A lot of people talk about going to cloud for cost reductions but there are some other very important underlying benefits to virtualizing your dedicated as well - even if you are only running one vm on it:

hypervisor with upgraded feature sets for management

ability to move your vm from server to server with no downtime or interruption for hardware upgrades or planned maintenance

high availability features including automatic reboot on another machine in under 2 minutes if you experience a physical machine failure. this ia a great lifestyle enhancement for tech support people - hardware outages go from hair on fire emergencies to standard break fix operations that are not service impacting - what a difference!

fault tolerance - some virtualiztion setups allow you to run full fault tolerance where the vm instance is running on 2 different machines at the same time so you dont even get a reboot if you have a physical hardware failure

ability to spin up more machines in a pinch in minutes and move them to their own hardware node when you get it up

ability to have all your images that you want to run - ie your iso's / distros with all the associated software packages in prepacked application bundles ready to drag and drop and spin up in minutes vs doing the old manual installs of multiple packages.

ability to move from san to san if you are running on a san with no interruption in your operation

increased flexibility in networking and movement of ips

Hardware upgrades - These are not exactly virtuzaliation benefits but are certainly best practices on hardware that you generally will find when you roll to a virtualized system at a provider:

SAN - getting on an enterprise san for your storage instead of internal drives - this is a huge upgrade in stability and survivability of your data. Most virtualized offerings providers have rolled out are SAN backed - make sure to ask if they are on an enterprise level san or a single node san like a simple nfs setup. A distributed highly available san that can survive multiple failures in different areas is the way to go these days. You will find the performance to be as good or better than the best internal raid setup you can build and the uptime will be better since you dont have to rebuild raid arrays that affect your entire data set when you have hardware failures on a large san. I compare failures on the systems to throwing a brick in a puddle or throwing a pebble in the ocean when looking at small internal raid setups vs being part of a large san.

virtuzlized networking - you will find that most of the virtualized supported services are on the latest network gear that supports full virtulization and therefore takes advantage of the latest and greatest ip swithcing and management. say goodbye to spanning tree and other problematic issues. Also you are likly to find that the servers are connected at multiple gig connections that are in a trunked format for support of high availability since its so easy in a virtualized environment.

Ability to run more on a single hardware platform - you will find that todays hardware will support more than a single instance can toss at it generally speaking - so when you can keep tossing vms at it you are able to effectivly utilize those resources which is good for all of us - that means less power used overall and less hardware manufactured. This is a very green thing to do!

Everything is moving this route in technology so don't get left behind!

What other benefits can you guys think of and if you are not doing this yet - what are some of your reasons for not getting updated on technology?

Great post.

:)
D

Informative, Nice work!! Very informative for non-techie people to know the advantages.

I also think the aspect of geography is quite important to get onto.. Moving a VM over 1000's of miles is no problem anymore.. It can all be done in a matter of minutes. No need to have costly moves of gear anymore.

And the Green aspect of having virtual servers.. Much more bang for the buck when bulking together massive amounts of servers on big boxes (I personally REALLY love the IBM 3850 models).

First of all, virtualization is great! But as soon as one starts mentioning the cloud and SAN or specifically "infrastructure sharing", all goes south. Keep in mind that historically, Web hosting providers have not always acted in their clients' best interest (overselling, service level, etc). There has to be a cultural adjustement before trust is reinstated. It's only then that the idea of the cloud and resource sharing can take off.

Regards
Joe / UNIXY

First of all, virtualization is great! But as soon as one starts mentioning the cloud and SAN or specifically "infrastructure sharing", all goes south. Keep in mind that historically, Web hosting providers have not always acted in their clients' best interest (overselling, service level, etc). There has to be a cultural adjustement before trust is reinstated. It's only then that the idea of the cloud and resource sharing can take off.

Regards
Joe / UNIXY

great point. that is always going to be an issue - no matter what you are talking about and has to do with what kind of provider you select. Do you gravitate towards the cheapest bottom of the barrel too good to be true deals? If you do then you will probably do this on cloud as well and likely eventually get burned.

There are some high quality providers out there that are building quality enterprise level infrastructure that is the real deal. There are also a lot of junk under the covers buildouts going on as well.

Ask the provider to give you a diagram of what they are doing with the list of the specific gear they use and how many vms they are putting on servers. Ask them how much they are overselling and what their sla and provisioning policies are. The good ones will be proud to tell you because they are investing. The junk ones will be evasive or just wont tell you. run dont walk away from them.

I personally think its the way to go and adds a huge amount of operational excellence to hosters that could never afford this level of enterprise build.

Very good info :agree:

What other benefits can you guys think of

Virtualization is a "green" technology. Some providers have gone as far as to advertise planting trees along with your purchase, but the real ecology improvement is in energy and infrastructure. As virtualization is becoming more and more widespread, there is talk about a shrinking demand for new data centers. Fewer machines means fewer resources to build machines, and less energy needed to run them.

Not only is virtualization good for business, it is also good for the environment.

I agree with the points about SAN storage. We use EqualLogic SAN units and they are leagues ahead of a local disk RAID setup. I'm not sure many customers take this into account though as it's still fairly specialist. Also, if the provider is running the full feature set of VMware then scheduled maintenance is a thing of the past as you can easily migrate servers between physical machines if you need to do any server maintenance.

It must be just me but i have never liked virtual servers. Shared hardware resources and a single point of failure is not something i would like to sell to my customers,... I'll admit it is cheap and eco friendly but most of my customers are after good service rather than cheap.

It must be just me but i have never liked virtual servers. Shared hardware resources and a single point of failure is not something i would like to sell to my customers,... I'll admit it is cheap and eco friendly but most of my customers are after good service rather than cheap.

If you dont like single point of failure then you should be off of dedicated servers. Virtual private data center setups on something like vmware take that equation out of the picture.

You can run 2 servers and have them back each other up.

When i say "single point of failure", i'm talking about having 10+ virtual servers go down because one physical server fails. Even if you double it up with failover, you still have a higher potential risk. You're also more open for security problems as one hack (for someone that knows what they are working wth) could potentially get the lot. You also have the problem of one customer running heavy scripts/database searches etc and chewing cpu/memory.

Overall, I'm guessing that our "target clients" are different. I'm working with customers that want good service and you're working with people that want cheap. I've had far too many pepole move over because their "server" ran out of (shared) disk space and their services crash because it ran out of (shared) memory.

When i say "single point of failure", i'm talking about having 10+ virtual servers go down because one physical server fails. Even if you double it up with failover, you still have a higher potential risk. You're also more open for security problems as one hack (for someone that knows what they are working wth) could potentially get the lot. You also have the problem of one customer running heavy scripts/database searches etc and chewing cpu/memory.

Overall, I'm guessing that our "target clients" are different. I'm working with customers that want good service and you're working with people that want cheap. I've had far too many pepole move over because their "server" ran out of (shared) disk space and their services crash because it ran out of (shared) memory.

I believe you have some fundamental misunderstandings of enterprise cloud / virtualization technology - at the risk of angering you - let me take a moment to try to highlight some features for you specifically in reference to your statements above.

Disk: you wont run out of disk when on an enterprise level san because it will be able to scale out. You also will have predetermined LUN sizes but they will be able to grow with a mouse click. You can also reserve disk for a vm and limit disk for a vm so you dont run out. You can oversell easier through thin provisioning as well if you so choose. Enterprise san is like having many many internal raid arrays in a big array. faster and more resilient.

CPU - you can limit the amount of cpu a customer gets down to the mhz and you can give them a defined ability to burst - software like vsphere will manage the resource for you and will get you the minimum and allow you to use up to what you allow that is available in burst.

memory - you can carve out the min and max they can get

security - you secure the individual vms - no one can hack and get all the vms at once. You also run the box from vcenter which is inherently more secure than at the os level. you will not see entire box compromises like you do with lower level tinkering around vm software.

High availability - you can run ha in resource pools which means those 10 machines you reference can be part of a ha "group" (resource pool, and if one of the boxes goes down, the vms on it migrate automatically to another box - they can go to one box or they can go to the best boxes available based on capacity through DRS.

Scheduled maintenance downtime - its a thing of the past since you can use vmotion to move the running vms to another box without even seeing a reboot.

user running something heavy that will cause problems for others - not in vmware - you can limit and scope the resources they can access.

Bottom line this is a good illustration between enterprise level offerings based on vmware etc compared to amature hour based on things like openvz.

I hope this helps - just post if you have more questions.

Very good post sailor.

I like to call it, "Disposable infrastructure" Fact that if you need a test enviroment, or an enviroment for a few days. You can easily create an environment and roll it out painlessly and simply.

That's a good instruction indeed.

It sounds like you're taling specifically about clouds here, I thought we were talking about virtualisation in general. When I made my post I was referring to virtual servers only, not clouds (not that I like clouds much eaither, I find them hard to make a profit on).

We moved from a single cpu (dual socket) system, and various opterons and such onto a dual socket xeon same as the other single but each core like .66ghz faster.

handles our entire load flawless on Xen, 1 VM isn't paravirturalized (Debian squeeze) runs low load too. 2 VM's are WHM/CPanel w/litespeed pushing sites that are on alexa (traffic stats) :|

While I agree that there are certainly advantages to this sort of architecture, this isn't necessarily an architecture for every season (none is). In my experience, most architects of large sites prefer shared nothing to shared storage architectures when possible. And anyone who says that enterprise SANs don't fail simply hasn't been around long enough. I've personally seen several fail, even those with multi-site replication (and being maintained by the vendor's support team no less), and there have been high profile outages or data loss attributed to enterprise SAN issues in everything from ATMs in the UK being down to T-Mobile customers losing their contacts database. These aren't homegrown jobs either, but high end arrays from EMC, HDS, etc.

Enterprise SANs may not fail often, but when they do, it's often spectacularly, as a lot of enterprises have so much of their critical data on them. I remember reading in a DR journal a few years ago that SAN outages where the third most common cause of application outages, so this is an industry wide issue. There's no doubt human error is the cause of many of these issues, but when you've consolidated all your data on a SAN, you're greatly magnifying the impact when something does go wrong.

While I agree that there are certainly advantages to this sort of architecture, this isn't necessarily an architecture for every season (none is). In my experience, most architects of large sites prefer shared nothing to shared storage architectures when possible. And anyone who says that enterprise SANs don't fail simply hasn't been around long enough. I've personally seen several fail, even those with multi-site replication (and being maintained by the vendor's support team no less), and there have been high profile outages or data loss attributed to enterprise SAN issues in everything from ATMs in the UK being down to T-Mobile customers losing their contacts database. These aren't homegrown jobs either, but high end arrays from EMC, HDS, etc.

Enterprise SANs may not fail often, but when they do, it's often spectacularly, as a lot of enterprises have so much of their critical data on them. I remember reading in a DR journal a few years ago that SAN outages where the third most common cause of application outages, so this is an industry wide issue. There's no doubt human error is the cause of many of these issues, but when you've consolidated all your data on a SAN, you're greatly magnifying the impact when something does go wrong.

Everything can have problems from time to time. An enterprise san uptime is far superior to internal storage at the end of the day and the risk for data loss is much lower as well.

Getting on an enterprise san is a no brainer.

Having your own private san is certainly the cats meow but most can not afford this so being on somones managed san is a great alternative and a huge stepup.