Page 1 of 17 123411 ... LastLast
Results 1 to 25 of 416
  1. #1
    Join Date
    Jul 2005
    Posts
    63

    thegridlayer.com

    I seeing the new Layeredtech service called "the grid layer" , is the hosting market switch to these kind of solutions?
    Last edited by domotre; 11-08-2006 at 10:28 AM.
      0 Not allowed!

  2. #2
    Join Date
    Jul 2006
    Location
    Detroit, MI
    Posts
    1,962
    Quote Originally Posted by domotre
    I seeing the new Layeredtech service called "the grid layer" , is the hosting market swicht to this kind of solutions?
    I have a feeling this is a gimmick - just like the MT "grid"
      0 Not allowed!

  3. #3
    Join Date
    Jul 2006
    Location
    Detroit, MI
    Posts
    1,962
    I just did some reading on AppLogic - the application they are using for load balancing.

    I think this solution is better then MTs, as they really just have a SAN. At least TGL will distribute the load, however, it is really co-location, not a grid.

    A grid divides a single problem up into pieces in order to develop the answer faster.
    • A cluster is not a grid.
    • co-located app servers are not grids.
    • Grids use MPI.

      It's a shame these companies are turning their marketing campaigns into jokes....
      0 Not allowed!

  4. #4
    Well, I just test the gridlayer from LT, what I can say is that:

    • Looks like a VPS, but with the ability to increase RAM, maybe Bandwidth, etc but without the need to order a physical server or move accounts, etc. You can keep the IPs addresses, and add more ram, more space.
    • I test it, the RAM control feature works great, I have 0.0 load average.
    • No need for raid, because the grid stores your data on multiple machines, so no downtime for you and your customers.
    • I’m not sure if your main account goes down (a), and a server (b) comes in replacement, server (b) will have the ips and the same configuration of server (a)
    • I haven’t seen the web based grid application, but you can add appliances on the grid and just pay for the services that you need. So I guess what you can get 20 packages, and then add appliances that you might need for a small fee. (Just pay for what you use).
    • I see this as a main advantage for the following reasons:
    If you want to provide your customers with 100% uptime without worrying of raid failures.
    I have H-sphere control panel, and doing a cluster is expensive, because I need to buy each server for each process plus raid configuration. It is more expensive to maintain also a cluster. With this option I can build my cluster with a little money.

    People may ask, why us Clustering so important? Well clustering offers you greater uptime, and less data loss, plus you have more control over what your customers is doing. With a cluster you can add a www server without having any downtime, or replace ram with any downtime.

    I am not sure if these grids include load balancing, and firewall services. Plus, since the grid is like a VPS, what happens if the server needs a hard reboot? Well guess what? Since the data is mirrored, there will be no problem with this, if the main server goes down, there will be another server to take care of you data, so the old limitations of the VPS doesn’t apply to this new technology.
      0 Not allowed!

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

    Is GridLayer a true grid?

    Quote Originally Posted by utropicmedia-karl
    I just did some reading on AppLogic - the application they are using for load balancing.

    I think this solution is better then MTs, as they really just have a SAN. At least TGL will distribute the load, however, it is really co-location, not a grid.

    A grid divides a single problem up into pieces in order to develop the answer faster.
    • A cluster is not a grid.
    • co-located app servers are not grids.
    • Grids use MPI.

      It's a shame these companies are turning their marketing campaigns into jokes....
    Karl,

    I am the architect of AppLogic. I think you are mixing grids with computational clusters, also called "Beowulf" clusters. Beowulf clusters use MPI to divide number-crunching problems into pieces to develop the answer faster. They are NOT grids. The Top500 site lists the 500 largest supercomputers in the world; more than half are clusters, and they call them so.

    I know Carl Kesselman and Ian Foster, the founders of the Grid Alliance and the authors of the Globus toolkit, since 2001. I have shown them AppLogic several times, and they both agree that it is a new type of a grid system - one that runs web applications.

    I also know the guy who invented the very term "cluster". His name is Richie Lary; he is one of the architects of the original VAX Clusters product and a chief storage architect of Compaq (now HP). Richie advises our team on the storage and high availability aspects of AppLogic. He will be the first to tell you that AppLogic is not a cluster.

    The commonly accepted definition of a Grid is provided by Ian Forster in his article "What is the Grid? A Three Point Checklist". The three points of this checklist are:
    • Computing resources are not administered centrally.
    • Open standards are used.
    • Non-trivial quality of service is achieved.
    The GridLayer grid meets all three characteristics:

    - Each customer can administer their portion of the grid
    - The grid is open and compatible with Linux and most existing software
    - The resulting scalability, high availability and ease of provisioning are non-trivial

    See also Wikipedia, which states: "
    Grid computing is optimized for workloads which consist of many independent jobs or packets of work, which do not have to share data between the jobs during the computation process. Grids serve to manage the allocation of jobs to computers which will perform the work independently of the rest of the grid cluster. Resources such as storage may be shared by all the nodes, but intermediate results of one job do not affect other jobs in progress on other nodes of the grid."

    AppLogic and the GridLayer grid clearly fit this definition as well. This is exactly what happens when you run multiple firewalls, load balancers, web servers, database servers, shared hosting apps, etc. on the same grid.

    So, we are not talking about a marketing campaign here.
      0 Not allowed!

  6. #6
    Join Date
    Jul 2006
    Location
    Detroit, MI
    Posts
    1,962
    Vlad,

    Thanks for taking the time to come visit - my background is Computer Science and theoretical math so I enjoy speaking with people about it. Just as a background, I did my thesis on combinatorics heuristics so am familar with the math behind graph design.

    While we know industry tends to take simple academic problems and create

    Karl,

    I am the architect of AppLogic. I think you are mixing grids with computational clusters, also called "Beowulf" clusters. Beowulf clusters use MPI to divide number-crunching problems into pieces to develop the answer faster. They are NOT grids. The Top500 site lists the 500 largest supercomputers in the world; more than half are clusters, and they call them so.
    Agreed, but when they run benchmarks/applications on those machines, they do not serve webpages - they perform discrete computational tasks. Additionally, grid's like Sun's do the exact same thing, in the exact same mannor, and are recognized as grids.
    I know Carl Kesselman and Ian Foster, the founders of the Grid Alliance and the authors of the Globus toolkit, since 2001. I have shown them AppLogic several times, and they both agree that it is a new type of a grid system - one that runs web applications.
    I appreciate you know them. Though do you realize there are several "Grid Alliances" out there? They are but one of the smaller ones. Appreantly, all the big companies are members of this one.


    The three points of this checklist are:

    • Computing resources are not administered centrally.
    • Open standards are used.
    • Non-trivial quality of service is achieved.
    Anyone can write an article and create a definition; I would disagree with that list.
    Open standards would not allow for a particular vendor to have a competitve advantage, and though it may be nice in a perfect world, there are many examples of closed technology outperfomring open standards. (SGI?)

    See also Wikipedia, which states: "
    Grid computing is optimized for workloads which consist of many independent jobs or packets of work, which do not have to share data between the jobs during the computation process. Grids serve to manage the allocation of jobs to computers which will perform the work independently of the rest of the grid cluster. Resources such as storage may be shared by all the nodes, but intermediate results of one job do not affect other jobs in progress on other nodes of the grid."
    How exactly do the SQL servers fit into the model? according to the above definition, the each discrete node would have to consist of it's own SQL instance and dataset, to be considered a grid - the only way to make each node and request independent. From your collateral, AppLogic does not provide that, or am I missing something?

    AppLogic and the GridLayer grid clearly fit this definition as well. This is exactly what happens when you run multiple firewalls, load balancers, web servers, database servers, shared hosting apps, etc. on the same grid.
    The only way AppLogic would meet that is if:
    • There is a basic unit of work defined in the system
    • Each node has full, independent resources, local to the node, to perform a unit of work.
    Kind Regards,
      0 Not allowed!

  7. #7
    Join Date
    Jul 2006
    Location
    Detroit, MI
    Posts
    1,962
    Bah - vB timed out on me so the last post is "best-effort"
      0 Not allowed!

  8. #8
    Join Date
    Nov 2006
    Location
    Southern California
    Posts
    173

    Grids, grids and grids

    Karl,

    Ah, now we have the start of a real discussion. You are absolutely right that the term "grid" is increasingly being defined by the industry rather than by academia. So I agree with you that the old academic definitions are no longer valid (and I have told Ian so).

    Within the industry, the word "grid" has increasingly come to mean "an aggregation of commodity computers that function as a single, easily scalable system". For example:

    - Oracle uses it in this sense in their 10g architecture.
    - IBM speaks about a grid architecture for the WebSphere application server.
    - Data Synapse, ActiveGrid and Appistry use it for scalable application servers
    - Exagrid uses the term for a scalable storage system.
    - Many people have identified the Google architecture as a grid.

    3Tera uses the word grid for a scalable system that runs web applications, both transactional and I/O-intensive.

    The common theme about the modern usage of the word "grid" seems to be:

    - built from commodity servers and networks
    - shares storage/processing/IO resources among multiple parallel tasks
    - scales easily to hundreds and thousands of servers
    - handles hardware failures transparently

    So, in this context, a grid that runs web services and web applications is not even new. The problem with all of the existing grid architectures is that they require you to rewrite or significantly modify your application for the grid. This is very expensive and risky, and so there aren't many takers.

    AppLogic is different in that it allows you to deploy, run and scale existing applications on a grid without modifications. This works for anything from simple virtual servers all the way to complete N-tier applications.

    So, what we have here is a grid that runs and scales the majority of commercially interesting applications right out of the box, handles hardware failures automatically and without losing data, and is easy to manage.

    Looks like something of value to me :-)
      0 Not allowed!

  9. #9
    Join Date
    Nov 2006
    Location
    Southern California
    Posts
    173
    Quote Originally Posted by Latin_Carrier
    Well, I just test the gridlayer from LT, what I can say is that:

    • Looks like a VPS, but with the ability to increase RAM, maybe Bandwidth, etc but without the need to order a physical server or move accounts, etc. You can keep the IPs addresses, and add more ram, more space.
    • I test it, the RAM control feature works great, I have 0.0 load average.
    • No need for raid, because the grid stores your data on multiple machines, so no downtime for you and your customers.
    • I’m not sure if your main account goes down (a), and a server (b) comes in replacement, server (b) will have the ips and the same configuration of server (a)
    • I haven’t seen the web based grid application, but you can add appliances on the grid and just pay for the services that you need. So I guess what you can get 20 packages, and then add appliances that you might need for a small fee. (Just pay for what you use).
    • I see this as a main advantage for the following reasons:
    If you want to provide your customers with 100% uptime without worrying of raid failures.
    I have H-sphere control panel, and doing a cluster is expensive, because I need to buy each server for each process plus raid configuration. It is more expensive to maintain also a cluster. With this option I can build my cluster with a little money.

    People may ask, why us Clustering so important? Well clustering offers you greater uptime, and less data loss, plus you have more control over what your customers is doing. With a cluster you can add a www server without having any downtime, or replace ram with any downtime.

    I am not sure if these grids include load balancing, and firewall services. Plus, since the grid is like a VPS, what happens if the server needs a hard reboot? Well guess what? Since the data is mirrored, there will be no problem with this, if the main server goes down, there will be another server to take care of you data, so the old limitations of the VPS doesn’t apply to this new technology.
    Unlike VPS, each grid server has it's own copy of Linux, so you can configure and install software whichever way you want.

    You can resize RAM, CPU and disk size dynamically, w/o migrating or losing data.

    All of your configuration, including IPs, etc. is independent of any specific physical server, and will survive server crash and/or replacement. In fact, TGL can migrate you from one grid to another on the fly and you will not notice it.
      0 Not allowed!

  10. #10
    Join Date
    Nov 2004
    Posts
    366
    It's interesting, nowhere on the site does it give you any indication of what type of CPU performance you can expect.

    I've actually very confused by the pricing, it seems to have the pricing models of normal dedicated servers, but in the features it says that cost is based on usage.
    " * Predictable cost, based on usage"
    but, then is says:
    "Unlimited CPU and server use which means lower cost and greater ROI."

    From the descriptions on the site, it sounds like it would be priced like amazon's EC2, which is based on bandwidth and cpu time, (priced down to the hour of cpu and gig of bandwidth) but, with TGL, it seems I can only purchase "packages"

    It sounds like this is not for me (who runs one large site on 10 machines, which aren't busy most of the time, but need the capacity for busy times of the day) and is more for a hosting company that runs many sites that needs to dynamically partition them as sites grow. Please correct me if I'm wrong.
      0 Not allowed!

  11. #11
    Join Date
    Jul 2003
    Location
    Texas
    Posts
    787
    Hello for anyone interested we will be having a WebEX demo this evening of Gridlayer and its features. Please feel free to join as at the details listed below. There will be a limit of 20 people and we will likely hold another one next week.

    http://3tera.webex.com
    Date: Friday, November 10, 2006
    Starting time: 5:00 pm, Pacific Standard Time (GMT -08:00, San Francisco)
    Meeting number: 804 158 857
    Meeting password: None

    Teleconference
    (800) 977-8002
    Pin # 890517

    We will be able to demo the features of TGL and take questions.

    Thanks,

    Jeremy
      0 Not allowed!

  12. #12
    Join Date
    Nov 2004
    Posts
    366
    Oh, I'd like to go, but bad timing for me.

    Can someone fill us in how it went?

    I'm still curious about the CPU info on these packages.
      0 Not allowed!

  13. #13
    Join Date
    Jun 2004
    Location
    Bay Area
    Posts
    1,320
    This all sounds very interesting, but for me, the average webhoster, very confusing.

    Lets say I currently have a Dual Xeon 2.4Ghz / 2GB / 2*73GB server. CPU load is arround 0.2 at night and 2.0 at peak hours.

    a. What package would I be needing?
    b. What would the total monthly costs be? There are some packages listed, but not 'unit prices'..prices per Mhz, MB ram or HDD. And if I understand correctly, that is the power of this system: Being able to fit your requirements exactly?
    c. How flexible is this? Would I only pay for the exact number of CPU-units I use? In other words: Would I pay less at night than I would do during peak hours?
    Would I be able to 'peak' CPU resources if other people in the grid do not use any? Would it cost extra?
    d. Does the system emulate a physical machine, like a VPS? Would I get access to a server that looks just like a normal machine, to which I have root access, and I can run my own scripts and tools? Would I be able to compile and use my own kernel?

    In a couple of months I need to move a popular website to a new host, and I must say that this sounds extremely well-suited for me. But I do need some more clarity in how it works. Maybe even a demo login, or simple demo movie of the interface or something. Right now it all is very vague
      0 Not allowed!

  14. #14
    Join Date
    Feb 2004
    Location
    Atlanta, GA
    Posts
    5,662
    thanks for the con call Jeremy.

    Very interesting...
      0 Not allowed!

  15. #15
    Join Date
    Dec 2005
    Posts
    550
    this is damn interesting. Just hope it is as reliable as Amazon EC instead of the MT Grid
      0 Not allowed!

  16. #16
    Join Date
    Jan 2003
    Location
    Texas, where else?
    Posts
    1,571

    Cool

    Quote Originally Posted by VladMiloushev
    Unlike VPS, each grid server has it's own copy of Linux, so you can configure and install software whichever way you want.
    As for all the stuff about grids it's beyond me but I keep a "play" VPS with a company I have a long history with just to try stuff-break things etc. without utilizing an entire server and I was able to use the *nix flavor of my choice as well as all other software versions, so maybe some people do VPS different than others
    New Idea Hosting NO Overselling-Business-Grade, Shared Only! New-In House Design Team.
    High Speed & Uptime; , DIY Pro-Site Builder-Daily Backups-Custom Plans, All Dual Xeon Quad Intel servers w/ ECC DDR3 RAM SCSI RAID minimums.
    We Concentrate on Shared Hosting ...doing one thing and doing it VERY well
      0 Not allowed!

  17. #17
    Join Date
    Nov 2004
    Posts
    366
    Oh, I found a demo of their virtualization interface

    http://www.thegridlayer.com/images/f...logic-demo.swf

    Looks like you can create some pretty complex systems, but still no word on cpu or performance of the whole thing. To me, it's pointless to have your whole system virtualized and redundant if it's too slow.
      0 Not allowed!

  18. #18
    Join Date
    Nov 2006
    Location
    Southern California
    Posts
    173
    Quote Originally Posted by digdugdog
    Oh, I found a demo of their virtualization interface

    http://www.thegridlayer.com/images/f...logic-demo.swf

    Looks like you can create some pretty complex systems, but still no word on cpu or performance of the whole thing. To me, it's pointless to have your whole system virtualized and redundant if it's too slow.
    I don't think making performance claims serves any purpose. You should check the performance yourself - LT is giving out free eval grid servers.
      0 Not allowed!

  19. #19
    Join Date
    Jun 2004
    Location
    Bay Area
    Posts
    1,320
    Quote Originally Posted by VladMiloushev
    LT is giving out free eval grid servers.
    Are you sure about that? I mailed for some extra info and a possible demo about two days ago - no word back yet.
      0 Not allowed!

  20. #20
    Join Date
    Nov 2004
    Posts
    366
    I emailed a few days ago, no reply.

    I don't think making performance claims serves any purpose.
    Performance matters a lot for some people. There is a reason people pay more for faster machines.

    Amazon EC2 at least gives a comparable measure for their virtual machines:

    "Each instance predictably provides the equivalent of a system with a 1.7Ghz x86 processor, 1.75GB of RAM, 160GB of local disk, and 250Mb/s of network bandwidth."
      0 Not allowed!

  21. #21
    Join Date
    Jul 2003
    Location
    Texas
    Posts
    787
    Quote Originally Posted by digdugdog
    I emailed a few days ago, no reply.



    Performance matters a lot for some people. There is a reason people pay more for faster machines.

    Amazon EC2 at least gives a comparable measure for their virtual machines:

    "Each instance predictably provides the equivalent of a system with a 1.7Ghz x86 processor, 1.75GB of RAM, 160GB of local disk, and 250Mb/s of network bandwidth."

    Hello,

    Our current TGL-2400 can provide a equalivent min 2.4ghz CPU to 2.8ghz range of CPU per node with 128MB to 2GB (1.90GB) of RAM per node available (The CPU per node can be cut down to 1/10 per appliance). There is also the ability to have upto a 240GB high availability storage slice on each node. The next TGL grid we release will be able to double this power per host and it should be released in the next 30 days and clients from the first grid will be able to migrate their appliances between the various versions of the TGL as upgrade paths.

    Any appliance can take the full resources of one of the 32 nodes or can be setup to use more then one if properly designed via the editor or other tools. We will be holding another WebEX next friday and I highly suggest you watch it as it will answer many questions and show you the full power of the TGL.

    I am sorry if we did not get back to you quickly via the email request. We have been swamped with people requesting information and demos of the stand alone GSC appliances.

    Thanks,

    Jeremy
    Last edited by Cirrostratus; 11-12-2006 at 02:30 PM.
      0 Not allowed!

  22. #22
    Join Date
    Jul 2003
    Location
    Texas
    Posts
    787
    Some other info. Each host on the TGL is also linked to our front end network with a dedicated 100mb/s and also has a second 1000mb/s link for host to host communication on the TGL. All appliances running on the grid have a default link of 100mb/s to the front end network. The bandwidth throughput can also be scaled per appliance.

    Here is the 'grid info' on one of our current 3 grids.

    Grid Name : GridLayer01
    Grid Description : Layered Technologies TGL01
    AppLogic Version : 1.2.3
    Total Servers : 32
    Servers Running : 32
    Servers Enabled : 32
    Applications Running : 16
    Total CPUs : 32.00
    Free CPUs : 23.86
    CPU Load : 1.13
    Allocated CPUs : 8.14
    Total Memory : 60384.00 MB
    Reserved Memory : 687.00 MB
    Free Memory : 42481.00 MB
    Allocated Memory : 17216.00 MB
    Total Bandwidth : 64000.00 Mbps
    Free Bandwidth : 63972.70 Mbps
    Allocated Bandwidth : 27.30 Mbps
    Total Disks : 32
    Total Storage : 7113507.70 MB
    Reserved Storage : 1600.00 MB
    Free Storage : 6273620.52 MB

    Thanks,

    Jeremy
      0 Not allowed!

  23. #23
    Join Date
    Jun 2004
    Location
    Bay Area
    Posts
    1,320
    Thanks Jeremy, that already clears up a lot

    Say I would go for the '60' package, install a webserver and run my site off it. That should run fine with 2.4Ghz / 512MB ram.
    However my site gets slashdotted...it would require a lot of extra CPU and memory resources. How would I proceed, and what costs are involved?
      0 Not allowed!

  24. #24
    Join Date
    Nov 2006
    Location
    Southern California
    Posts
    173

    Performance matters

    Quote Originally Posted by digdugdog
    I emailed a few days ago, no reply.



    Performance matters a lot for some people. There is a reason people pay more for faster machines.

    Amazon EC2 at least gives a comparable measure for their virtual machines:

    "Each instance predictably provides the equivalent of a system with a 1.7Ghz x86 processor, 1.75GB of RAM, 160GB of local disk, and 250Mb/s of network bandwidth."
    Sorry, Digdugdog, I did not express myself clearly. Of course performance matters. What I meant was that performance claims have little value - better to get an eval and test it yourself.

    Vlad
      0 Not allowed!

  25. #25
    Join Date
    Apr 2002
    Location
    Orange County, CA
    Posts
    338
    I think there are two ways to use LayeredTech's Grid system. First is like a traditional VPS where a generic AppLogic "container" runs an OS and a control panel. One can SSH into that container and install / customize software (Apache, MySQL, email MTA etc) within it. In this case the VPS container only runs on one grid server, and its maximum performance is limited by that particular server's resource (CPU, RAM, etc).

    The second (and better) way is to "tear apart" the traditional VPS and put individual web components into different AppLogic containers. For example, there should be multiple Apache servers that run on different grid servers. Another container will run the MySQL server, and yet another run the email MTA. Then the entire system could be scale up or down as needed. This is probably the intended use for the grid, and most suited for large-scale projects.

    For someone who already has 100+ clients on a traditional VPS or dedicated server, switching to the grid using the first way is easy - no different than moving to a new dedicated server or VPS. The second way (in which the VPS is "teared apart") might be more difficult. I imagine with so many clients, some PHP scripts will need to be coaxed into working through multiple servers (MySQL on a different server than Apache, for example).

    I also noticed the small memory allowance given by the GridLayer packages. The 10 and 20 packages only have 256MB RAM. If I split off a minimum Apache-MySQL setup into two Apache containers and two MySQL server containers, then each will have only 64MB of RAM. That is not enough to run an OS plus an application (Apache or MySQL). Perhaps those two packages are meant to run everything within one single container like a traditional VPS.

    Please correct me if I am mistaken.
    Last edited by Yikes2000; 11-13-2006 at 02:19 AM.
      0 Not allowed!

Page 1 of 17 123411 ... 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
  •