Page 1 of 2 12 LastLast
Results 1 to 40 of 62
  1. #1
    Join Date
    Feb 2011
    Posts
    39

    Thoughts about Cloud vs VPS Hosting?

    In terms of overall performance and scalability, which of these two is the better choice?

    Also, which of these two is the cheaper route?

  2. #2
    Join Date
    Feb 2003
    Location
    NY
    Posts
    11,521
    It all depends really on who your market is. There are some who are pro cloud and some who are still pro VPS. To me its more of a matter of preference, but knowing how technology sectors move so quickly, Cloud is probably a safe bet or at the very least good to have knowledge of.

    A lot of business technology is moving to the "cloud" with remote access, work from home, programs of that nature which are utilizing cloud technologies...

  3. #3
    Join Date
    Feb 2011
    Posts
    39
    I see. Thank you!

  4. #4
    Join Date
    Dec 2004
    Location
    New York, NY
    Posts
    10,574
    If you just want performance and scalability, honestly a normal VPS should do just fine for you. You don't necessarily need the typical cloud infrastructure.

  5. #5
    Join Date
    Feb 2011
    Posts
    39
    Thank you for your reply!

  6. #6
    Join Date
    Mar 2009
    Location
    Here Today - Gone to Maui
    Posts
    6,262
    Cheaper depends on your market. Does your traffic spike by events, seasons or promotions? You need to match your specific requirements to the pros and cons of each. Generally speaking, VPS would be cheaper while cloud could scale easier.
    ProlimeHost- Dedicated Server Hosting & KVM SSD VPS
    Three Datacenter Locations: Los Angeles, Denver & Singapore
    SuperMicro Hardware | Multiple Bandwidth Providers | 24/7 On Site Engineers

  7. #7
    What is the difference between cloud hosting and cloud computing?

  8. #8
    Join Date
    Sep 2009
    Posts
    361
    Based on what I saw over at BirdHosting.com and how Applogic works for them, Cloud could potentially trump VPS in almost every arena.

  9. #9
    Join Date
    Apr 2010
    Location
    Nepal
    Posts
    40
    Cloud is very flexible, where as VPS is stable and efficient. In Cloud you can have various advanced functions and settings compared to vps.
    Shared Hosting, Reseller Hosting, VPS and Dedicated Servers
    www.AQuickHost.com
    Affordable, Fast, Reliable and 100% Satisfaction Guaranteed
    Host with us today!

  10. #10
    Join Date
    Jan 2011
    Location
    Canada
    Posts
    934
    To cut through the marketing hype, you should think of it not as whether it's Cloud (new) versus VPS (old), but instead, look at how the components are architected.

    The main components you need to consider are:

    1) File object storage
    2) Front ends (eg: Apache)
    3) Load balancers
    4) Database nodes
    5) Monitor and control

    A good cloud implementation, should have each of these separated and redundant. Unfortunately, not all hosting flying under the cloud banner will give these so be warned.

    If you want to do an apples to orange comparison, then of course, a cloud host that sets up these will have better performance and stability than putting it all in one VPS. The truth is though, if you have the expertise you can setup the above on any system, whether its VPS or dedi.

    Whether you want to do that yourself is another matter.

  11. #11
    Join Date
    Apr 2002
    Location
    Auckland - New Zealand
    Posts
    1,572
    Moving to a service such as someone using Applogics cloud, gives several advantages over the typical single or home made SAN 'cloud providers' and VPS hosters that put everything in 1 box. If you don't want single point of failure or lengthy storage synchronisation rebuilds, then choosing someone running applogics cloud (done right) looks like it will be the better option right now. Looking forward to Openstack maturing as well, as that is going to be another silver bullet by the look of things.

  12. #12
    You may want to look into Cloud VPS Hosting.

  13. #13
    Join Date
    Aug 2000
    Location
    Phoenix, Arizona
    Posts
    558
    tchen is completely right. I have shopped around and talked to many of the "big guys" selling cloud hosting and I found that when I pressed them on the technical specs they were all over the board. Even Amazon's EC2 cloud has zero redundancy of data unless you specifically know how to do this yourself and purchase separate load balancing instances as well as a second instance for your data to be replicated to.

    I also found that many providers are simply doing the same old stand alone, single server VPS hosting and only rebranding it as "cloud". While technically it is a cloud in that it is a virtual machine, it doesn't have redundancy or high availability. Even if someone is using a centralized SAN storage unit, unless they have a really high end unit, that has data redundancy built into it, you still do not have HA. If the SAN fails, EVERYTHING fails. Single point of failure.

    Applogic is a good product but still very early in it's development cycle. Their documentation is horrible so it all depends on how smart your host is who is running and managing AppLogic.

    So the bottom line, is ask lots of questions and know what you are getting. Make sure that your VM will failover if the node fails but also make sure that there is redundancy with the storage so that you don't loose everything when the SAN fails.

    Most importantly, no matter what you are doing, BE SURE AND BACK UP YOUR DATA off of the VPS or Cloud.

  14. #14
    Join Date
    Aug 2009
    Location
    INDIA
    Posts
    43
    As money is considered VPS will be cheaper than Cloud. Cloud is still in the early stages, need to evolve ( facilities,hardware softwares etc..) In the future surely cloud will be become more affordable.
    reloadIN - Onshore & Offshore Web Hosting Solution
    VPS | DEDICATED | CLOUD | VPN SERVERS
    █ Server Location - USA|RUSSIA|GERMANY|NETHERLANDS|INDIA

  15. #15
    Join Date
    Feb 2003
    Location
    NY
    Posts
    11,521
    I think in the future Cloud will become the new norm. For instance companies are moving to more Work From Anywhere programs which utilize the Cloud and licenses that access the Cloud which eventually will become the mainstay on the hosting industry (Until the Cloud becomes data).

  16. #16
    Join Date
    Mar 2004
    Location
    Cheshire, UK & WA, USA
    Posts
    202
    Guys,

    I think you have to be careful here comparing two things that are not in competition.

    Cloud is an ethos/structure (for birdhosting anyway) which allows us to pool resources together from multiple devices / locations and allocate them to different users / applications.

    VPS is simply a virtual private server, so you get the features of a dedicated server but without having to pay out for the full hardware.

    Are you maybe comparing hosting a website on a "cloud host" vs hosting it on a VPS?

  17. #17
    To StevenG,

    What would be your definition of a someone who is "running applogic's cloud (done right)". I have been testing this platform for some time and have to admit, running single SATA drives in Applogic (as IP SAN) is something that is scarring me the most. I'm all old fashioned and RAID for me is the last bastion in defending from single (SATA) drive failures but Applogic is not 'supporting' that approach.

  18. #18
    Join Date
    Jan 2010
    Posts
    1,028
    Quote Originally Posted by gocloud View Post
    To StevenG,

    What would be your definition of a someone who is "running applogic's cloud (done right)". I have been testing this platform for some time and have to admit, running single SATA drives in Applogic (as IP SAN) is something that is scarring me the most. I'm all old fashioned and RAID for me is the last bastion in defending from single (SATA) drive failures but Applogic is not 'supporting' that approach.
    You can run a local RAID if that is what you prefer, but also keep in mind that your data is mirrored across multiple servers it's not just on one. So assuming you're not running a local RAID and the server dies, another one that has the volume stream will run your content and start rebuilding that stream on another server after the failure or when you replace the downed server.

    In the worst case scenario you are running a non-local RAID1 (Ie: two IP SANs with the same data).

  19. #19
    Join Date
    Apr 2002
    Location
    Auckland - New Zealand
    Posts
    1,572
    Cloud to me isn't about having expensive SAN's and complicated synchronous replication or single points of failure.

    I prefer the openstack/applogic approach, with commodity boxes.
    The data is mirrored and replicated between the disks available to the storage pool. There isn't any need for raid either in that situation.
    That is what cloud storage is to me. The XIV kind of approach to SAN, without SAN.

    You can use hardware raid if you want, but for me there is little point in it, these approaches eliminate the need.

  20. #20
    Join Date
    Aug 2000
    Location
    Phoenix, Arizona
    Posts
    558
    we used applogic for a solid 8 months or so before canceling the whole contract. Not only was the documentation so terrible that it was almost worthless, when you actually do have to "fail over" it is not immediate. It can take 15 or 20 minutes to fail over if it even fails over. We found that often the volumes would get degraded in some way in an event like that and we ended up having more downtime with AppLogic then we did with the same clients on stand alone dedicated servers.

  21. #21
    To StevenG and CloudWeb,

    Thanks for your input but just last night I lost a volume (in a grid with 4nodes/3x sata drives per node) after a single disk failure in srv2 forced me to migrate data around and in the migration process (300GB+), which took almost 18 hours to finish, I had another disk failure in srv3 where some of volumes being copied to.

    It may be coincidence but this incident forced me back to drawing board how best to organize storage and protect it from single drive failures like this.

    To Atjeu,
    I understand your 'pain' with documentation (or lack of it) still for me Applogic is very much unique and interesting platform to get it binned.
    I'm not there yet but I guess by getting your Applications more complex it will certainly result in increased restart time. I'm not sure if this can be addressed purely by throwing more hardware (faster drives?) into it or is it Applogic design 'issue' (ot the way how their scripts work).

    StevenG or CloudWeb could have a better view as they seam to 'use' Applogic a lot.

  22. #22
    Join Date
    Aug 2000
    Location
    Phoenix, Arizona
    Posts
    558
    that sounds just like some problems that we had. We also had a tech go through their cloud certification program in California ($2000) so it wasn't like we were a passive user. I think AppLogic has some really good concepts. The graphical user interface for provisioning is what sold me on it to begin with. I just think it is a bit early in it's development cycle.

    Since they were bought out by CA I would hope that they will sink some more cash into the product and bring it up a notch. But the comparison between using Citrix Xenserer or VMWare is just not even close right now though. Both VMWare and Xenserver have flawless documentation and you can learn them in 10 minutes. AppLogic will take you days or the better part of a week to setup initially and get all your images and everything set. Setting up Windows ISOs was an absolute nightmare.

  23. #23
    Join Date
    Jun 2008
    Location
    San Diego, CA
    Posts
    510
    Quote Originally Posted by atjeu View Post
    we used applogic for a solid 8 months or so before canceling the whole contract. Not only was the documentation so terrible that it was almost worthless, when you actually do have to "fail over" it is not immediate. It can take 15 or 20 minutes to fail over if it even fails over. We found that often the volumes would get degraded in some way in an event like that and we ended up having more downtime with AppLogic then we did with the same clients on stand alone dedicated servers.
    Sounds like 2.7.8.

    Those problems have long been fixed in 2.8.9 and 2.9.9. Failover is virtually instantaneous now and we have had no issues since these updates were released.


    EDIT - I cannot disclose much, but the 3.0 release will be the most significant for this platform in over 30 months. It is very impressive right now.

  24. #24
    Join Date
    Aug 2000
    Location
    Phoenix, Arizona
    Posts
    558
    we were using 2.7 and then 2.8. That is really good to hear. Strange that they didn't talk to me about that when I canceled everything though.

    The other issue was that there is no easy way to do an upgrade the next big release. has that been fixed?

  25. #25
    Join Date
    Apr 2002
    Location
    Auckland - New Zealand
    Posts
    1,572
    I'm not using Applogic at all, I was more referring to cloud storage philosophies.. commodity boxes, no san, no fancy raid cards, easily scalable. Hypervisors these days are 2 a penny and have their own pluses and minuses.. where it's at for me is the storage level.

    I've long awaited the storage layer to grow up, and with openstack and applogics approaches, it finally is coming to reality for the masses.
    I am going to build a cloud storage platform for a large isp in nz, in the coming months and I'm taking my philosophy as the approach for that.
    We already run vmware clusters and utilise san, that's a totally different platform and not anything remotely associated with cloud to me, anyway.

  26. #26
    Join Date
    Jun 2008
    Location
    San Diego, CA
    Posts
    510
    Quote Originally Posted by atjeu View Post
    we were using 2.7 and then 2.8. That is really good to hear. Strange that they didn't talk to me about that when I canceled everything though.

    The other issue was that there is no easy way to do an upgrade the next big release. has that been fixed?
    Upgrades still require you to work directly on the grid controller and the nodes. So downtime is a must for them. What we often do is a "staggered" upgrade. Bring up a new grid in the same DC with the new software, then migrate volumes over to it. We can coordinate with customers and it requires far less downtime for them (usually seconds to a few minutes). As customers move off the old grid and resources are freed up, you can shrink the grid a node at a time, adding those nodes to the new grid. In the end, it takes longer, but is transparent to the end user.

    I wish I could tell you about 3.0. Lets just say a LOT of the things above you voiced will be fully addressed. Think outside the box about everything you know about AppLogic.

  27. #27
    Join Date
    Aug 2000
    Location
    Phoenix, Arizona
    Posts
    558
    Thats really good to hear brent. I know CA's resources were a big help to them.

    StevenG- vmware and Xenserver are definitely a cloud as long as you have your storage replicated on multiple nodes either in the same facility or in multiple facilities. There are a handful of ways to accomplish this. Some are super expensive but there are open source options or cheaper options to handle replicating the storage.

  28. #28
    Join Date
    Apr 2002
    Location
    Auckland - New Zealand
    Posts
    1,572
    StevenG- vmware and Xenserver are definitely a cloud as long as you have your storage replicated on multiple nodes either in the same facility or in multiple facilities.
    They can be a cloud, but whether its vmare, xenserver, kvm, xen whatever, thats not the part that really defines cloud. Most aren't cloud that claim it, especially in budget hosting. Charging hourly for a service connected to a single budget iscsi san.. or any premium spof san is not cloud.

    I prefer the multiple nodes of disk, replicated and highly available, running on commodity hardware, than any san ideas. See lots of people failing in the storage area all the time, nearly every vps/cloud provider out there has a horror story to tell about their storage technology that they've been trying to implement or have in the past few years.

    SAN has it's place, for high performance, or for specific application requirements, pci compliance, lots of reasons. To me, it's not the storage platform for my cloud.

  29. #29
    Brent,

    Would you say that replacing sata drives in my nodes with ssd sata drives (with Intel SLC drives) I would be able to see (availability and) performance gains in comparison to sata drives (currently 1TB WD).

    My overall applogic storage usage is really modest, altogether less then 150GB, and it is not expected to grow a lot more so I do not need 1TB drives.

    I believe I read in some other posts of yours that you used SSDs (or maybe it as SAS drives with HBA) in some applogic cloud deployments.
    If that's the case then,
    Does raw power performance gain influence the applogic volume migration process and overall grid performance (starting/stopping applications, running applications, etc.)?

  30. #30
    Join Date
    Jun 2008
    Location
    San Diego, CA
    Posts
    510
    So, you will see mixed results with JUST upgrading your drives. And here is why I say that.

    AppLogic, as you know, uses a "backend" network that is dedicated to keeping all the server nodes synchronous (I.E the Local IP SAN). If you have a lot of changes to your local drive(s), then you are limited by the speed of that network to keep them synchronous. 1Gb Ethernet is 125MB/s, which is far slower than the read speed of a good SSD and slower than the write speed of those Intel drives, especially if you RAID them up. So you could see your systems just putter along at 50-70MB/s (1/2 the max speed since data moves both ways), while you wait on them to sync.

    For this reason, we have been investigating different backend options, such as 10GBE and Infinniband, as we run into these issues as well.


    If, however, your data usage is low, and you do not do a lot of writes, then you will see pretty good gains.



    You will need to describe your application a bit more for me to be able to give a better answer than that.

  31. #31
    Join Date
    Mar 2011
    Location
    Switzerland
    Posts
    3
    Hello Atjeu,

    interesting. May I ask you 2 questions; when were those 8 months (for me to know which applogic versions you used), and what are you using today for your cloud backend? Thanks.

    Cheers,
    Michael


    Quote Originally Posted by atjeu View Post
    we used applogic for a solid 8 months or so before canceling the whole contract. Not only was the documentation so terrible that it was almost worthless, when you actually do have to "fail over" it is not immediate. It can take 15 or 20 minutes to fail over if it even fails over. We found that often the volumes would get degraded in some way in an event like that and we ended up having more downtime with AppLogic then we did with the same clients on stand alone dedicated servers.

  32. #32
    Join Date
    Aug 2000
    Location
    Phoenix, Arizona
    Posts
    558
    we used AppLogic from January of 2010 to about august of 2010.

    We have tried a few different things but today we are using the Enterprise version of Citrix Xenserver with a replicated SAN for central storage. I have found it to be 1000 times easier to work with than AppLogic which also makes it a lot easier to train a client on if they do a private cloud of their own. With Xenserver you can train someone in about 10 minutes. I would guess that even with the improvements that have been made with AppLogic, that it still takes a while for someone to really know how to use it.

  33. #33
    Brent,

    How's your 10GBE (and Infiniband) testing going? What 10GBE switch and LAN adapters you choose for testing and would you recommend any at stage that you are currently on with testing.

  34. #34
    Join Date
    Jun 2008
    Location
    San Diego, CA
    Posts
    510
    These are interesting questions, which actually have a somewhat complex answer.

    We had run into several snags with IB. Basically, we can compile the drivers for IB into the base kernel for AppLogic 2.9.9, but cannot bind the adapters to the Gird OS itself. CA is currently working on a fix for this issue, but it is completely out of our hands and we do not expect a patch until after they release the 3.0 milestone (which is a HUGE HUGE upgrade we are dying to see as well).

    For this reason, we have started to deploy 10GBE to improve our backend network, even though the cost per port is 2-4X higher than 40GB IB. One issue we have run into is that the kernel that AppLogic 2.9.9 uses is not compatible with some features implemented in the AMD Magny Cours CPUs, when running either 10GBE or IB. Basically, certain low-level register calls cause the kernel to panic if you are using these controller. Unfortunately, these CPUs constitute 75% of our deployed hardware, and this has been a big setback for us.

    On a happy note, we just today got a new grid setup with 10GBE and Intel E5540 CPUs with Silicom 10GBE cards (supported out of the box by AppLogic). The base OS is installed, and we will be testing in the next 2-3 days. Hopefully we will not see the problems we did with the AMD-based servers, as 3 other AppLogic MSPs have verified they have deployed 10GBE on Intel hardware in the past.

    For switches, we went with Fulcrum Micro 10GBE CX4 switches. They have the highest I/O performance of any switch out there for 10GBE, and they are very reasonable priced, if you can find them (they are not a big market company).




    If all goes well, then we will be deploying this 10GBE grid in our Seattle DataCenter over the weekend and if it pases testings we will be releasing our HA, Load Balanced, cPanel Cluster at that time. We are so excited about this deployment, as it has been 6 months in the making. This cluster is like nothing anyone has EVER deployed for cPanel. It does not run off a single VDS. Instead, it is 100% redundant, with different components of the stack being able to be spread across multiple nodes. It is High Availability (you can have all but one web server, in-gateway, load balancer, and mysql server fall out of the cluster, and the end user will never notice). And it is FAST. We have deployed every tweak we could think of to make the performance outstanding. It runs an accelerated PHP, an optimized MYSQL, and to top it all off runs UNIXy's Varnish plugin.

    Oh, and since it runs on AppLogic, I can literally move copies of this cluster to our other DataCenters with a one-line command for rapid re-deployment. Pretty freaking cool if you ask me.

  35. #35
    Join Date
    May 2007
    Posts
    451
    Most Typical VPS's are shared resources. So depending on how loaded their VPS servers are depends on what kind of speed your going to be getting.

    VDS's (Virtual Dedicated Servers) offer 100% dedicated resources. What Applogic offers is us to completely dedicated Disk/Memory/CPU to a given application.

    So depends on what your want/need. Also depends on how "loaded" a given server is.

    Good luck with choosing a provider.

    Thanks,

  36. #36
    Cloud is better than VPS, however cloud is a new technology it is always good to try something new.. my site is currently running on cloud hosting. never face a problem till date... its Awesome and super fast.

  37. #37
    Hi Brent,

    You are deeply into 'heavy' stuff with Applogic
    Hoping that you will be able to find interested parties/clients to utilize the solution that you are building.
    I too read before of all kind of problems with IB and specifically lack of driver support. 10GBE seams sure bet for the time being but still there are lot of 'unknowns' in Applogic case. Any particular reason why you didn't choose Force10 XFP 10GBE suggested/certified 10GBE switch with Intel 10GBE LAN cards?

  38. #38
    Join Date
    Jun 2008
    Location
    San Diego, CA
    Posts
    510
    From our experience, when you get above 6-8 cores per server node (our typical node is 16 cores), you start running into back-end network bottlenecks that prevent the nodes from keeping in sync as quickly as we would like. A 1GbE pipe (125MB/s) is simply not enough bandwidth, as most HDDs this day and age can burst data that fast. And that is before you get into RAID configs on your local machine.

    This manifests itself as increased iowait, even though the actual disk sub-system is more than fast enough for the requests being made to it by clients. That is why we went to 10GBE/IB, to remove this bottleneck and give us room to grow for future SSD-enabled storage options on the IP-based SAN.

    The cards we chose to use are based upon the Intel 10GBE chipset, just not Intel branded.

    The switch, well, is not just an amazing piece of hardware for its I/O, but we got a deal on a pair of them that I simply could not pass up. It made the "pain" of swallowing 10GBE a lot less, but the deployed cost for 10GBE was still substantially more than 40GB IB, and you wind up with just 1/4 the bandwidth.


    The problem with IB as it stands currently with AppLogic is no longer drivers. With the help of the Mellanox developers, we have a kernel that now has drivers compiled into it and the performance is amazing. The problem currently is that AppLogic cannot bind the virtual network ports to the physical IB adapter. Until they are able to solve that problem, we are dead in the water on IB.

  39. #39
    Brent,

    When you say 'you start running into back-end network bottlenecks that prevent the nodes from keeping in sync as quickly as we would like'..
    Do you actually refer to volume(s) repairs/checks/migration?

  40. #40
    Join Date
    Jun 2008
    Location
    San Diego, CA
    Posts
    510
    Anything involving I/O synchronization. So if you access a VDS that has a large database, you will see that VDS be less responsive than if the same VDS were on a system with more back-end bandwidth.

    Vol-repairs and migrations WITHIN the cloud also take longer to complete.


    These are not to the level that our customers notice a difference, but when performing maintenance or planning a larger deployment, we have to keep them in mind.

Page 1 of 2 12 LastLast

Similar Threads

  1. Replies: 7
    Last Post: 08-26-2011, 06:28 PM
  2. Thoughts on hardware for Cloud
    By ewitte in forum Cloud Hosting
    Replies: 23
    Last Post: 09-24-2010, 03:02 PM
  3. Replies: 0
    Last Post: 05-17-2010, 01:11 PM
  4. Thoughts on Godaddy's new MacOSX Server Cloud Hosting
    By vitalis in forum Cloud Hosting
    Replies: 1
    Last Post: 04-17-2010, 07:32 PM
  5. Replies: 0
    Last Post: 01-25-2010, 01:18 PM

Posting Permissions

  • You may not post new threads
  • You may not post replies
  • You may not post attachments
  • You may not edit your posts
  •