Cloud Linux is not Cloud Hosting

Portal Home > Knowledgebase > Articles Database > Cloud Linux is not Cloud Hosting

Posted by NexDog, 01-26-2011, 09:04 AM
Seeing quite a few hosts that use CloudLinux state on their sites that they offer Cloud Hosting. CloudLinux is just an OS, whereas Cloud Hosting depends on architecture. I could make a brand new Linux distro and call it WindowsLinux, doesn't mean it is Windows, heh, Discuss.
Posted by layer0, 01-26-2011, 09:57 AM
I haven't noticed hosts doing that. Are you sure they are not actually using a 'cloud' environment in addition to CloudLinux (I know some hosts that are doing this)? But anyway, even if they are simply using CloudLinux, the simple answer is that this is all for marketing purposes. I consider 'cloud' to be a marketing term at this point in any case.
Posted by warlock-m, 01-26-2011, 10:34 AM
Speedysparrow was using VPS.net, so they were offering true cloud hosting solutions according to the vps.net architecture.
Posted by JasonD10, 01-26-2011, 10:48 AM
Sometime last year this was also discussed. CloudLinux is now branded as the "Cloud Ready OS". Cloud can be SaaS, PaaS, and IaaS. The problem comes in when most people here refer to IaaS, and rightfully so. Web Hosting providers are normally IaaS, sometimes PaaS. Companies that you have mentioned above appear to be claiming Cloud when in fact they are lacking the underlying inrfastructure to truly be, Cloud. Even if they are selling PaaS and not IaaS, the infrastructure of their business must be Cloud as a standard shared web hosting platform, vps platform, etc are not Cloud if they lack the required components of Cloud.
Posted by JasonD10, 01-26-2011, 10:52 AM
You are welcome to your opinion, but Cloud is not a marketing term. It is much more defined and there are standards being set. Besides companies like Tier 1 Research and the 451 Group, the US Government, and even providers are continuing to properly define and establish the definition. Marketing terms are things like "compute cycles" and "overselling".
Posted by MostHost, 01-26-2011, 10:57 AM
Excellent point. The importance of standards for something as dynamic as 'cloud computing' and 'cloud hosting' can't be overstated in order for consumer to really adopt the technology. The clearer everything is explained to end users, the better for everyone involved.
Posted by iseletsk, 01-26-2011, 12:48 PM
As defined in wikipedia: http://en.wikipedia.org/wiki/Cloud_computing "Cloud computing is location-independent computing, whereby shared servers provide resources, software, and data to computers and other devices on demand, as with the electricity grid." I am sorry, but I cannot see a place where "cloud" is nwell defined term in any way or form. While elasticity, scalability, reliability are often considered to be "cloud" related qualities -- even that is not easily defined. For example: Try to scale single instance on AWS -- you have to create new instance, and even that is not scalable beyond one real server -- so you have to get more instances. Reliability -- same story here, your instance on AWS can go down at any moment. Most likely it will be down more often then dedicated server. Elasticity -- well, what does that term really mean? So, I kind of disagree with bashing people for something that doesn't agree with "your" definition of the cloud. And just for the record: I don't think that putting CloudLinux on a single shared hosting server makes that shared hosting a cloud. In my view, for it to be considered a cloud. Most likely -- I am incorrect, but that is my wishful thinking of how cloud hosting should be defined.
Posted by arisythila, 01-26-2011, 01:20 PM
I also agree, There is what I believe to be a certian standard, sort of like PCI Compliance Standards, or SAS 70, or some of the others. To me there are a set of rules you must abid by to be a "cloud" hoster. 1. Centralized Storage 2. N+1 In the event a switch fails, do you have a hot standby? 3. Automatic failover (WITHOUT HUMAN INTERVENTION) 4. Scalability 5. I think being able to do load balancing and stuff is pretty important, but I dont think this is a limiting factor. If you can obtain these sort of standards, Im sure there are a few things i've missed. But these are the basis. Marketing... LOL. I guess it depends on what kind of cloud your thinking of..
Posted by JasonD10, 01-26-2011, 01:24 PM
The wikipedia article also states that device and location independence is a key feature of Cloud. Cloud Linux is not device or location independent, it's just an OS that is not utilizing multiple servers or locations. This is a part of the requirement that goes into further defining reliability as there needs to be a level of redundancy in the environment.
Posted by iseletsk, 01-26-2011, 01:25 PM
Michael, Who came up with this "standard"? Which governing body approved it? Where I can read about it?
Posted by iseletsk, 01-26-2011, 01:28 PM
As I stated -- I don't consider server with CloudLinux to be a Cloud. Yet, for shared hosting customer, device and location independence is there. They really don't need to know if they are hosted on particular server (you can always move them to another shared hosting server) nor location (do most customer's care in which DC they are?). Reliability: Would multiple power supplies satisfy reliability? What about multiple network connections? RAID disks? Try AWS instance -- see how reliable it is. Yet, AWS is considered to be a cloud. I am sorry, but Cloud is used as a marketing term. It was created as a marketing term -- and there is very little in terms of technical definition of what "cloud" should really be. And just in case -- no, CloudLinux is not a Cloud. I agree with that
Posted by JasonD10, 01-26-2011, 01:39 PM
NIST released a definition of Cloud in October of 2009 which included the following statement: Note 1: Cloud computing is still an evolving paradigm. Its definitions, use cases, underlying technologies, issues, risks, and benefits will be refined in a spirited debate by the public and private sectors. These definitions, attributes, and characteristics will evolve and change over time. Rackspace, Amazon, and other large Cloud infrastructure's were built more or less as you stated and they are NOT per instance redundant across multiple servers, and/or locations. They are essentially a VPS provider. When the host server that goes down when you are on that server, so does the instance. Rackspace still to this day clearly states on their (new) website that servers are RAID10, and in the event of a host node failure the customers will go down in which they will try to resolve it in 3 hours or less. Their instances do not utilize centralized storage, and they are not redundant across multiple physical servers. The debate has been last year that the level of redundancy required is complete server independence. This has been discussed and is becoming the standard. Customers are now expecting this level of service and rightfully so. Every major Cloud platform vendor has this capability now, and they are all instructing their new partners and customers to utilize it. The standards will continue to evolve and single points of failure of an entire server, switch, or other traditional point of failure crashing will be eliminated in Cloud.
Posted by iseletsk, 01-26-2011, 01:53 PM
Hmm, based on this definition from NIST: http://csrc.nist.gov/groups/SNS/clou...ud-def-v15.doc Shared hosting with CloudLinux is Cloud (I still don't agree with it, but I am going against well defined standard). Lets see the "Essential Characteristics": Well, that has been there since end of 90th. You can get shared hosting account via complete self-service process. I guess no explanation needed -- hosting had it always Multi-tenant model - check, resources: processing - check, memory -- check, network bandwidth -- check, virtual machines -- not with a single server, but it doesn't state that it has to include all resources This can be done with LVE, you can scale a single customer by increasing limits, on the fly. And if you are on 16 core server -- the ability to scale is pretty huge. You are still bound by single server, but scalability doesn't have to be infinite. Storage, bandwidth -- had been done for a long time. With CL -- you can also measure CPU (processing), connections and memory (in latest beta version). Lets see service models: That works for shared hosting. Create PHP app, deploy it. Does it mean that CL on a dedicated server is a cloud? Fine, I think I would have to cave in, and accept that it does. It goes against my belief of what cloud is, but based on this definition, shared hosting with CL & dedicated server is a cloud. TBH: Based on this definition, pretty much any shared hosting is a cloud.
Posted by arisythila, 01-26-2011, 02:02 PM
Ahhhhhhh, I just now read this, and i know why this is a battle now... "CloudLinux -- The OS that can make your Shared Hosting stable" Igor being the CEO of this company, and we are setting higher standards that he can achieve with this "OS that makes shared hosting stable" Its all coming Evident.. :-)
Posted by iseletsk, 01-26-2011, 02:09 PM
Michael -- sorry I didn't realize you haven't seen my signature at the end of each message. I am sorry if I confused you or bored you. I thought we had meaningful conversation on what is cloud, and if CL can be considered a cloud when it runs on a single server (note, I don't think it can). Yet, if you think it is waste of your time, I apologize, and stop posting.
Posted by arisythila, 01-26-2011, 02:33 PM
Yes, but instead of argueing. You should take into heart what cloud hosters are currently doing. Also what kind of standards we have put into place. Thanks, EDIT: This way you can go back to the drawing board and make your product meet and exceed those standards. Last edited by arisythila; 01-26-2011 at 02:36 PM.
Posted by iseletsk, 01-26-2011, 03:14 PM
So, what is your idea of a cloud hosting?
Posted by JasonD10, 01-26-2011, 03:18 PM
We're not all going and spending large sums of money to create a higher level of reliability because it's fun, it's because people are expecting this of Cloud. Redundant power supplies, RAID, and all of that is fine but servers still fail and they fail hard. Centralizing the storage, releasing an instance/vm from a single server, and other environmental flexibility is doing great great things for this business. I say embrace it with open arms as continuing to deny these benefits will only leave you behind.
Posted by arisythila, 01-26-2011, 03:19 PM
When i think if cloud hosting, I think scalable, and high availability. If srv1 crashes, its got to provision and run on srv2. Im not saying a dedicated cannot be reliable, but just I think our main thing is having as few single points of failures as we can. Thats our main goal.
Posted by iseletsk, 01-26-2011, 03:35 PM
1. As of today, I see cloud often being less reliable then dedicated servers. I know it goes against what cloud is/its architecture and everything else, but due to complexity, human error, SAN issues, network issues, and 100s of other reasons -- I see that majority of cloud services are having issues. Often those issues spawn the whole cloud, and not just a single server. And this happens with major companies who know what they are doing. Rackspace, Amazon, name any cloud company -- I am sure they had major downtime. I am also a strong believer that as platform matures -- this will change, and cloud will become more reliable then dedicated servers. Last edited by iseletsk; 01-26-2011 at 03:40 PM.
Posted by arisythila, 01-26-2011, 03:45 PM
Last year we were able to achieve 100% uptime. Most of our multi tiered customers (clustered customers) that are running multiple in gateways, load balancers, and application servers noticed ZERO downtime last year. It was kind of funny, I had a server crash awhile back System Uptime : 373 days, 16 hours, 5 minutes Controller Uptime : 42 days, 7 hours, 21 minutes Last Server Failure Time: Sat Aug 28 08:43:22 2010 (srv2) So in this Grid, we had srv2 fail. Applogic, saw that it failed and it would not respond. So applogic logged into the IPMI powered down the srv, powered it back up, rebuilt the volumes, and it was back in good running state again... Mind you, We got ZERO calls about this server crashing. I didn't even know it happened until septermber when i did my rounds..
Posted by JasonD10, 01-26-2011, 04:10 PM
You have heard about major Cloud problems because it effects many customers, but you don't hear about the tens of thousands of servers crashing every day that are not in a Cloud because it only effects one server, or one application. There have been far fewer of these outages on superior technology with complete redundancies where even the failure of the Cloud engine results in no downtime of any running Cloud components. Fact. We have been in business for 13 years primarily operating as a Managed Hosting provider. We host many traditional infrastructures, still to this day we host more off the cloud than on the cloud. Fact. Of that business, these customers spend significantly more to create a high available environment than they would in Cloud. Fact. Those customers routinely have large scale and costly outages that require manual intervention of physical engineers and emergency engineers in the middle of the night to correct conditions. Fact. Our Cloud's have never gone down unless we wanted it to which is 1-2x/year for our major upgrades. During those time, the most stringent customers have multiple clouds so they remain 100% available as only one is done at a time. Fact. As Michael has stated, it is automatic when failures do occur. We have a failure two days ago in a Cloud. The Cloud sensed the failure, restarted all instances on another available server, started rebuilding the data streams, logged into the failed server via IPMI and rebooted it, and it came back up and cleared the data streams and started over as a fresh new server in the Cloud. Single points are failure are eliminated if the provider chooses to do so. There's no reason to see full Cloud outages if every component, including the engine, is redundant.
Posted by ramnet, 01-26-2011, 04:20 PM
Cloud is a marketing term plain and simple. That's a Cluster. Not a cloud really unless you define them to be the same. Cloud isn't anything new - they have been around since the 1950's as Clusters - some marketing people invented the Cloud word for the ignorant masses who find words like cluster and processor and such scary. Same thing happened with hacker vs cracker. A cloud supposedly means a high availability cluster - so why not call it that? marketing.....
Posted by iseletsk, 01-26-2011, 04:29 PM
Care to quote source of your statistics? [quote] I can see that. They are not prepare to deal with complexities of running a cloud. Hosting provider is much better to deal with it. What about individual clients on your cloud? Do you want to say that no individual customer on your client didn't have downtime due to disk, network, or what ever other reason? If that is the case -- GREAT!
Posted by iseletsk, 01-26-2011, 04:30 PM
Congrats! Excellent statistics? Is that shared hosting server running in VM on applogic?
Posted by arisythila, 01-26-2011, 04:40 PM
I can see some people using cloud as a marketing term. Then I can see others using it how it should be used. Welcome to your opinion.
Posted by arisythila, 01-26-2011, 04:41 PM
That is one of our Clouds in Seattle. I've moved people from 3Tera's OLD OLD grids that had over 900 DAYS of uptime. I don't really call that a marketing term. Something that we were never able to obtain without complex clusters on dedicated servers
Posted by iseletsk, 01-26-2011, 04:44 PM
But is it really shared hosting server? Or is it just a VM running single site for someone (or something similar).
Posted by arisythila, 01-26-2011, 04:44 PM
[QUOTE=iseletsk;7245685]Care to quote source of your statistics? I think hes just quoting what he sees in his business with Cloudweb. I can see similar things in our company. We've effectively converted a lot of dedicated server customer from dedicated infrastructure to cloud infrastructure. By moving our dedicated customer base from dedicated servers to Applogic, we were effectively able to cut our monthly tickets down from 100-200 tickets a month to about 10-20 tickets per month. Customers are generally happier than when they were on dedicated servers. Thanks,
Posted by iseletsk, 01-26-2011, 04:45 PM
Weren't we talking about cloud as applicable to shared hosting and not dedicated? Or are we comparing apples to oranges?
Posted by arisythila, 01-26-2011, 04:54 PM
Its the infrastructure itself. System Uptime : 373 days, 17 hours, 11 minutes Controller Uptime : 42 days, 8 hours, 27 minutes Last Server Failure Time: Sat Aug 28 08:43:22 2010 (srv2) HA State : ok Total Servers : 12 Servers Running : 12 Servers Enabled : 12 Applications Running : 49 Total CPUs : 192.00 Free CPUs : 70.51 CPU Load : 5.54 Allocated CPUs : 121.49 Total Memory : 288.00 GB Reserved Memory : 4.11 GB Free Memory : 83.39 GB Allocated Memory : 200.50 GB Service Memory : 482.00 MB Total Bandwidth : 24.00 Gbps Free Bandwidth : 5.56 Gbps Allocated Bandwidth : 18.44 Gbps Total Disks : 4 Total Storage : 35.89 TB Free Storage : 13.48 TB So, We were able to effectively maintain at lease N+1 on our infrastructure. Allowing all 49 running applications to continue running. without downtime.
Posted by JasonD10, 01-26-2011, 04:56 PM
If I took the time to research it, the numbers would be much higher but I was being conservative speaking only to the US markets. Exactly, and as a host it's easier and better to do it in a Cloud. It depends on the application but many customers will experience no downtime in any of those events. If the client only has one instance (one OS/kernel) and the host node they're running on crashes, they will go down for approximately 3-5 minutes as their instance is restarted on another available server. This is the same for all technologies. There is no memory sharing technology to spread a standard Linux or Windows server across multiple servers, so we're limited by today's technology.
Posted by Sparrow-Sean, 01-26-2011, 05:10 PM
So you are slandering us because we are using CloudLinux?! We have never stated to our customers that we are a 'Cloud' provider because we use CloudLinux? You have no idea how we virtualize our cloud or our environment but you are willing to stake that you know how we operate our business and are willing to make a statement on us not being a Cloud Provider I think you really need to learn how we offer our service, and how we actually give our services basing them under a cloud environment. I believe this comes down to bashing a company without forehand experience or knowledge of how we have our business setup.
Posted by arisythila, 01-26-2011, 05:12 PM
Tell the community what your using.....
Posted by Spiral-Eric, 01-26-2011, 05:13 PM
Just for your information, we were powered by VPS.net (Cloud) and are now powered by the GNAX Enterprise Cloud in Atlanta. I hope this will put all speculations to an end.
Posted by arisythila, 01-26-2011, 05:17 PM
What cloud infrastructure are they using? What hardware are they using? What does their hardware look like? (how is it setup?)
Posted by Tyl3r, 01-26-2011, 05:20 PM
Why are you being rude about this? Obviously you're not very well informed if you don't know what hardware/setup they are using. OnApp, read about it.
Posted by JasonD10, 01-26-2011, 05:21 PM
OnApp? Their website says vCloud?
Posted by arisythila, 01-26-2011, 05:21 PM
I'm sorry, I am just very direct about when i ask questions. I'm not even the one that called him out. I don't mind what he uses. OnApp is a good solution.
Posted by Tyl3r, 01-26-2011, 05:24 PM
I'm talking about VPS.net
Posted by Tyl3r, 01-26-2011, 05:24 PM
I would suggest using Google then and researching it, instead of trying to be a bully on an internet forum.
Posted by ramnet, 01-26-2011, 05:25 PM
To quote http://www.gnax.com/ GNAX Cloud * Powered by VMware vCloud Director
Posted by JasonD10, 01-26-2011, 05:26 PM
Your comment was out of turn that's where the confusion came in .. They already responded that they're not using VPS.net any more, but in fact GNAX who states to be using vCloud (although it looks to be beta as well).
Posted by JSCL, 01-26-2011, 05:27 PM
No no. We offer OnApp for people and organizations wishing to start up their own private (or public) Cloud. For our NETDEPOT Cloud product (which we offer through NETDEPOT.com) is the product that is being used by SpeedySparrow and is based on VMWare vCenter with VMWare's vSphere Cloud OS. Here's a quote from our site which sums it up nicely for you: If you don't believe it's Cloud - take a look over our website and do some reading for yourself http://www.netdepot.com/cloud-hosting.html Thanks. Last edited by JSCL; 01-26-2011 at 05:31 PM.
Posted by arisythila, 01-26-2011, 05:27 PM
Your the one that came in here. If your complaining because someone called you out, Tell people why the person that called you out is wrong, and back it with facts. Thats all. Thanks,
Posted by JSCL, 01-26-2011, 05:31 PM
Our GNAX Cloud product is in BETA but our NETDEPOT Cloud offering is definitely in full swing
Posted by JasonD10, 01-26-2011, 05:39 PM
Thanks for the clarification
Posted by Sparrow-Sean, 01-26-2011, 05:49 PM
Actually CloudWeb we still do use them in conjunction with NetDepot, so we are actully onside to both providers, Nerdie is 100% correct, what we do not like is others trying to slander our business for not being a true cloud even though they have not got the real facts. Shows there are some will speculate quite heavily without prior hand knowledge.
Posted by arisythila, 01-26-2011, 05:55 PM
Again, I'm not the one that called you guys out. I think it was pretty unprofessional. But since you wrote. I wanted to know what you guys were running on. Good to know what other hosting providers are running on. I've heard a lot of good things about SpeedySparrow. Thanks,
Posted by Spiral-Eric, 01-26-2011, 06:01 PM
It actually looks more like competitor bashing... you yourself provide 'cloud hosting' - should be rant and question what you are on?
Posted by JasonD10, 01-26-2011, 06:02 PM
Well, I tried speaking with you in August to ask about this which my inquiry was ignored as I had a customer coming from you specifically asking about your infrastructure in comparing us. I personally reviewed your website and only found mention of Cloud Linux, and nothing about your underlying infrastructure so I was left to my assumptions since you did not respond to my inquiry. How really can I be faulted for such a thing when I tried to clarify this to one of your customers when this was ignored. Thankfully, I have never said a word of it until now which it seems may have been in haste. What I find odd is that you will gladly speak up and explain this here, but make no mention of your actual Cloud infrastructure on your website. It is a red flag but I understand now with you moving hosts and not owning your infrastructure that may be why it was hidden. In any case, I do apologize and thank you for setting the record straight.
Posted by arisythila, 01-26-2011, 06:12 PM
Sure, Your welcome to ask me anything you'd like. We are offering co-location services, and Cloud hosting currently. I openly share pretty much any aspect of our business. I've very proud of the infrastructure that we've built up over the years. Also wanted to mention that i didn't bring you into this thread.
Posted by NexDog, 01-26-2011, 10:01 PM
It wasn't my intention to "call someone out" and I think we should be able to discuss each other's services without getting all defensive. This thread is actually very productive for SpeedySparrow as you get to discuss your service which gets you traffic and sales. The only bad publicity is no publicity as the old marketing maxim goes. I'm not knocking CloudLinux either. I think CloudLinux is absolutely awesome and I'm thinking about using it on a new brand I'm working on. It doesn't qualify as real Cloud Hosting in my view but SpeedySparrow have confirmed they are on a real Cloud, so all good. If it was me I'd probably pimp that fact a bit more on the site. I think CloudLinux (OS) and LiteSpeed (HTTPD) powered Cloud Hosting (infrastructure) is the future of web hosting and SpeedySparrow is already there so hats off to them.
Posted by arisythila, 01-26-2011, 10:53 PM
Agreed! I don't mind if someone questions our equipment. Lets do it!
Posted by KarlZimmer, 01-27-2011, 11:15 PM
I agree, if this is well defined, where is the definition that 50%+ of people would agree on? There are certainly "clear definitions" but there are thousands of them and they don't agree with each other. There is currently no standard, and as far as the term is being used now it is mostly marketing. yes, there are some things now that I'd consider a cloud, but that doesn't mean it matches everyone's definition, as there is no clear definition.
Posted by Sparrow-Sean, 01-28-2011, 02:47 AM
Cloudweb, I have sent you a private message, please do read it carefully and ask as you wish. I do want to state however, I am not ignoring you and I do not check my private messages all that often, the message you sent is actually still sitting in my inbox and I was not aware of it so please do not make the assumptions that I was being ignorant and ignoring your message. As mentioned in my private message, there are other alternate options you could of chosen from, but none of these were taken so ... I am sorry, not to sound rude or anything as it is not my intention to do so, though, some people cannot just co-locate their systems, sure companies grow and expand to achieve this at a later date of their expansion(s). I know MDDHosting started off with Softlayer and are now Co-locating themselves but I am not sure how we can be put down when half the businesses on this forum run off either a VPS/Dedicated System. We are not in this business for games, otherwise I would of been out of it a long time ago @NextDog, Thank you! I did not want to sound so arrogant towards you but personally claiming without knowing is never a good thing to do as many others will tell you, have been there before /end./ Last edited by Sparrow-Sean; 01-28-2011 at 02:53 AM.
Posted by ramnet, 01-28-2011, 03:15 AM
Well said Migrating off legacy gear is a royal PITA tho I must admit. .... Obviously CloudLinux isn't Cloud Hosting although it does have some cloud qualities - and that seems to be the key point. Hosts that have hosting with some cloud qualities are calling themselves Cloud Host's all too often when they are really only "Could-Like". Of course, that all goes back to defining what a could is and what is required to make a complete cloud instead of only being a partial cloud. I do however think most would define a cloud as a HA Cluster. Last edited by ramnet; 01-28-2011 at 03:19 AM.
Posted by MikeDVB, 01-28-2011, 03:19 AM
Anybody who really wants to go colo can do it, although it may cost them a bit more (such as interest on financing for example). Some facilities will finance your gear for you or you may be able to get money from a bank, small business organization, your government, or even private investors. It all comes down to selling your business plan and how going colo will allow you to do X, Y, and Z as well as paying back your backer in a timely fashion and how you will make it worth while for them. Now I will agree that going colo can be much more daunting and can very easily seem impossible when in reality it's just more difficult and requires more work/effort than simply leasing a server from a provider.
Posted by cartika-andrew, 01-28-2011, 09:48 PM
The term "Cloud" is certainly a marketing term - but, make no mistake about it, there is legitimacy to the concept, the architecture and the technology that is underlying the marketing. Just because some choose to abuse the term from a marketing perspective, it does not mean that the intention and potential of the "cloud" movement should be disregarded. Too many people have spent far too much time and money trying to make high availability and elasticity - not too mention resource accountability - a reality in this industry. Having it summarily dismissed is not appropriate. As a provider that has been offering "clusters", in all shapes and sizes for over a decade, I can tell you with absolute certainty that a cloud and a cluster are not the same thing. A cloud is a specific infrastructure, a cluster is how services are distributed. In actuality - people can and do cluster in the cloud. The 2 are not nearly the same thing. I apologize, but comparing "cracker vs hacker" to "cloud vs cluster" is not being rational. Cloud computing actually started around 15 years ago. IBM sold this as "on demand". Prior to this, the concept did not exist. It has taken almost 15 years for the concept to reach the masses, and it has been branded as "cloud" vs the original IBM "on demand" because a cloud does not mean a cluster - plain and simple. You could have all services (web, dns, email, db, etc) on a single cloud instance. This is NOT a cluster. This simply means that these services are all hosted on a single instance, on top of cloud infrastructure. Conversely, you could also have a cluster, where some services are hosted on cloud infrastructure and some are not. For example: DNS - cloud node 1 and 2 WEB - cloud node 3 and 4, load balanced DataBase - Dedicated Server 1 and 2, load balanced eMail - VPS 1 and 2, load balanced Above I have described an advanced, load balanced cluster, utilizing a combination of 3 different infrastructure technologies (cloud, VPS and Dedicated). It wold be wrong, absolutely wrong, to describe a cloud as a cluster and a cluster as a cloud. Frankly speaking, anyone who would do so simply does not understand the fundamental differences between the 2. They are not even the same thing - apples vs oranges and all that...
Posted by cartika-andrew, 01-28-2011, 09:58 PM
I should also note that I am a firm believer of the cloudlinux product. Igor has already addressed that he does not consider his OS on a dedicated server to be a "cloud". However, what cloudlinux does do, and does very well, is apply some basic cloud principles to all sorts of infratructure (dedicated, vps, cloud). When used within a cloud environment, it enables granular virtual environments within cloud instances. Absolutely amazing product - almost a cloud within a cloud type of concept. It is also revolutionizing shared hosting and how it is handled. Although I do not agree with calling shared hosting using cloudlinux as a "cloud" on its own - it certainly, philisophically at least - adds some "cloud" type of elements to even the most basic infrastructure - simply by installing a specific OS. People get far too caught up on this cloud marketing scheme - at least in my opinion - but, end of the day, it is a movement towards elasticity, redundancy and resource management. CloudLinux, despite the fact that it alone does not enable "cloud" - it plays a very important role in all of these areas, even in single server setups.
Posted by arisythila, 01-28-2011, 10:18 PM
Would you consider this a cluster, on a cloud? http://img521.imageshack.us/img521/7...10128at616.png
Posted by iseletsk, 01-28-2011, 10:28 PM
The concept is a bit older. Here is an interesting tidbit from Bill Gates schoool days: Of course the concept is even older, and dates back to multics (think 1965). Don't get me wrong. Cloud is different, but each time we try to put technical definition on it, we run into some technology that was defined/made to serve same purpose before. This is why for me cloud is a new name for bunch of old & new technologies.
Posted by cartika-andrew, 01-28-2011, 10:42 PM
that configuration is a cluster.. and I think that is an applogics screen cap you are looking at - so, yeah, I would say this is an example of a cluster within a cloud.. Last edited by cartika-andrew; 01-28-2011 at 10:48 PM.
Posted by cartika-andrew, 01-28-2011, 10:58 PM
Hey Igor - thanks for the historical links I had no idea this concept went even further back then the IBM "ondemand" offering. I think we are agreeing though that the "cloud" movement is a movement towards utility computing, HA, elasticity and resource accountability. This certainly isnt new, but, its application as a service to the masses is absolutely much more modern. Not completely unusual to attach a "buzz word" to this sort of movement - its necessary. It is unfortunate everyone is getting wrapped up in the "marketing" of the concept - because essentially, the underlying architecture, infrastructure and technology are very clear advancement points for the industry and humanity on the whole. There are several elements here and several complimentary products and I think people have too much of a micro view here. Operating Systems, infrastructure, networking - absolutely everything will need to adapt towards this movement. Anyway, I was surprised to read this thread and some of the comments - people clearly are not seeing the big picture here. Your OS is a very important step in this process. I am hopeful, and, I expect - future versions of your OS to help push this process and evolution - as you have already done to date with CloudLinux. Maybe in the future, we will start to see cross computational support, etc, etc... thanks Igor - and keep up the good work Last edited by cartika-andrew; 01-28-2011 at 11:06 PM.
Posted by RossH, 01-28-2011, 11:05 PM
In one post you say there is no standards and then in the next you point to a NIST document, what is NIST? (National Institute of Standards and Technology) In no ones mind except maybe your own can shared hosting be considered cloud computing. Shared hosting can sit on top of Cloud IaaS which can give it some benefits. While you quoted from the document you left some key things out. Cloud Infrastructure as a Service (IaaS). The capability provided to the consumer is to provision processing, storage, networks, and other fundamental computing resources where the consumer is able to deploy and run arbitrary software, which can include operating systems and applications. The consumer does not manage or control the underlying cloud infrastructure but has control over operating systems, storage, deployed applications, and possibly limited control of select networking components (e.g., host firewalls). What NIST tries to convey is basically cloud is a set of resources (servers, storage, network, etc.) that allows multi-tenancy, elasticity, measured service, pooled resources and self service. If you are arguing shared is SaaS or PaaS I'd say you are trying to convolute the definitions of cloud for marketing purposes. Last edited by RossH; 01-28-2011 at 11:12 PM.
Posted by cartika-andrew, 01-28-2011, 11:12 PM
Hi Ross, with all due respect - His new OS is certainly not a defining element of the "cloud". It is however an important piece or component of the cloud. He has developed an OS, that when layered on Cloud infrastructure, takes the entire concept to a new level. Hopefully he continues along this path, because he is onto something significant here. Having said this, I would guess that he would be the first to admit that his OS does not enable, on its own, a cloud. But, do not dismiss what he is developing here. Right now, the cloud is limited by the systems running on it - and primarily by operating systems, web servers, DB platforms, etc, etc.. If Igor keeps developing his OS the way he is, it will play a fundamental role in the evolution of this industry... Last edited by cartika-andrew; 01-28-2011 at 11:22 PM.
Posted by JasonD10, 01-28-2011, 11:16 PM
Excellent reply, you beat me to it. Igor definitely has done a great service, and is developing some very interesting technology that is a great compliment to Cloud. I think we all agree on what Cloud Linux is, it's intention, and the limitations at this point so hopefully this thread can be put to rest soon. Last edited by JasonD10; 01-28-2011 at 11:30 PM.
Posted by RossH, 01-28-2011, 11:35 PM
With respect to Igor I do not know him or have ever used his product although I have heard it is a great product so I have no opinion either way. I hope he does continue to develop a great product that helps us all manage the operating system as we need to start controlling the applications we run from an OS level. So basically, I do not disagree with anything you have said. What i had a problem with is quoting certain parts of a document while making a statement that is contradictory to that document only adds to the confusion of cloud. I may have just read Igor's statement out of context or maybe he has defined his stance better in other areas. I think we need to be true in what cloud is and what the intentions of the standards bodies and committees are trying to define (which is hard). If anyone would like to learn more check out: http://www.nist.gov/itl/cloud/index.cfm http://www.cloudsecurityalliance.org/ http://www.dmtf.org/standards/cloud
Posted by cartika-andrew, 01-28-2011, 11:49 PM
I dont disagree with you Ross - but, right now, a "cloud" is severely limited at the software level. We all agree with the standards you have linked to and quoted above - no questions asked. But, as of right now, a cloud is limited by the operating system running on each instance. Microsoft Azure has developed an OS which enables the cloud infrastructure to a level beyond what most of us, including the links you referenced above, would make a reference to. Unfortunately, Azure is extremely and programatically limited. The market desperately requires an open standards OS that goes beyond what current cloud infrastructure can support. Specifically, we require granular control of resources within a cloud instance (which cloudlinux already provides) and we also require cross computational support across multiple physical instances within a cloud - which currently no one except Microsoft Azure support. Azure however, is so limited, that it is functionally useless - except for completely new development projects, that are willing to accept the Azure limitations. What Igor is doing here is developing a *.nix standard OS that will be able to take advantage and expand on current cloud infrastructure. This should be promoted - and in my opinion, heavily.. it will change our world - and well beyound the boundries of the hosting industry.. cheers man - very much appreciate this conversation..
Posted by RossH, 01-29-2011, 12:05 AM
Andrew, I wouldn't say cloud is limited at the OS level. For all intents and purposes cloud is still meant to run applications, you don't stand up a cloud server to run a Redhat server you stand it up to run an application on top of the OS. The big thing is can your software (web app, database, etc.) scale across multiple instances and be redundant. This is where we see the difference of PaaS/IaaS right? IaaS: I have to configure/manage the resources to make my application scale. The intent is not to scale beyond the physical hardware limits but to make it so I can scale up quickly to support my application. PaaS: the provider is supposed to scale this for you. For example I upload application xyz to Google App engine and if coded properly it should scale across their platform without me having to deal with any of the hardware limitations. SaaS: provider gives me the application and makes it so it can scale past normal hardware limitations. You are right this is a very good conversation to have.
Posted by cartika-andrew, 01-29-2011, 12:29 AM
Hi Ross, sorry to say, it seriously is. For example, we run nodes in our cloud that are comprised of 24 CPUs and 96 GB RAM. What happens if someone needs more resources then this? well, the answer is - fire up 2 instances and load balance across the instances. Ideally, you would just pull CPU and RAM from alternate instances and scale up the instance as far as you needed. This is simply not possible yet, and it is an OS level restriction - nothing more.. sorry, I disagree - the limitation is at the redhat level, not the cloud level. We can currently provide cloud instances that are redundant. We can seamlessly move an instance to another physical server that has the required resources to accommodate requirements, we can seamlessly move an instance to another physical server in case of failure or in case of a lack of resources currently available on a physical server - what we cannot do is pull 24 CPU and 96 GB RAM from one server and 12 CPU and 48 GB RAM from another server and serve it up as a single instance. This however is an OS and software limitation and not an infrastructure limitation. no, this simply doesnt exist yet. I think you are confusing some definitions. The ONLY OS that can accommodate what you are describing is MS Azure - and they are so limited, that their actual real world utilization is so limited, that they are functionally useless. Cross computational support is not limited at the virtualization or cloud layer - they are indeed limited at the OS and coding level. No one is able to run a common script - lets say Wordpress - at a cross computational level - it simply does not exist yet. In order to accomplish this, you would need an OS capable of accomplishing this and wordpress would need to recode their application to accomplish this. This is indeed the future, but, its not here yet. What the cloud does enable is legitimate HA environments, legitimate elasticity (within the confines of the capacity of a single physical node or multiple nodes utilizing legacy load balancing technology) as well as hard resource accountability. Nothing more - but, do not be confused by this - it is a serious and legitimate advancement.. We have large hopes for something like CloudLinux which appears set to expand these limitations even further.. absolutely...
Posted by RossH, 01-29-2011, 12:59 AM
You don't need to scale beyond 24 CPUs and 96 GB RAM to run Redat/CloudLinux/Ubuntu/etc, you need to do that because the application you are running on top of the operating system needs more resources. It is an application level restriction if you can't scale to two machines to run your applications. An operating system is meant to managed the node of hardware it is on, not to scale to multiple hardware nodes to make it so your application can scale without hardware limitations. I get what you are saying and have often thought about a system where it doesn't matter the physical hardware but the "node/vm" you are on can scale across multiple hardware pieces. This is a great idea in concept but practically it is very hard to do. At any point you have to ask "scale to what" as should a single operating system be able to scale to 24,000 servers and be able to manage it all, we just aren't there yet and I wouldn't consider it an operating system limitation as thats not what they were designed for. I think you will see platforms (think PaaS) that sit on top of the OS that help you scale this but at an OS level it is very difficult. Please see above, again operating systems aren't really meant to scale across hardware but I think you will see platforms that run on top of them that will help you scale across hardware for any applcation. This is where I have to blatantly disagree with you as I think you need to go look up the definition of SaaS. SaaS doesn't care about the OS level, for all intents and purposes it doesn't exist. You seem to be talking about something entirely different than what SaaS is supposed to be. When you think of SaaS think of things like Google Apps Mail, Salesforce, etc. An application that scales across their cloud seamlessly. When I buy google apps it isn't on one hardware node, it is across many (thousands?) nodes and seamlessly given to me. Again here is a definition from NIST: Cloud Software as a Service (SaaS). The capability provided to the consumer is to use the provider’s applications running on a cloud infrastructure. The applications are accessible from various client devices through a thin client interface such as a web browser (e.g., web-based email). The consumer does not manage or control the underlying cloud infrastructure including network, servers, operating systems, storage, or even individual application capabilities, with the possible exception of limited user-specific application configuration settings. I don't disagree but again I don't have a thought about Cloud Linux one way or another, just the definition of cloud. Last edited by RossH; 01-29-2011 at 01:09 AM.
Posted by cartika-andrew, 01-29-2011, 01:12 AM
but Ross, there is no way to scale without hardware limitations - at least not without OS and software support.. I agree, but, if the applications cannot handle it, neither can the applications running within that OS. I can tell you with certainly, right now, the current cloud infrastructure can already support it. its the OS and applications which cannot... Sorry Ross, applications run within OS's. not the other way around... Software as a Service correct? unfortunately, software is run on an OS, not the other way around - software is reliant upon the OS it runs on... the OS is reliant upon the infrastructure... no Ross, it is not - what google apps is running, is a very large and proprietary cloud, similar to what the rest of us are running this is exactly what I said... I agree with your definition of a cloud - but, you are assuming that the infrastructure behind some of these offerings are much further ahead then they really are - I assure you, they are not...
Posted by RossH, 01-29-2011, 01:32 AM
You seem to be missing the point of cloud, the point is not to scale the operating system but the ability to scale the application across either of the three service models. In no ones definition is the purpose of cloud to scale the operating system. I'd take a step back and seriously read a lot of the documentation out there to make sure what you believe needs to happen for cloud is matching the definitions. Again, operating systems do not really define/limit cloud, you need to drop that notion. They can help you at a application level to deal with cloud (e.g. GFS) but they aren't an enabler of cloud. If your application wasn't written to scale then that is the fault of the application and not the operating system, you should look at a platform that can support you. I restated "application" to platform before your post (I knew confusion would arise which is why I restated it), again think platforms here. Platforms that sit on the OS are what you need to be thinking about. Yes Software as a Service and no it isn't what you said, you went down the OS path which has nothing to do with SaaS. Again, I think you really need to take a step back and think about what SaaS is and re-read the definition. SaaS is a application given to me by a provider that is scaled seamlessly by the provider. OS, hardware, network has nothing to do with it. I think we will agree to disagree on SaaS , I'm not taking a dig at you but if you want to discuss this offline I would love to. You seem to be stopping at the operating system in most of your definitions, which isn't the definition of cloud. Cloud doesn't care about the operating system, it cares either about the infrastructure, platform or software. There are obviously operating systems, platforms, applications, etc. that can help you do this but it doesn't really define cloud nor do operating systems limit cloud. Last edited by RossH; 01-29-2011 at 01:44 AM.
Posted by cartika-andrew, 01-29-2011, 01:46 AM
Ross, I can see the point of an application ruling the Operating System, I really can. This traditionally, and more to the point, up to common technology advances, has not happened. It is a brilliant idea - imagine an application that can transcend the OS's it resides upon. If I was a smarter person, I would take on this battle and develop this technology - sadly, I am not... unfortunately, as of today, applications run within operating systems - there is no way to deny this. An operating system is installed onto infrastructure and recognizes the infrastructure it is run on. Applications are then installed within this infrastructure. Ideally, I agree with you - applications would run at some level "higher" then the OS and would be able to pull resources from whatever infrastructure they required. All I can tell you is the Azure project is the closest to accomplishing what we are discussing, and it is an OS level solution http://www.microsoft.com/windowsazur...k/default.aspx There is simply nothing else which I know of that enables an application to scale beyound the OS. Even Azure only enables you to scale beyoud a physical server, within said OS. I understand your points - clearly I do - and if I am offside, I would be the very first to admit it - but, can you show me any example of where an application resides above the OS and on its own controls the infrastructure beneath it? I agree, the provider handles it... but, it has EVERYTHING to do with the OS, hardware and network...
Posted by RossH, 01-29-2011, 01:58 AM
Andrew, You keep saying Azure as it is an OS, Azure is PaaS even admitted by Microsoft: Windows Azure Platform is a Platform as a Service (PaaS) offering (http://technet.microsoft.com/en-us/edge/ff945094) or here (http://blogs.msdn.com/b/zaneadam/arc...ed-at-pdc.aspx) or here (http://www.microsoft.com/showcase/en...6-11e0501b2ff2) or many of the documents in here (http://www.microsoft.com/windowsazure/whitepapers/) From their intro document: Windows Azure provides what’s commonly called Platform as a Service (PaaS). Rather than offering a cloud replica of the on-premises world, it offers a higher level of service that’s meant to make life easier for both developers and administrators. What follows walks through the five components that collectively provide this service.. I have never tried Azure or know nothing about how it was built but I believe it is a platform built upon the existing Microsoft operating system family, not something new. I have never used Azure but I'm sure it is a fine platform. Again, take a step back and re-think your definition. Last edited by foobic; 01-29-2011 at 03:34 AM. Reason: not pure, sure!
Posted by cartika-andrew, 01-29-2011, 02:03 AM
good enough Ross - believe it or not, we actually agree... I think the last few posts can be ignored, as i dont think much of what we were saying is fundamentally any different.. same concepts, different terminology.. cheers man...
Posted by iseletsk, 01-30-2011, 04:18 PM
What I find interesting is that each time someone new comes in into discussion, they come up with something different that is important attribute for the cloud. Yet, for me it is not clear what is really important, and what is not. For example, scalability (elastiticy). Is it important? If yes -- then does it have to scale beyond single hardware server? Why? I mean if you promise customer that his eCommerce application (shopping cart) will scale from 1000 customers per month to 50,000 customers per month -- it would be enough of scalability. Is reliability important? Yet, once again, what kind of reliability are we talking about? Should customer paying $20/month expect same reliability from their cloud as customer paying $2,000/month? What about $5/month? Resource pooling -- is that important? Which resources? How important is it to "pool" hardware resources? Multi-tenancy -- is it important? If yes, then why do we have private clouds? Sometimes I see companies setup a private cloud to serve a single application. Is that truly a cloud? Etc, etc, etc... It all can be sliced and diced in so many ways. Some things that hosting industry was doing for ages, if would be done today -- would be called "cloud". Yet, because it was done for ages -- we don't term it that way. Think of Hotmail -- if it would be done today, it would clearly be SaaS cloud. Yet, no one really thinks of hotmail as a cloud. Why not?
Posted by MikeDVB, 01-30-2011, 05:02 PM
I agree with Igor entirely - until there is a clearly defined and regulated definition for the term "Cloud" as it pertains to hosting ... everybody is going to have their own notions of what makes a cloud and what does not and what a cloud should have. Just as with "semi-dedicated" hosting - some see it as a VPS, some see it as a high end shared account, some see it as a hosting account on a "cloud" infrastructure... There is no clear industry-wide definition of either of these terms.
Posted by RossH, 01-30-2011, 05:17 PM
There is....NIST...it has already been discussed in the thread.
Posted by MikeDVB, 01-30-2011, 05:21 PM
I understand that - but until *everybody* begins using what NIST has outlined as the definition of Cloud Computing - it's no different from me posting up guidelines as to what defines Cloud hosting. There are a lot of standards organizations and boards that try to standardize things to no avail. Now don't get me wrong - I'm not saying I agree or disagree with NIST's definitions of Cloud Computing... I'm just saying that there are TONS of "Cloud" products out there that do not match the NIST definition but that doesn't stop companies from labeling the products as "Cloud".
Posted by RossH, 01-30-2011, 10:18 PM
By your posting I do not think you have the full grasp of what NIST is, it is basically a non-regulating standards body of the U.S. government. They are the ones defining what cloud is for government use. Saying your interpretation of cloud would be equal to theirs is wholly unrealistic and silly. With that being said the U.S. government doesn't own the world so there may be some more stringent definitions from say an EU body. Of course there are a ton of companies marketing their products as cloud. That is marketing and they want the business that comes from the hype. To say the definition isn't relavant because people are misusing the cloud term for marketing purposes seems somewhat self serving. I can call a dog a goat but it doesn't make it so and if I get enough people calling a dog a goat because goats are the new hotness to own doesn't make it so. Last edited by RossH; 01-30-2011 at 10:22 PM.
Posted by MikeDVB, 01-30-2011, 10:20 PM
Actually what I was saying is that their definition of what "Cloud" is, is irrelevant to this conversation. I was just saying that if everybody agreed that NIST's definition of it was correct and stuck to it then sure - it would be a valid argument. Exactly - like I said previously - the definition differs wildly from provider to provider. My goal of the post that you quoted about NIST was simply a hypothetical tangent of the conversation.
Posted by iseletsk, 01-30-2011, 10:45 PM
I think the way NIST defines the cloud -- bunch of service hosting providers were doing for years would fall under that definition. It is broad, not exact, and covers most of the hosting services. It would fall into either PaaS or SaaS category with ease.
Posted by RossH, 01-30-2011, 10:51 PM
Mike...I don't think we can go any further in this discussion as you wouldn't be willing to accept anyones definition of cloud because companies use the term to market it for different things. Standards bodies set the definition not the hosting companies and their marketing machines. If you can't figure out what cloud is by the definitions set forth or are unwilling to accept them then I cannot help you.
Posted by MikeDVB, 01-30-2011, 10:53 PM
Not looking for help, but thanks anyhow
Posted by RossH, 01-30-2011, 11:09 PM
Igor, I've read your post a few times and am still confused by it. The characteristics you mention are ESSENTIAL to what cloud is (except multi-tenancy), it isn't open to debate on what cloud is. The true idea behind cloud is not single cloud but multiple clouds. I have a "cloud" (pool) of servers that uses a "cloud" (pool) of storage that has access to a "cloud" (pool) of networking. All of this together forms the cloud that we are talking about today which is why they mention resource pooling. You pool compute, storage, network together to form a cloud to drive the application you need to run. Now for the service models you mention, private cloud, you'll notice in the NIST definition that multi-tenancy isn't mentioned as an essential characteristic but to be able to do it as part of the resource pooling. You can have a private cloud dedicated to a customer that fits the cloud model because there are still pools of resource to allow the customer to scale the application. Again think of IaaS as a pool of resources; storage, network, compute. Then PaaS is a layer that sits on top of IaaS to help you scale that application without having to worry about managing IaaS and SaaS is they give you the application that sites on top of IaaS and scale it for you. Private clouds are typically for larger financial customers that don't want their resources shared. You'll notice in NISTs definition they say: On-demand self-service. A consumer can unilaterally provision computing capabilities, such as server time and network storage, as needed automatically without requiring human interaction with each services provider. Not local storage. The idea behind cloud is that you have a set of resources that include multiple physical nodes that can be provisioned/de-provisioned on demand by a client to help them scale.
Posted by RossH, 01-30-2011, 11:23 PM
If you have ever been a part of one of the standards meetings you would understand that they are intentionally left broad as to not restrict the providers in how they provide service and also because there are a lot of business interests at stake. I don't disagree with you that many services even today could fall into PaaS/SaaS territory and honestly in my opinion those shouldn't have been included in the cloud definition. However they are and providers still need to meet those definitions. Convoluting the term cloud only drives away the people/companies willing to put forth the capital to provide a superior service to their customer and push forward innovation.
Posted by erickmiller, 01-30-2011, 11:23 PM
Just a quick question regarding this failed server and how Applogic handled it. Why would you want a failed server to come back online, or for that matter, be automatically rebuilt? Wouldn't you want to find out exactly what happened and repair it (such as doing memory tests and replacing bad memory, since that's a common culprit for crashed servers) before putting it back into service? Do Applogic nodes fail that often due to software crashes? Or was this a hardware issue? Just curious if anything. Eric
Posted by iseletsk, 01-30-2011, 11:38 PM
Ross, 1. Essential characteristics are things like Resource Pooling or Rapid Elasticity. Not resource pooling of memory, network, processing, storage, etc... Those are just examples of possible resource pools. Nothing in this standard defines that cloud has to pool that particular set of resources. Same with elasticity -- it doesn't specify how much elasticity there should be. Note critical phrasing: It should "appear" to be unlimited... Just an example: Shared hosting providers were proving unlimited disk space and bandwidth for ages, because it "appears" to be unlimited to their customers. 2. Why did you remove multi-tenancy? Why "except multi-tenancy"? Private cloud can still be multi-tenant. And yes, I read the definition, but those are examples, "such as" doesn't mean for me same thing as "it has to include". Those are examples of particular essential properties. yet, those are just examples. You might not need network storage, and still be a cloud. You might not need to allocate server time, and still be a cloud. And I am sorry, but I disagree. You don't need to sit on top of IaaS to be PaaS. Why would you? Why can't I run my own hardware/datacenter not as a cloud but as a simple cluster that I setup, and provide PaaS? Why would it be any less of a PaaS, then if I run it on top of IaaS?
Posted by che09, 01-30-2011, 11:58 PM
CloudLinux is the first commercial OS that meets the specific needs and demands of web hosting. Generally, a cloud-hosted website is opearting on multiple connected servers. Instead of limited to a single server like what we have in traditional hosting services (dedicated/shared hosting), the website now has the access to multiple servers. Virtually, the processing power is unlimited as you can always add a new server and scale up.
Posted by JasonD10, 01-31-2011, 12:01 AM
In my experience through the past 13 years of administrating servers, most all crashes are not related to hardware failures. Maybe I'm just lucky, or maybe I just build my servers with excellent hardware and my burn-in procedures help reduce this as servers go through a 5 day period of burn-in and tests before being put into production. However, we still see periodic server crashes which are mostly due to high load and some form of software problem, kernel, driver, or other issue. Hardware failures of course do occur, but more often than not when they do occur it's a complete failure. This kind of action does not need to happen and you can configure your settings accordingly, but I too take advantage of the auto rebuilding and am quite happy with the results. AppLogic itself is very stable and rarely does a node crash on mine. We did have a hardware failure two weeks ago in a Cloud, and the backplane board stopped functioning showing no drives. In this case (complete failure of a critical component) AppLogic of course could not put the server back into operation.
Posted by Richboos, 01-31-2011, 12:53 AM
The government standards body defining something doesn't mean things with be agreed upon and inter-operable. Think about the W3C with HTML and CSS. What a disaster that was for years with "This site best viewed in Netscape/Internet Explorer/Etc." The cloud becoming an agreeable standard may take as long as IPv6 to be 100% implemented.
Posted by erickmiller, 01-31-2011, 01:14 AM
We perform significant burn-in tests as well, but have seen memory failures creep upward with larger DIMM sizes, but after years of operation under high load. I guess I haven't seen that many software failures on ESX. There has been the occasional driver that caused a kernel fault, but it's pretty rare (twice in 5 years?), which is why I asked about AppLogic's stability. It just doesn't happen much on our ESX boxes. I guess I should say, ESX on HP hardware. I'm not too familiar with other hardware with ESX long-term. We also had a hardware failure in an HP DL385 G5p recently where a motherboard went bad somewhere. No obvious cause of failure, but replacing the board was the only thing that solved the problem. The unit simply powered off and never powered on again until the board was replaced. Funny how things like this can run for years and all of a sudden, something gives. With memory failures, we know ahead of time if ECC corrections have occurred so we can replace memory before failure, but it doesn't always happen that gracefully, unfortunately. Sometimes memory simply fails entirely before showing signs of errors. Eric
Posted by RossH, 01-31-2011, 01:46 AM
Again you seem to be missing the point. Yes resource pooling is defined in the NIST document if you took the time to read it: Resource pooling. The provider’s computing resources are pooled to serve multiple consumers using a multi-tenant model, with different physical and virtual resources dynamically assigned and reassigned according to consumer demand. There is a sense of location independence in that the customer generally has no control or knowledge over the exact location of the provided resources but may be able to specify location at a higher level of abstraction (e.g., country, state, or datacenter). Examples of resources include storage, processing, memory, network bandwidth, and virtual machines. It doesn't say example of possible resource pools, it says examples of resource pools. It doesn't specify elasticity because it isn't their job to specify that, it is up to the consumer to find a comptent provider to provide them the adequate resources to scale. I'm honestly not even going to go into the shared hosting model with you any more as it is self serving for you and you keep trying to discuss it as it is legitimate to call shared hosting cloud for your marketing. Because multi-tenancy isn't part of the definition. Private cloud cannot be multi-tenant which shows you severely lack the ability to grasp the definitions of the technology. Go look at the definition: Private cloud. The cloud infrastructure is operated solely for an organization. It may be managed by the organization or a third party and may exist on premise or off premise. This means one company/organization owns/rents that cloud, no one else. It cannot be multi-tenant by definition. Either you haven't read the definition or lack the ability to comprehend it. You must have network storage to fit the first definition of cloud: On-demand self-service. A consumer can unilaterally provision computing capabilities, such as server time and network storage, as needed automatically without requiring human interaction with each service’s provider. I'm fine with anyone disagreeing with the definition as long as they have a coherent argument but to say that the definition doesn't say what it defines is pure ignorance and anyone should steer clear of a company that argues as such. You can disagree all you want but the leaders in the industry will keep following the standards. Last edited by RossH; 01-31-2011 at 01:52 AM.
Posted by RossH, 01-31-2011, 02:34 AM
Government standards bodies did not try to define HTML/CSS last time I checked so your argument is pretty much null/void. IPV6 is fully defined, no one is arguing what IPV6 is. It may take some time for it to be implemented but we won't have arguments like we are having today such as "What is cloud." There is a big difference, if you can't see it then I'd suggest you read some more on the standards and definitions.
Posted by Richboos, 01-31-2011, 03:36 AM
I didn't say that a government standards body defined HTML/CSS. I also didn't say that IPv6 wasn't defined. What I mean is that certain standards won't pickup until vendors and consumers start implementing them. This could mean that we still could witness many more variations on "what is the cloud" regardless of what the NIST defines. That is why I mentioned the W3C(not government, but still a valid point).
Posted by ramnet, 01-31-2011, 03:39 AM
Well Al Gore thinks he invented the Internet..... Private industry doesn't always agree with or follow along with what the government is doing.
Posted by niroh, 01-31-2011, 04:42 AM
is it the os you are talking about ? i had some reviews that it is good especially when it comes to limiting the cpu resources
Posted by iseletsk, 01-31-2011, 08:35 AM
Ross, I actually think you are missing the point. Examples means just that -- EXAMPLES. It can be any of those, or none of those. Hence Examples. And multi-tenancy can still exist for one organization. One organization can have multiple departments, multiple distinct projects -- and still have multi-tenancy. Two projects on same cloud is still multi-tenancy. Please, try to come to this with open mind. Cloud definition is short, not exact and very broad. It was defined that way. I would agree that if there is NO resource pooling, it would not qualify as a cloud. Yet, I believe it was on purpose that particular resources that HAD TO BE PULLED were not specified. Examples are there to show most common scenarios, but they are not essential qualities. They are just examples. Oh, and btw, don't make any mistakes. leaders WILL NOT follow standards. Leaders will DEFINE standards. That is how it always been. They are the leaders. Standards are made later on, for the followers. And one more thing Ross -- for some reason you think it is self serving for me -- yet, I don't know why. I really don't care if shared hosting is a cloud or not. It makes no difference to me. I just see bunch of people who do cloud, and they thing that the only way to do cloud is the way they do cloud, without realizing there are so many other ways to do cloud as well. What ever you do in your day job is just ONE WAY to run cloud. Not the only way. Another cloud might look nothing like yours, serve completely different purpose, be used in completely different way -- yet still be a cloud. Last edited by iseletsk; 01-31-2011 at 08:41 AM.