients. From clusters with any hardware of their choice, dedicated servers in the same way, VMs in an existing HA cluster or high performance shared hosting. A unique feature is the load balancing HA with a 100% uptime SLA while being extremely cost effective.
With a startup budget of approximately $350,000, I have considered RackSpace and SoftLayer. But they seem to be an unnecessary investment. Startup with a colocation in a private suite with the same networking capabilities of RackSpace and SoftLayer is just as a great choice.
As you might take into consideration, management of the servers is not a concern here.
But I'm having a hard time choosing my hardware. If anyone has the experience, I'd love to get some advice for:
Webservers
Database servers
Load Balancers
Firewalls
Switches
SAS 15k drive brands
Servers should be MADE for performance, looking forward to Quad Processor servers.
Thanks for your advice.
-
07-08-2012, 12:50 PM #2The Guru!
- Join Date
- Nov 2007
- Location
- India, USA and Amsterdam
- Posts
- 2,581
If managing infrastructure isn't a problem, then collocation would be a cheaper solution in the long run!
If you are planning for some mission critical cloud or HA cluster, I would recommend you to use some software that has some paid support incase your infrastructure goes down due to bug (VMware, Applogic, OnApp etc) you can get back your infrastructure back online quickly.
Quad processors might be overkill for offering cloud/cluster. Most of the time saturation takes place in RAM and I/O. Dual Hexa core processors should be a right balance between performance and price.
Just my 2 centsLast edited by nix101; 07-08-2012 at 12:58 PM.
-
07-08-2012, 01:14 PM #3Web Hosting Master
- Join Date
- Apr 2004
- Location
- Singapore
- Posts
- 1,234
Hi,
For a turn key or validated deployment, check this: http://media.netapp.com/documents/ds-3105-flexpod.pdf
VMware vSphere + Cisco UCS + Netapp Storage.
Fully HA, everything virtualized, enterprise ready.Alan Woo, alan [@] ne.com.sg
= NewMedia Express Pte Ltd (AS38001)
= IP Transit, Colocation & Dedicated Servers in Singapore | Hong Kong | Tokyo | Seoul | Jakarta |
= Singapore Speedtest speedtest.sg
-
07-08-2012, 05:43 PM #4Always there
- Join Date
- Jun 2004
- Location
- Europe
- Posts
- 3,822
Deploying and running, as well as developing on Openstack is very human resource intensive, so i do assume you have experienced system administrators as well as a programming team and a project manager running your Openstack solution.
A budget $350 000,- is not a lot since you need to deploy a test cluster, make sure to do the necessary (Automation, API and security) developments and once you are ready to setup a production environment, your team should know all the ins and outs of Openstack. This should include the knowledge on the best hardware platform to roll out for the production environment.
That is a nice Virtualization V2.0 platform but not a true cloud platform. Did you even take a look at Openstack and tried to compare the solution above with what openstack offers?Last edited by swiftnoc; 07-08-2012 at 05:49 PM.
█ Swiftway.net Your Business deserves our Quality - Experts on Hand since 2005. Europe & US locations, we operate our own network AS35017 Support response time <15 minutes 24/7
█ Introducing our new Entry level server line ! Support response time <15 minutes 24/7. Technology Fast 50 & Fast 500 award winning for multiple years, Your Business deserves Swiftway Quality.
-
07-09-2012, 10:06 AM #5Web Hosting Master
- Join Date
- Apr 2004
- Location
- Singapore
- Posts
- 1,234
Yes, openstack can do a lot more, I understand vmware solutions is just a subset of what openstack cover.
vmware with vcloud director and vcloud api can achieve what he want to offer as well, a cost effective load balance HA hosting with 100% uptime and SLA.
from business perspective, I will choose a turnkey/validated deloyment which inherent less risk, compare to openstack deployment which subject to more variables.
And since he have a hardtime on choosing a combination of hardware, an validated deployment will remove a lot of integration and compatibility issue.Alan Woo, alan [@] ne.com.sg
= NewMedia Express Pte Ltd (AS38001)
= IP Transit, Colocation & Dedicated Servers in Singapore | Hong Kong | Tokyo | Seoul | Jakarta |
= Singapore Speedtest speedtest.sg
-
07-09-2012, 10:58 AM #6WHT Addict
- Join Date
- Jan 2012
- Posts
- 136
-
07-09-2012, 11:16 AM #7Web Hosting Master
- Join Date
- Oct 2006
- Location
- US/EU/UK
- Posts
- 4,886
I'd not suggest you to consider offering services with "with a 100% uptime SLA" despite that you might set up an infrastructure that allows you to deploy HA cluster. The theoretically possible uptime is not higher than 99.995% on an infrastructure level.
HostColor.com ★★ Edge Infrastructure - US Dedicated Servers & Europe Dedicated Hosting ★ since 2000
In 50 U.S. Edge Data Centers & 80 POPs worldwide
24/7 Support ★★ Support Tickets - LiveChat - Phone
-
07-09-2012, 06:57 PM #8Temporarily Suspended
- Join Date
- Mar 2011
- Posts
- 122
Really what makes you say that I have a internal nova and swift cluster and while setup isn't automated it's really no big deal from a provisioning/maintainability POV
$350k is certianally reasonable your test custer onyl needs to be a proxy a switch, a hv and a handfull of nodes (~$5-10k) and if you know some basic principles of how openstack works and what your load may be it's easy to pick a suitable server base.
-
07-10-2012, 01:49 AM #9Always there
- Join Date
- Jun 2004
- Location
- Europe
- Posts
- 3,822
The topic starter seems to wants to code additional functions in Openstack and likely wants to provide a control panel to clients as well. I got this feeling from reading the original port.
Specific:
It seems reasonable to assume, that programming resources are required.
Sure, $350k is reasonable to provide for test cluster hardware, but $350k is a low budget if you need to hire system administrators, develop functionality in Openstack, develop a control panel and roll out a Openstack production environment on top of building a test cluster.█ Swiftway.net Your Business deserves our Quality - Experts on Hand since 2005. Europe & US locations, we operate our own network AS35017 Support response time <15 minutes 24/7
█ Introducing our new Entry level server line ! Support response time <15 minutes 24/7. Technology Fast 50 & Fast 500 award winning for multiple years, Your Business deserves Swiftway Quality.
-
07-10-2012, 05:30 AM #10WHT Addict
- Join Date
- Jan 2012
- Posts
- 136
Refer to the Atomia Platform [http://www.Atomia.com/]. Exactly what I'm looking for, already integrated and designed to be used with OpenStack.
What kind of a budget do you recommend from your perspective of an OpenStack system and why? We do have a flexible budget, and this shouldn't be a big problem.
-
07-10-2012, 11:39 AM #11Web Hosting Master
- Join Date
- Apr 2004
- Location
- Singapore
- Posts
- 1,234
If you have 2 identical web server serving identical static content, theoretically this can achieve 100% uptime in typical scenario even one web server go down. Base on this concept, with certain redundancy build in, it is achievable to provide 100% uptime on infrastructure.
-
07-10-2012, 03:23 PM #12Junior Guru Wannabe
- Join Date
- Jan 2012
- Posts
- 76
-
07-10-2012, 03:27 PM #13WHT Addict
- Join Date
- Jan 2012
- Posts
- 136
Separating the datacentres and running cron jobs across them is a hard job.
Though in theory due the redundancy of the infrastructure and the availability of multi-homed bandwidth, 100% uptime is possible (in theory).
As you said, power of the datacenter might affect the cluster. It is a tier IV datacenter with the least amount of downtime possible and the most failover options for power.
Though if a few customers demand it, I can for them have servers placed in another datacenter undergoing global load balancing. This (if done) would always be on request and customers have to understand that syncing between the servers can be time and bandwidth consuming. And using bandwidth every 10 minutes for a cron job isn't really something a server owner wants.
-
07-10-2012, 03:47 PM #14Always there
- Join Date
- Jun 2004
- Location
- Europe
- Posts
- 3,822
Sounds good in theory, in practice situations do occur that take both feeds in a rack down. Example:
A faulty PSU trips break on PDU 1 and the resulting spike in power on PDU 2 trips the breaker of that PDU.
Rack down.
Good idea, use an Anycast DNS service that provide realtime monitoring and you should be set.
This is a factor that will eat into the initial investment, building a fiber infrastructure (IRU, best relative return) or to rent waves between the datacenters - this should allow data replication between the physical locations at relatively small costs.█ Swiftway.net Your Business deserves our Quality - Experts on Hand since 2005. Europe & US locations, we operate our own network AS35017 Support response time <15 minutes 24/7
█ Introducing our new Entry level server line ! Support response time <15 minutes 24/7. Technology Fast 50 & Fast 500 award winning for multiple years, Your Business deserves Swiftway Quality.
Similar Threads
-
Choose one ColoUnlimited or HE Electronice for Colocation
By SuperbLayer in forum Colocation, Data Centers, IP Space and NetworksReplies: 1Last Post: 10-06-2011, 07:07 PM -
New Infrastructure
By Serverplan in forum Colocation, Data Centers, IP Space and NetworksReplies: 13Last Post: 04-03-2008, 05:40 PM -
Top End Colocation Infrastructure - Month to Month commitments
By Jaybirdsmith in forum Dedicated Hosting OffersReplies: 0Last Post: 10-25-2005, 11:31 AM -
Infrastructure Advice
By tbroeker in forum Hosting Security and TechnologyReplies: 6Last Post: 10-05-2003, 01:14 PM -
Another reason not to choose United Colocation
By dandanfirema in forum Dedicated ServerReplies: 17Last Post: 01-06-2003, 04:33 PM