Page 3 of 17 FirstFirst 12345613 ... LastLast
Results 51 to 75 of 416
  1. #51
    Join Date
    Oct 2006
    Posts
    68
    PHP5, DB4 and PGSQL8, basically this is it
      0 Not allowed!

  2. #52
    Join Date
    Dec 2004
    Location
    New York, NY
    Posts
    10,710
    I've been testing out LT's grid and I have to say it truly is an excellent product. If they keep it up with the high performance and reliability, I'm sure this can be seen as a great alternative to a VPS -- as that is essentially what it is (powered by Xen after all), other than the fact that it's very redundant
    MediaLayer, LLC - www.medialayer.com Learn how we can make your website load faster, translating to better conversion rates for your business!
    The pioneers of optimized web hosting, featuring LiteSpeed Web Server & SSD Storage - Celebrating 10 Years in Business
      0 Not allowed!

  3. #53
    Join Date
    Nov 2006
    Location
    Southern California
    Posts
    173
    Quote Originally Posted by TigerHosting
    PHP5, DB4 and PGSQL8, basically this is it
    So, how large a cluster would you like to have?
      0 Not allowed!

  4. #54
    Join Date
    Nov 2006
    Location
    Southern California
    Posts
    173
    Quote Originally Posted by layer0
    I've been testing out LT's grid and I have to say it truly is an excellent product. If they keep it up with the high performance and reliability, I'm sure this can be seen as a great alternative to a VPS -- as that is essentially what it is (powered by Xen after all), other than the fact that it's very redundant
    Thank you on behalf of both our teams - 3Tera, who built the software, and LT, who implemented the grid and keeps it up and running. The grid servers are just scratching the surface of what AppLogic can do - stay tuned as we roll out the true capabilities of the grid in the weeks to follow...

    BTW, LT and 3Tera are trying to set up an online WebEx demo next week to show off what the system can do. If you are interested, keep an eye on the posts here - someone should be posting the day/time, etc. over the weekend.
      0 Not allowed!

  5. #55
    Join Date
    Nov 2006
    Location
    Southern California
    Posts
    173

    How AppLogic compares with MT, Mosso and other grid offerings

    I have been seeing quite a few posts where people are trying to compare the Grid Layer, Media Temple, UtilityServe, Mosso, Amazon EC2, etc. As with any new technology, there is not enough data and speculations abound. So I decided that it may be useful to provide some information that will make these comparisons easier.

    As many of you know, I am the CEO of 3Tera and the architect of AppLogic - the grid OS that powers the Grid Layer and UtilityServe. As such, I cannot and do not speak for any of the hosting providers involved. It is also not appropriate for me to speak about the offerings that are not based on AppLogic until I understand better the technology involved. What I can do, however, is give you a concise overview of the kind of services that AppLogic supports today or is likely to support in the next 90 days or so. Hopefully, this will help the community compare the different services available on the market.

    I. GRID SERVICES

    Any hosting provider implementing a grid on AppLogic will be able to provide any or all of the services described below if they choose so. Note that AppLogic is simply the OS that runs the grid - each hosting provider is free to define their services based on their market focus and business priorities. If you like something and your favorite host does not offer it yet, let them know that the demand is there :-)

    Basic infrastructure:
    ------------------------
    - grid servers (available today)
    - grid racks (2-16 grid servers connected with a private switch, coming soon)
    - grid clusters (firewall + load balancer + 2-8 web servers + database server + NAS, coming soon)

    These are basic building blocks on which you can install your software. Grid racks are useful for things that have their own clustering software, such as game servers and proprietary apps. Grid clusters are for web applications.

    Web hosting infrastructure:
    ----------------------------------
    - Cpanel clusters (in Beta now)
    - SiteKreator clusters (in integration)
    - Ensim clusters (in planning)
    - H-sphere clusters (in planning)

    These are completely integrated clusters that are ready to run and allow you to manage up to 16-32 servers in a fully transparent & load-balanced fashion (like having a single huge server with up to 64 CPU, 128 MB RAM and 10TB of disk space)

    Application infrastructure:
    --------------------------------
    - LAMP clusters (available today)
    - J2EE clusters (coming Q1 2007)
    - RoR clusters (coming Q1 2007)
    - others to follow...

    These are pre-integrated clusters, with all middleware in place, clustered, configured and ready to go. Just add your HTML/XML, scripts, code & database. Useful for Web apps / SaaS apps like SugarCRM, WordPress, etc.

    Grid infrastructure:
    -----------------------
    - shared grid accounts (coming soon)
    - dedicated grids (available today)

    The grid accounts give you the ultimate flexibility in building and controlling your infrastructure. You can deploy and control multiple servers and clusters from the list above, customize each of them as you see fit, build your own infrastructure appliances, integrate applications, etc.

    The main difference between the two services is that the shared grid account will enable you to do this on a larger shared grid, with utility pricing models. The dedicated grid is yours alone, so I suspect you will have to pay based on the number of servers in the grid. On the other side, with a dedicated grid you can be sure those CPUs are there when you need them :-)

    Pre-packaged applications
    ---------------------------------
    - SugarCRM cluster ( up to 10 servers, available today)
    - WordPress cluster (up to 10 servers, available today, thanks to Gabriel Kent from International News Media)
    - others to follow...

    We have packaged these applications to test the demand for such solutions. Tell us what else you need and we will add it to the list if the interest is high enough.

    2. COMMON CHARACTERISTICS

    All of the services described above can be provided in a "mix and match" fashion from the same grid. Because of this, they all have a number of common characteristics that come directly from the way AppLogic works:

    Instant provisioning
    -------------------------
    In AppLogic, it takes a single command to provision any server, infrastructure or application. You choose the app/service, specify IP addresses, root password, and the amount of resources (CPU, memory, disk and bandwidth), and the system manufactures a private copy of the infrastructure for you, complete with all firewalls, load balancers, servers, NAS boxes, gateways, etc.. With a proper portal, you can pick a cluster or other service, pay by credit card and have it running on the grid in 5-10 minutes)

    Universal high availability
    -------------------------------
    When something crashes on an AppLogic grid, whether a physical server or a virtual appliance, the data remains available at all times and the affected appliances are restarted instantly somewhere else on the grid without human intervention, and without having to migrate/move any data. This means that ANY server or cluster running on the grid is highly available (today, there is an up to 5min delay when a physical server dies - much less if appliance of server simply crashes and comes back; with a bit of hardware assist, this will soon be be reduced to less than 30sec).

    Instant scalability
    ----------------------
    ANY AppLogic infrastructure scales in at least two ways: (a) you can change CPU, memory, bandwidth and disk space, and (b) you can add/remove servers from a cluster. Both are done on your exising server/cluster, without requiring any data/code migration. Each can be done in under a minute if the hosting provider implements a web portal that lets people control their grid accounts directly rather than through a ticketing system.

    Private software/configuration
    --------------------------------------
    One thing to keep in mind is that in AppLogic, only the hardware is shared - while you are running on a grid, you always get your own private firewalls, load balancers, web servers, app servers, mail servers, database servers, etc. They are tuned and configured as you want them. This means that there is simply no way for someone else's load to affect your operation. Also, there is no way for the host to oversell - AppLogic enforces resource quotas rigorously.

    Easy migration
    ------------------
    Say you need to add three more servers to your Cpanel cluster, and the grid on which you are running is full. Normally, the hosting provider will simply add several servers to the grid, and you will have the resources you need. But what if the grid is completely maxed out, and adding servers is not an option? In AppLogic, you can migrate a whole cluster from one grid to another with a single command. Depending on the amounts of data stored on the cluster, this will take from several minutes to tens of minutes. In less than an hour, you will be running on a brand new grid. Or in a different datacenter...

    Global deployment and business continuity
    ------------------------------------------------------
    You can set up your application once, and then deploy it to several datacenters with a single command. This saves huge amounts of work when you have customers worldwide. Also, you can run one copy of the app on the East Coast and another on the West Coast and make the two synchronize their data. For a few hundred extra dollars a month, you can build a global load-balancing and/or failover into your service.

    Utility pricing
    ----------------
    AppLogic has a metering system which can tell the hosting provider who consumed how much resources and when. This is important because it allows the hosting provider to let customers adjust their resources directly - you can boost up CPU/memory/storage/bandwidth when you need them, and cut down when you don't.

    While utility pricing is still very new, many people hope that it will finally provide a viable alternative to the overselling which is so rampant in the hosting industry. With each endividual customer able to scale up/down their guaranteed resource quotas explicitly in order to serve high loads and save money in times of lower loads, the hosting providers will be able to keep the utilization high by balancing load between different customers who have different peak times, while providing decent SLA to each customer.

    I hope this post was helpful to you. If you have any questions, feel free to post them here.

    Vlad
      0 Not allowed!

  6. #56
    Join Date
    Mar 2002
    Posts
    328
    Vlad

    I am a software developer and trying to understnad Applogic. In simple terms, you create a VM and all our applications run on it, for storage it use a central storage system such as SAN. In the event the hardware failes the VM is moved to a new system. Am i correct on this ?

    How do you handel when we want to have multiple VM for the same web site. Right now we got a LB in front and multiple servers talking to a single data base ? Can this type of app moved to a grid based system like yours ?

    Thanks

    Jay
    CTO, objectcube.com
      0 Not allowed!

  7. #57
    Join Date
    Dec 2004
    Location
    New York, NY
    Posts
    10,710
    Vlad,

    Thanks for the extremely informative post!
    MediaLayer, LLC - www.medialayer.com Learn how we can make your website load faster, translating to better conversion rates for your business!
    The pioneers of optimized web hosting, featuring LiteSpeed Web Server & SSD Storage - Celebrating 10 Years in Business
      0 Not allowed!

  8. #58
    Join Date
    Dec 2005
    Posts
    550
    a prepackaged app of PHP LAMP + APC + Memcached would be nice
      0 Not allowed!

  9. #59
    Join Date
    Feb 2004
    Location
    Atlanta, GA
    Posts
    5,662
    Quote Originally Posted by VladMiloushev
    These are completely integrated clusters that are ready to run and allow you to manage up to 16-32 servers in a fully transparent & load-balanced fashion (like having a single huge server with up to 64 CPU, 128 MB RAM and 10TB of disk space)


    Sorry Vlad, just had to get ya on that typo 128mb holy cow thats alot of ram
      0 Not allowed!

  10. #60
    Join Date
    Nov 2006
    Location
    Southern California
    Posts
    173
    Quote Originally Posted by jay23
    Vlad

    I am a software developer and trying to understnad Applogic. In simple terms, you create a VM and all our applications run on it, for storage it use a central storage system such as SAN. In the event the hardware failes the VM is moved to a new system. Am i correct on this ?

    How do you handel when we want to have multiple VM for the same web site. Right now we got a LB in front and multiple servers talking to a single data base ? Can this type of app moved to a grid based system like yours ?

    Thanks

    Jay
    CTO, objectcube.com
    Jay, AppLogic is designed to run distributed web applications, primarily transactional and I/O intensive apps. The underlying hardware consists of commodity servers, each with directly attached storage, connected via a backbone Gigabit switch. There is no need for propriatary hardware of any kind - no firewalls, no load balancers, no SAN, no NAS - just commodity servers.

    A typical AppLogic application consists of multiple virtual appliances (e.g. firewall, load balancer, multiple web servers, a couple of database servers, several NAS boxes, mail server, etc.) Each appliance is fully integrated in software, and runs in a separate VM. All storage and network interactions are also completely virtualized. The application consists of an infrastructure assembly and one or more application volumes on which you store the code, database and content that are specific to the app.

    AppLogic packages the whole app into a single entity which you can instantiate (provision), start/stop/resize/copy/ etc. You can change dynamically the resource allocations (CPU, memory, bandwidth, storage) of the whole app, or of individual component. You can also start & stop a component, and ssh into it at any time. You can migrate the whole app from one grid to another, or from one datacenter to another, with a single command.

    I looked on your web site, and yes, apps like yours are our bread and butter.

    Take a look at the AppLogic demo: http://ns1.netclime.net/online-demo.html. You will see exactly how it works.
      0 Not allowed!

  11. #61
    Join Date
    Nov 2006
    Location
    Southern California
    Posts
    173
    Quote Originally Posted by WireSix
    Sorry Vlad, just had to get ya on that typo 128mb holy cow thats alot of ram
    Well, you got me here Still, my 128 Megabytes (MB) sure beat your 128 Megabits (mb)

    Vlad
      0 Not allowed!

  12. #62
    Join Date
    Nov 2006
    Location
    Southern California
    Posts
    173
    Quote Originally Posted by grabmail
    a prepackaged app of PHP LAMP + APC + Memcached would be nice
    If you want a scalable app, you can start from our standard LAMP cluster (it is PHP- based) and add APC & Memcached yourself. Contact Jeremy at Grid Layer, or Jessie at 3Tera if you want to see the LAMP cluster yourself. Jeremy can also get you on their eval grid. Neither of these would cost you a dime unless you decide to deploy your app on the grid.

    Vlad
      0 Not allowed!

  13. #63
    Join Date
    Feb 2004
    Location
    Atlanta, GA
    Posts
    5,662
    Quote Originally Posted by VladMiloushev
    Well, you got me here Still, my 128 Megabytes (MB) sure beat your 128 Megabits (mb)

    Vlad

    nice catch
      0 Not allowed!

  14. #64
    ok, so can I close all this servers we running in 4 data centers and put it in single gridlayer account?
    It will be hosting-heaven to get all this thousands of domains run under single DNS and single WHM access and just ask TL to add more CPU, Space and RAM!!?

    is that possible? and just have single apache setup, single cpanel setting, single mod_security config, .... that take effect on allllll accounts!? .. single skin license, ....

    woow .. is that really possible with that system??

    I have asked LT this questions but no reply yet, may be VladMiloushev give me an answer
      0 Not allowed!

  15. #65
    Join Date
    Nov 2006
    Location
    Southern California
    Posts
    173
    Quote Originally Posted by 2Mhost
    ok, so can I close all this servers we running in 4 data centers and put it in single gridlayer account?
    It will be hosting-heaven to get all this thousands of domains run under single DNS and single WHM access and just ask TL to add more CPU, Space and RAM!!?

    is that possible? and just have single apache setup, single cpanel setting, single mod_security config, .... that take effect on allllll accounts!? .. single skin license, ....

    woow .. is that really possible with that system??

    I have asked LT this questions but no reply yet, may be VladMiloushev give me an answer
    2Mhost, I don't know how many servers you are running today, so I can't tell if you can replace ALL of them with a single cPanel cluster. The cluster version which is in Beta now is configured as follows: it has a firewall, a squid-based cache for static content, a load balancer, 8 Apache servers, a separate MySQL server and a shared NAS storage box.

    All of this is managed from a single instance of cPanel running in a separate virtual appliance. The 8 Apaches work as a single, scalable, load-balanced web server (single config file, single log file). The NAS box combines up to 2TB of storage into a single file system on which the customer accounts are placed.

    This setup scales from approx 2 servers to about 14-15 servers by simply adjusting CPU/RAM/storage and turning spare Apaches on. The accounts are load-balanced across the 8 Apaches, so it is pretty hard to get it overloaded. And, it is all highly available, so no more server rebuilds in the middle of the night :-)

    If this proves popular enough, there is a design to scale it further to about 64 servers. This will require us to also cluster the MySQL server and the NAS box for performance, so it involves a bit more work.
      0 Not allowed!

  16. #66
    Surely I'm going to order one grid, to put the accounts of 2 servers on in, what it's the difference doing it now and the BETA you are telling? thanks.
    Honesting.es honest european provider.
      0 Not allowed!

  17. #67
    Join Date
    Nov 2006
    Posts
    64
    Could you perhaps provide some technical information on this? What is the overhead? Do you use heartbeats to be able to automatically restart crashed nodes? Do "appliances" like firewalls, load balancers, etc really exist as hardware devices or are these simply VPS's tuned to handle these tasks?

    You suggest 30 second - 5 minute transitions on commodity hardware without a SAN. Is this the individual VPS boot time? Are you able to move a VPS onto a different node seemlessly without shutdown/downtime as VMWare can? What happens if my VPS uses 20gb of disk and it needs to migrate from node A to node B without some sort of shared storage?

    Also, can you explain how this can work for applications that don't scale well? My databases could scale to master-master, but beyond that it would require some major redesign. What if I need more memory allocated to a VM than the host node can support? Same question for I/O bandwidth and processor utilization.

    Finally, how would a public-facing load balancer device function? Would this be a VPS on a single node which would direct the traffic, or could the load balancer exist on more than 1 node at a time? What happens when my traffic exceeds the capability of the "load balancer" moving packets?

    Thanks
      0 Not allowed!

  18. #68
    Join Date
    Nov 2006
    Posts
    64
    Oh, and how stable are the host nodes generally? I've found Xen to be pretty flakey, so I find it hard to believe crashes aren't going to occur...often...
      0 Not allowed!

  19. #69
    Join Date
    Nov 2006
    Location
    Southern California
    Posts
    173
    Quote Originally Posted by pueblosnet
    Surely I'm going to order one grid, to put the accounts of 2 servers on in, what it's the difference doing it now and the BETA you are telling? thanks.
    I was referring to the Beta-testing of the CPanel cluster offering. We are testing it with 3 customers now and expect to roll it out in the next few weeks. It will be available on the LT grid the same way separate grid server packages are available today.

    If you are interested, my suggestion would be to get yourself a couple of grid servers now so that you can play with them and get familiar with the system. When the CPanel cluster is ready for production, I am sure that LT will not mind upgrading you. Ask Jeremy to be completely sure.

    If you plan to order a dedicated grid, go ahead - you should get the CPanel cluster upgrade automatically when it becomes available.

    Vlad
      0 Not allowed!

  20. #70
    Join Date
    Nov 2006
    Location
    Southern California
    Posts
    173
    Quote Originally Posted by flaggg
    Oh, and how stable are the host nodes generally? I've found Xen to be pretty flakey, so I find it hard to believe crashes aren't going to occur...often...
    Flaggg, we are not offering Xen, we are offering AppLogic. While AppLogic uses Xen internally, this is a version we maintain. We have grids today that have been running for 40-50 days under load without a server going down. When it goes down, it comes back in 5min without losing data.
      0 Not allowed!

  21. #71
    Join Date
    Mar 2002
    Posts
    328
    Quote Originally Posted by VladMiloushev
    Jay, AppLogic is designed to run distributed web applications, primarily transactional and I/O intensive apps. The underlying hardware consists of commodity servers, each with directly attached storage, connected via a backbone Gigabit switch. There is no need for propriatary hardware of any kind - no firewalls, no load balancers, no SAN, no NAS - just commodity servers.

    A typical AppLogic application consists of multiple virtual appliances (e.g. firewall, load balancer, multiple web servers, a couple of database servers, several NAS boxes, mail server, etc.) Each appliance is fully integrated in software, and runs in a separate VM. All storage and network interactions are also completely virtualized. The application consists of an infrastructure assembly and one or more application volumes on which you store the code, database and content that are specific to the app.

    AppLogic packages the whole app into a single entity which you can instantiate (provision), start/stop/resize/copy/ etc. You can change dynamically the resource allocations (CPU, memory, bandwidth, storage) of the whole app, or of individual component. You can also start & stop a component, and ssh into it at any time. You can migrate the whole app from one grid to another, or from one datacenter to another, with a single command.

    I looked on your web site, and yes, apps like yours are our bread and butter.

    Take a look at the AppLogic demo: http://ns1.netclime.net/online-demo.html. You will see exactly how it works.
    Vlad

    Thanks for this info. I spent time on your web site also. So now I understand that each app (FW, Web, SQL) are run on their own VM. Some question

    1. If there is no SAN, how are we managing HA
    2. Lets say I got a LB App, what happens if that goes down, do the Virtual IP address of that App is taken over by another LB VM
    3. Also on your web site I could not find a list of apps supported by your VM's. Is their a list some where ?

    Thanks

    Jay
      0 Not allowed!

  22. #72
    Join Date
    Nov 2006
    Location
    Southern California
    Posts
    173
    Quote Originally Posted by flaggg
    Oh, and how stable are the host nodes generally? I've found Xen to be pretty flakey, so I find it hard to believe crashes aren't going to occur...often...
    And, let's make one thing clear. There are exactly two ways to build a reliable system: (a) make EVERY element of the system 99.999% reliable; (b) make the system as whole 99.999% reliable.

    The first approach results in what we call "enteprise-class" hardware and software. This is why an Oracle database license is $40,000 per CPU, and the HP 4-way server it runs on is another $50,000.

    The second approach results in a grid. This is what Google does. It is not that their servers don't fail - they fail all the time. It is that those failures are handled by the grid transparently, so the users are not affected.

    Say you have a Dell server with an MTBF of 1 year (it will fail on average once per year). What happens if you put 100 of them in a single system? The MTBF goes down to about 80 hours, which means a server will fail every 3 days or so. 1,000 servers in a single grid, and MTBF is now 7 hours - a server fails every day. The only way to deal with this is by building the software that controls the system so that hardware failures become transparent.

    Bonus: once you can handle hardware failures transparently, software crashes are handled the same way :-)

    Or, you can pay for an Oracle license on an HP enteprise server. It's your money.

    Vlad
      0 Not allowed!

  23. #73
    Join Date
    Nov 2006
    Location
    Southern California
    Posts
    173
    Quote Originally Posted by jay23
    1. If there is no SAN, how are we managing HA
    You don't need a SAN to manage HA. AppLogic creates a redundant, mirrored storage array from the disks attached to the servers.

    Quote Originally Posted by jay23
    2. Lets say I got a LB App, what happens if that goes down, do the Virtual IP address of that App is taken over by another LB VM
    I don't understand the question. The load balancer is part of the application...

    Quote Originally Posted by jay23
    3. Also on your web site I could not find a list of apps supported by your VM's. Is their a list some where ?
    Anything that runs on CentOS or Fedora with 2.6 kernel right out of the box. If you need another distro, you can put it into a virtual appliance yourself as long as it supports 2.6 kernel.
      0 Not allowed!

  24. #74
    Join Date
    Nov 2006
    Location
    Southern California
    Posts
    173
    Quote Originally Posted by flaggg
    Could you perhaps provide some technical information on this? What is the overhead? Do you use heartbeats to be able to automatically restart crashed nodes? Do "appliances" like firewalls, load balancers, etc really exist as hardware devices or are these simply VPS's tuned to handle these tasks?

    You suggest 30 second - 5 minute transitions on commodity hardware without a SAN. Is this the individual VPS boot time? Are you able to move a VPS onto a different node seemlessly without shutdown/downtime as VMWare can? What happens if my VPS uses 20gb of disk and it needs to migrate from node A to node B without some sort of shared storage?

    Also, can you explain how this can work for applications that don't scale well? My databases could scale to master-master, but beyond that it would require some major redesign. What if I need more memory allocated to a VM than the host node can support? Same question for I/O bandwidth and processor utilization.

    Finally, how would a public-facing load balancer device function? Would this be a VPS on a single node which would direct the traffic, or could the load balancer exist on more than 1 node at a time? What happens when my traffic exceeds the capability of the "load balancer" moving packets?

    Thanks
    Flaggg, these are all great questions. The best way to answer them is to show you the system and discuss how it works in the context of a real grid running a real app.

    GridLayer and 3Tera are setting up a WebEx demo for sometime this week. If you can attend, we can answer all of your questions there. If you can't wait, or would like to discuss in a more private setting, send email to info@3tera.com and request a demo. Be sure to mention my name in the message, as we are almost overwhelmed by the interest we are getting the last couple of weeks...
      0 Not allowed!

  25. #75
    Join Date
    Mar 2002
    Posts
    328
    [QUOTE=VladMiloushev]You don't need a SAN to manage HA. AppLogic creates a redundant, mirrored storage array from the disks attached to the servers.

    I don't understand the question. The load balancer is part of the application...

    QUOTE]

    Thanks for this info. So you are replicating in real time 2 servers storage (dont you think thats going to make things slow compared to a SAN).

    As for the LB question, I though LB runs on its own VM so the quesiton is what happens that vm goes down ?


    Jay
      0 Not allowed!

Page 3 of 17 FirstFirst 12345613 ... LastLast

Posting Permissions

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