Do I need more than Proxmox? I mean: a Cloud Computing...

[wmunguiam]

Hello dear friends,

I am a network engineer, working with Routing Protocols, VPNs, IPv6, multicast and monitoring/managing some application servers like Videostream/Videoconference servers.
I Work for a consortium that has a presence in several countries and maintains an efficient IP connectivity between all.

We are planning to develop a project consisting of:
-Installation of five servers
-Each server will be located in a different country(PoPs)
-Each server will host many virtual machines(based on KVM/Xen) (We are defining if: XenServer,VMwareESX or Proxmox!)
-In principle, those server will conform a cluster.
-Virtual machines will serve Videostreaming and some website servers.
-Is very important for me to have Virtual Machine motion feature between servers ( not necessary in live mode)

My query is:
-According to the scenario described, a solution of Cloud computing will improve the performance of all of these facilities? Or Proxmox can cover this requirements?

PD: Sorry If I am confused with the term "Cloud computing", but what I am looking for is some solution for management virtual Machines.

 

[fortechitsolutions]

Hi,

Just my 2 cents worth,

IMHO the term cloud computing is partially worthless due to the excessive 'buzz' on the topic; virtually anything can be called 'cloud computing' these days (at least at a marketing level).

In the 'purest' sense, cloud computing might be considered .. a form of offsite managed virtual server platform / well suited for highly scalable platforms with very 'peaky' resource requirements; where "you the consumer of cloud computing services" don't have physical access to the hardware providing the virtual server platform (nor do you need/ want / benefit from such). Part of the benefit of the model of 'cloud computing' is abstraction of capital cost & associated maintenance-related costs needed to meet resource requirements for 'peaky' service requirements. Typically a 'cloud platform' will be amenable to scriptable / automated deployments; and is likely best suited for services which scale 'horizontally' inherently / are designed to scale in this manner.

So for you,

- you plan to own your own hardware
- you want to manage it, and offer services from it
- effectively you are setting up a 'private cloud' - more or less server virtualization environment spanning many physical sites.

If you *really* trust your network between points, you might choose to make a 'single logical virtualization cluster' spanning all sites. This demands reliable (redundant?) network; "trusted pipes" (i.e., not insecure data transmission via public internet for management network ideally)

Possibly an 'easier' model would be,

-5 physical sites
-2 servers per site, running 'cluster virtualization environment' of 2 nodes: each site is a mini-cluster facilitating live migration, failover, etc between the 2 machines at that site
-system images could be migrated between physical sites if desired, either via offline migration or manual migration methods
- consistent platform across all 5 physical sites simplifies management burden
- assuming your VM's do not require 'huge' CPU requirements, then .. the capital cost of 2 x single-socket quadcore boxes .. is insignificantly different from 1 x dual-socket quadcore box. Or maybe your budget is such that dropping an extra few $k for 2 servers per site is a non-issue in the bigger picture of costs, maintenance, services:capital influx etc.


I'm not sure that a 'cloud' spanning all 5 sites could enhance performance in any way (in terms of VM throughput or capacity). Possibly the only / primary benefit would be entirely unified management of all systems from a 'single point' - and that VMs would be live-migrate capable between all sites (in theory). In practice, Live-migration depends on centralized SAN style (iSCSI or fibre) storage (or possibly NFS server) - and inherently needs excellent bandwidth from this shared storage to *all* virtualization hosts in your cluster; so unless you have nice solid gig-ether network between all 5 of your physical deployment sites (or really long fibre optic SAN cables - I'm not so sure I would be keen to recommend this route!

Anyhow. These are just some off-the-cuff comments and observations. I'm sure others will have something to say as well, possibly in an entirely different direction from the comments I've made here


Tim Chipman
Fortech I.T. Solutions
http://FortechITSolutions.ca

 

[corsari]

Any update on this foggy argument?

Thank you

R.

 

[adder_uk]

cloud = host
it's just a computer!
it could be anyway; more likely to be in a building/somebodies garage/cupboard than in the clouds! ;-)

 

[corsari]

Ok.

And what about delivering services, shares and services in parallel.

Made by two servers, I mean. Not a cluster made by primary and secondary server. A system where the servers are both primary as one unique.

If one dies nothing happen, just a warning. No interruptions and no migration need, this because they should work parallel...

Hope that gives the idea.

R.

 

[fortechitsolutions]

(I'm posting this reply here - I got an email query directly from R.Corsari but since the email and posted query are similar vein, I'm posting my response here, in case it is of any interest / value to anyone else. --Tim)

--paste--

Hi,

many virtualization platforms allow "live migration" of VMs with minimal downtime... so in such a case, it is possible to have a management layer which reboots a downed VM in the event of a hardware fail that takes a VM out of service. In such a case your downtime is limited to the reboot time of the VM, plus some 'detection of failure" timeout.

VMWare and Citrix XenServer, for example, both offer this kind of function - in their commercial flavours. XenServer does support LiveMigration in the free edition, but NOT the 'monitoring and automatic restart" - it is up to the system admin to bring it online themselves . VMWare also has a "high availability" feature where you can have 2 identical VMs up and running 'in lockstep' but one is 'not visible to the public' unless the other VM goes down.. in such a case, you have 'effectively zero' downtime; no interruption of service (since no reboot..) and active sessions continue as if there was no change. But this is very much (!) a non-free feature in VMWare.

ProxmoxVE supports live migration of KVM based VMs; but I don't believe there is automated management features yet to make self-restart of downed VMs a reality. I believe something along those lines is in the cards with ~2.0 release for ProxVE but I'm not positive.

Of course, application-level 'tricks' can be done to ensure service continuity; but this typically requires either a custom app layer, or other infrastructure to provide redundancy (for example, some web sites use 'front end load balancers' which redirect queries to back-end web-servers in a round-robin manner; thus if a single back-end web server goes down, this is no impact to service). But this isn't really a 'cloud' so much as an 'app server cluster with redundancy and a front-end load-balancer'...

Some web service providers do assert they offer 'cloud' based web hosting; and I'm sure this is possible - assuming a fairly constrained service hosting offering (ie, static web content, or well contained web content ..) - but again this becomes more or less "application level redundancy" coupled with 'custom OS/app redundancy and data replication' as a means to providing more assurance of uptime in the event of a hardware fail.

Presently, I don't think there is any free, fully automatic way of offering a 'service cloud' which is fully fault tolerant (ie, restarts VMs on its own). You could develop something yourself of course but that is probably more effort than it is worth.

Probably in the short term, either ProxVE or XenServer free edition are closest to what you want: a no-cost virtualization platform; which can be run on relatively few pieces of hardware; and supporting live-migration of VMs for no added cost. ProxVE can do this without a "SAN / shared storage device" thanks to availability of DRBD features in ProxVE. XenServer would need either NFS or iSCSI (or SAN, but likely is out of your budget scope?) storage target to provide live migration functionality.

Right now - there is no 'easy, magic, free" way of having complete / instant failover of "any/all services" ; without either (a) designing your specific app service platform with this goal in mind; (b) paying money for a commercial VM implementation, or (c) something else I haven't thought of here which is relevant

I hope this is somewhat clear / makes sense. For possible use to others in the forum I'm also posting this text (not your email though) there .. hope that is OK,


Tim

-=-=-=-=-=-=-=-=-=-=-=-=-=-
Fortech I.T. Solutions
http://FortechITSolutions.ca
-=-=-=-=-=-=-=-=-=-=-=-=-=-

 

[dmichaels]

Can you tell what exactly the difference between cloudhosting and a virtual private server is? I always thought it is the same. And what would you recommend for video managing. What do you think about this web host?

 

[fortechitsolutions]

Hi,

In my experience, "Virtual Private Server" (VPS) is a term typically used to denote an "OpenVZ / Container based" VM; although strictly speaking this is not written in stone. The term has been around for longer than the now-in-vogue "Cloud Computing" terminology. Often / in the past, VPS has been also used in the context of 'value price hosting services' - sometimes, but not always, with lower-spec virtual servers (smaller ram or CPU share allocations) than might normally be present in other hosting solutions.

Re: your question - 'video managing' - I have no idea what you mean exactly - if you want to give more detail that would help me comment.

Re: IX Web host - I had a quick look at their site - they look like a typical service provider with fairly typical pricing structures. Ultimately I have no experience with them, and the best info you might get will come from someone who has been a client with the provider. There seems to be a fair bit of variability in this market - some vendors 'oversubscribe' more than others, and the level of performance you get out of the service will vary accordingly. So 'testimonials' actually are very important. Barring that, a 30-day free trial with no-questions-asked account termination to let you 'kick the wheels' is one way to go, assuming you don't mind burning some time/effort in setting up the service.

Hope this helps a bit,

Tim Chipman
Fortech I.T. Solutions
http://FortechITSolutions.ca

 

[BiagioParuolo]

I think to have 2 or more Datacenter with cluster nodes, but for disaster recovery, the best solution is transfer VM backup/snapshot files among DC, because Live Migration need a lot of bandwith.

 

[iaMsEo2011]

We can say that a business could still stand without cloud computing, but not as long as of those with it and those who consider it important.

 

[William Smith]

Proxmox virtual environment ia an easy to use Open source virtualization platform for running virtual Appliances and Virtual Machines. You want to form a cluster of servers. A server cluster is a group of servers working together as a single system to provide high availability of services for clients. When a failure occurs on one computer in the cluster, resources are redirected and the wrokload is redistributed to another computer in the cluster. You can use server clusters to ensure that users have constant access to important server-based resources. Cloud computing is much more than this, and thus, can offer better performance. Cloud Computing is a new way of providing and consuming IT services.

Cloud computing is a model for enabling convenient, on-demand network access to a shared pool of configurable computing resources(e.g., networks, servers, storage, applications,and services) that can be rapidly provisioned and released with minimal management effort or service provider interaction. Cloud model promotes availability and is composed of five essential characteristics, viz., on-demand self-service,broad network access, resource pooling, rapid elasticity and measured service. Cloud consists of four deployment models private cloud,Community cloud, Public Cloud, hybrid cloud. Key enabling technologies include fast wide-area networks; powerful, inexpensive server computers, and high performance virtualization of commodity hardware.