Page 1 of 2 12 LastLast
Results 1 to 40 of 53
  1. #1

    Experts: Clustered Servers

    http://www.adaptec.com/worldwide/pro...id_family_tool

    1)I would like to have a "clustered server" , how the server connected to the External RAID storage?
    2)What is HBA? DO we need fibreChannel ?

    3)Do we need to have 2 RAID controller on the External Storage?

    4)Do we need to have a RAID controller for each of the server?

    5)How having 3 servers for this clustered setup, instead having 2 server ?

    6)If would like to get a Dell solution, which server and RAID recommended? Any extra RAID card/fibrechannel we need?

    7)How the server OS will boot from the External Storage, we need to have a boot loader to boot each of the server?

    8)Can we partition the external storeage to 3 partition, and each of the server boot from their own partition.?

    -Thx
    qazakax .at. yahoo.com

    Cheers..We are using LINUX OS.

  2. #2
    What will be the use of this "Cluster"
    Datums Internet Solutions, LLC
    Systems Engineering & Managed Hosting Services
    Complex Hosting Consultants

  3. #3
    Join Date
    Feb 2003
    Location
    Detroit
    Posts
    836
    DO we need fibreChannel
    Probably not. The vast majority of what we see it network attached storage units.

    If you post what your trying to do with the cluster, more people can probably give input. In most of the clusters I see, each unit has it's own boot drives, with some sycronization depending on what it's doing. Usually, there is a massive storage array on a gigabit internal network or something simular. But, there are a million and one ways to build a cluster. How you do it depends on budget and expected results.
    managedway
    WE BUILD CLOUDS

    Cloud Computing | Fiber Optic Internet | Colocation

  4. #4
    there really are not a million ways to build a cluster.

    nas is usually out of the question because of performance and brokenness reasons - linux nfs implementation is *broken* and obviously insufficient if you are trying to run a database over it. if you want an active/active setup *and* your apps support that, you need shared scsi. if you are happy with an active/passive failover cluster, a real-time lock-level data replication solution like nbd will work.

    paul
    * Rusko Enterprises LLC - Upgrade to 100% uptime today!
    * Premium NYC collocation and custom dedicated servers
    call 1-877-MY-RUSKO or paul [at] rusko.us

    dedicated servers, collocation, load balanced and high availability clusters

  5. #5
    You asked a similar question before. And I think I've gave you the answer.
    Now, anybody who needs this solution/answer are welcome to contact me by e-mail or PM.

    Peter Kinev.
    Open Solution, Inc
    http://opensolution-us.com

  6. #6
    Join Date
    Nov 2002
    Posts
    2,780
    I think I actually chat with him on this project before.

    He is looking to mount databases over the network. Perhaps SCSI over LAN/WAN is what he wanted. Anyone have input on this?

  7. #7
    Originally posted by rusko
    if you want an active/active setup *and* your apps support that, you need shared scsi.
    rusko,

    In case of node A failed, can node B access the disk array in node A by shared scsi bus?

  8. #8
    with shared scsi, your shared data is on the shared disk in an external enclosure, so yes. with that said, trust me when i say shared scsi is going to be too expensive to be financially viable if you are a hosting service provider.

    you didn't answer my question though, do you need active/active (impossible with mysql unless you use something like emic) or active/passive failover?

    paul
    * Rusko Enterprises LLC - Upgrade to 100% uptime today!
    * Premium NYC collocation and custom dedicated servers
    call 1-877-MY-RUSKO or paul [at] rusko.us

    dedicated servers, collocation, load balanced and high availability clusters

  9. #9
    Originally posted by Mfjp
    I think I actually chat with him on this project before.

    He is looking to mount databases over the network. Perhaps SCSI over LAN/WAN is what he wanted. Anyone have input on this?
    if you mean iscsi, it's a bit tricky. i would venture a guess that it would be too slow. with a two-node cluster, i would think something like raid 1 over nbd would work (if he had the patches to fix nbd =).

    paul
    * Rusko Enterprises LLC - Upgrade to 100% uptime today!
    * Premium NYC collocation and custom dedicated servers
    call 1-877-MY-RUSKO or paul [at] rusko.us

    dedicated servers, collocation, load balanced and high availability clusters

  10. #10
    Originally posted by rusko
    you didn't answer my question though, do you need active/active (impossible with mysql unless you use something like emic) or active/passive failover?

    paul
    Oh sorry... yes I'm looking at the active/passive model, I can do it with replication software like nbd or Pratima?

    Anything else I need to take care while setting up this cluster?

  11. #11
    BTW can these replication software do 2-way replication between 2 or more nodes?

    for example, a load balanced setup: node A to node B and node B to node A, they replicate each others?

  12. #12
    Originally posted by nowisph
    Oh sorry... yes I'm looking at the active/passive model, I can do it with replication software like nbd or Pratima?
    nbd is broken. not familiar with pratima, but it is asynchronous, so no way you will have consistency using it - it would be the same as using rsync. there is some open source stuff but you would need my custom (proprietary) patches to fix it so it works.lifekeeper is an option, though i don't have experience with it.


    Anything else I need to take care while setting up this cluster?
    yes, you need to do research and understand what you are doing.would you go designing an airplane when your knowledge of physics boiled down to newton's laws of physics? no. why is async not acceptable for this purpose? what is the difference between coherency and consistency? before you are able to answer that, you have no business engineering clusters. i am sorry if this offends you, but i am a firm believer in being brutally honest when it counts.

    paul
    * Rusko Enterprises LLC - Upgrade to 100% uptime today!
    * Premium NYC collocation and custom dedicated servers
    call 1-877-MY-RUSKO or paul [at] rusko.us

    dedicated servers, collocation, load balanced and high availability clusters

  13. #13
    Originally posted by nowisph
    BTW can these replication software do 2-way replication between 2 or more nodes?
    replication is one way. you are probably thinking of multicasting replication. if so, there are ways to do it depending on what you are using for replication itself.


    for example, a load balanced setup: node A to node B and node B to node A, they replicate each others?
    have you not read what i said? you can *not* do load balancing with sync replication, since you do not have write (or even read in some cases) access to the shared resource, which means anything but the primary node can *not* be active.

    paul
    * Rusko Enterprises LLC - Upgrade to 100% uptime today!
    * Premium NYC collocation and custom dedicated servers
    call 1-877-MY-RUSKO or paul [at] rusko.us

    dedicated servers, collocation, load balanced and high availability clusters

  14. #14
    Join Date
    Nov 2002
    Posts
    2,780

  15. #15
    http://www.redhat.com/software/rha/cluster/manager/

    This will be something i m looking at..?

    Yes, obviously, i have to thanks Jay for his idea and opinion.

    The 2 Application server will connect to the External RAID storage using HBA, which i dun now what is HBA, ?
    HBA = FibreChannel or just switch?

    And i have talk to Dell technician, he say Dell PowerVault SCSI only can support to 2 servers, i think i will going to this:

    1x Dell750
    -1GB ram
    -1x 36SCSI
    -4 NIC
    -to do the firewall and balancing

    2x Dell 1650
    -2GB ram
    -1x 36SCSI(just for the boot)
    -RAID controller
    -HBA adapter
    - 4 NIC
    -to do the App server, redundancy and failover,

    -1x Dell PowerVault SCSI
    -4x36GB U320SCSI (for RAID 5)
    -1 HotSpare (Redundancy)



    Dell750 <---- lan-----> dell1750 x 2 <---HBA---> RAID
    ^----------hba-----------^



    -qaz
    sorry for my poor drawing ","

  16. #16
    hba stands for host bus adapter. you don't need to buy redhat's solution, the components are freely available. your config is bunk. and you still don't understand what you are doing. *please* do some research, at least start by reading redhat's whitepaper on their solution. don't have time now, will post a bit later in more detail. rtfm while i'm gone, please!

    paul
    * Rusko Enterprises LLC - Upgrade to 100% uptime today!
    * Premium NYC collocation and custom dedicated servers
    call 1-877-MY-RUSKO or paul [at] rusko.us

    dedicated servers, collocation, load balanced and high availability clusters

  17. #17
    Join Date
    Nov 2002
    Posts
    2,780
    I still think you should run your Database on your Raid Array and connect to it using TCP/IP over the network.

    Not that I'm saying what you're doing isn't good, but the expandaiblity and flexibility is somewhat limited in how you do this.

  18. #18
    Originally posted by Mfjp
    I still think you should run your Database on your Raid Array and connect to it using TCP/IP over the network.

    Not that I'm saying what you're doing isn't good, but the expandaiblity and flexibility is somewhat limited in how you do this.
    jay,

    nfs is too slow and broken for this; trust me, it's been tried =]

    paul
    * Rusko Enterprises LLC - Upgrade to 100% uptime today!
    * Premium NYC collocation and custom dedicated servers
    call 1-877-MY-RUSKO or paul [at] rusko.us

    dedicated servers, collocation, load balanced and high availability clusters

  19. #19
    Join Date
    May 2004
    Posts
    351
    rusko:
    So for an application such as mysql to have an active configuration, would it be required to allow writing to the same file by two servers at the same time?
    Also, when you say application support, would you consider what I mentioned application support or actually doing some sort of balancing where the query is exectued on half of the dataset within on server and the other half on the other server?

  20. #20
    cord,

    you mean active/active. and yes, to have a full active/active config you would need to have full write access at both nodes, which is impossible with mysql since locks are not persistent, even if you did work around all the other issues. you would be looking at some serious corruption because of concurrency issues.

    splitting the dataset in half will not help if it is the same database. think about it, you now need to propagate changes both ways. this is a Very Bad Idea (tm).

    essentially, the best you can do with mysql currently is an active rw/active ro x N configuration. your problem is going to be propagating changes from the master (the rw copy) to the read-only copy as well as directing queries at the appropriate node. this can be done in an application transparent manner (no need to change souce code for your scripts) by proxying and parsing the queries (emic does this, albeit very badly) or by building that into your application (ie scripts in this case).

    in short, it is almost always cheaper and better to throw bigger iron at mysql rather than more iron.

    this and other related clustering questions seem to come up fairly often lately. perhaps i'll put together a few whitepapers as introductory reading if/when i have the time. for now, i am too busy engineering and implementing clusters for our customers =]

    paul
    * Rusko Enterprises LLC - Upgrade to 100% uptime today!
    * Premium NYC collocation and custom dedicated servers
    call 1-877-MY-RUSKO or paul [at] rusko.us

    dedicated servers, collocation, load balanced and high availability clusters

  21. #21
    Join Date
    May 2004
    Posts
    351
    well thanks for the reply rusko. Everything you said seems common sense to me.

    What I mean about queries was read only queries splitting across sections of the dataset. For example, if you do a SELECT blah FROM table WHERE blah=12, would it not benefit if we split the dataset in N over N servers where each server searches its 1/Nth of the database?

    It is interesting that such support for active/active does not exist in mysql. One would think it would be a very natural request made by many people. It really makes me question the use of mysql for high volume/mission critical environments.

  22. #22
    Join Date
    Nov 2003
    Posts
    384
    cord: there is a cluster solution coming up, type in "cluster" at @mysql.com, read what they whitepaper has to say, give these guys a call and be ready to pay some $$ but it'll be worth it.
    You might also get a mysql guru - you can have 100% mysql uptime with ordinary scsi boxes 4GB RAM which are replicating each other (a master + 2 slaves with low_priority_updates takes you very very far and you can "swich on the fly" if the master breaks...)

  23. #23
    Join Date
    May 2004
    Posts
    351
    This seems intersting. I am in no ways looking at clusters for, mostly I am educating myself. It would be such a waste to let free information fly by without taking advantage eh?
    I have made a point of following rusko around to see what he has to say. Dont tell him though

    The problem I foresee with splitting datasets is the following:
    databases are designed such that it is impossible to know in advance what kind of splitting would benefit you most. In other words, depending on the type of query, you have to make differnt divisions. I would assume that one can write a semi-intelligent system to split the dataset on-the-fly, but I guess it is not efficient or else someone would have already done it (that is if it hasnt been done).

  24. #24
    Join Date
    Nov 2002
    Posts
    2,780
    Originally posted by rusko
    jay,

    nfs is too slow and broken for this; trust me, it's been tried =]

    paul
    Yes, i agree with you, that's why i said connect to the mysql host over tcp/ip rather than mounting the sql store using nfs...

  25. #25
    Originally posted by Mfjp
    Yes, i agree with you, that's why i said connect to the mysql host over tcp/ip rather than mounting the sql store using nfs...
    how is this clustering mysql then? you just have a separate mysql box =]

    p
    * Rusko Enterprises LLC - Upgrade to 100% uptime today!
    * Premium NYC collocation and custom dedicated servers
    call 1-877-MY-RUSKO or paul [at] rusko.us

    dedicated servers, collocation, load balanced and high availability clusters

  26. #26
    Originally posted by cord
    well thanks for the reply rusko. Everything you said seems common sense to me.
    that's because it is =]


    What I mean about queries was read only queries splitting across sections of the dataset. For example, if you do a SELECT blah FROM table WHERE blah=12, would it not benefit if we split the dataset in N over N servers where each server searches its 1/Nth of the database?
    no, not if your database is optimized correctly for the queries you are running. for example, in this case, the complexity of the search is not O(n), but rather close to O(1) if you have the table indexed on blah. think of it as a hash table as opposed to a linked list.


    It is interesting that such support for active/active does not exist in mysql. One would think it would be a very natural request made by many people. It really makes me question the use of mysql for high volume/mission critical environments.
    erm. mysql is *not* appropriate for high volume and mission critical use. that doesn't stop people from trying to use it for those purposes though. every single blogger on a shared hosting account considers the availability of his illiterate musings to be mission critical. feh.

    mysql started out as a very simple attempt at an rdbms lookalike. essentially, it was a flat file database with a non ansi sql compliant query parser on top of it. it is still not ACID compliant. it still can not handle concurrency well. essentially, mysql is good for dipping your toes in the database world without getting involved in the complexities of transactions, stored procedures and triggers and for play projects. since it is not encumbered with being an actual rdbms, it's select query performance is quite good. there are a lot of things wrong with mysql before we even come to the clustering bit.

    with that said, there are few rdbms out there that can be clustered. oracle rac is one.

    paul
    * Rusko Enterprises LLC - Upgrade to 100% uptime today!
    * Premium NYC collocation and custom dedicated servers
    call 1-877-MY-RUSKO or paul [at] rusko.us

    dedicated servers, collocation, load balanced and high availability clusters

  27. #27
    Originally posted by freeflight2
    cord: there is a cluster solution coming up, type in "cluster" at @mysql.com, read what they whitepaper has to say, give these guys a call and be ready to pay some $$ but it'll be worth it.
    You might also get a mysql guru - you can have 100% mysql uptime with ordinary scsi boxes 4GB RAM which are replicating each other (a master + 2 slaves with low_priority_updates takes you very very far and you can "swich on the fly" if the master breaks...)
    ahh, how nice it would be if people knew what they were talking about. oh well, one day maybe..

    their clustering solution is available. it is *in-memory only*. in other words, if your whole database can fit into each node's ram, only then would you be able to use mysql's clustering. it has other limitations as well. this product is not a solution for you and i, who are typical mysql users. it is a meak attempt to muscle in on cach/e (pronounced casheh) and other in-memory dbs often used in telco environments, ie for sms routing.

    we know about mysql replication - we are as good or better, at reading documentation. asynchronous replication can not provide consistency, only coherency with an often unacceptable time bound. this only works if your dataset is non-deterministic.

    let's see if i can clear it up with two examples.

    google. google has several datacenters, each with it's own (diverging) copy of the index database. naturally, they would have a very hard time keeping them all completely in sync in realtime (consistent). if you search on www, www2 and www3, you are likely to get slightly different results. this does not matter, as it does not interfere with your usage of their service.

    paypal. people often ask me why paypal has to take their systems down for a few hours of maintenance now and again. 'google is never down, why can't paypal do the same thing?'. the answer is simple - their databases need to be consistent (coherency with a 0 time bound, ie realtime). if they were not, for example if they were to follow your advice and use replication, i could log in, use one cluster to make an $69 payment to your sister, jump to another one (before that transaction gets replicated to it) and withdraw all of my funds (including the $69 i just sent). either paypal or your sister are out of $69.

    in short, replication is acceptable only under certain (very rare) conditions where you do not care about the consistency of your data.

    paul
    * Rusko Enterprises LLC - Upgrade to 100% uptime today!
    * Premium NYC collocation and custom dedicated servers
    call 1-877-MY-RUSKO or paul [at] rusko.us

    dedicated servers, collocation, load balanced and high availability clusters

  28. #28
    Join Date
    Nov 2002
    Posts
    2,780
    Originally posted by rusko
    how is this clustering mysql then? you just have a separate mysql box =]

    p
    OH, I didn't actually read thru the whole thread. I wasn't referring to any clustering. But if it was for HA setups, I know there are some third party software out there that would allow a standby, active node setup. It isn't load balanced, but it would be high availability.

  29. #29
    Originally posted by cord
    This seems intersting. I am in no ways looking at clusters for, mostly I am educating myself. It would be such a waste to let free information fly by without taking advantage eh?
    I have made a point of following rusko around to see what he has to say. Dont tell him though
    i am being intellectually stalked? you make me blush =]


    The problem I foresee with splitting datasets is the following:
    databases are designed such that it is impossible to know in advance what kind of splitting would benefit you most. In other words, depending on the type of query, you have to make differnt divisions. I would assume that one can write a semi-intelligent system to split the dataset on-the-fly, but I guess it is not efficient or else someone would have already done it (that is if it hasnt been done).
    this is a bit off-topic and rdbms is certainly not my primary area of expertise, but let's see if i can explain why this is unnecessary with modern rdbms.

    if you have a database with no indexing, all of your searhes will be O(n) or worse, meaning the complexity of the search will be linearly proportional to the number of rows in the dataset being searched. if i had a table with 1000 rows and was looking for one row where id=666, the dbms would have to look at all 1000 rows (worst-case scenario, i'd much rather not examine best-case because it gets us into some more involved math).

    if you have an index, in this case on id, the search will be O(1) provided that your index implementation is O(1). this means, essentially, that the dbms could hand me that appropriate row without going through the 999 others matching ids against the one supplied in my query. in other words, the complexity of the search is not proportional to the number of elements in the dataset. in practice, certain index searches will be O(1+log^n), but let's leave that out for now.

    this brings me to my point. the only reason you would want to split the dataset is to reduce the 'n' (number of elements). if your database is correctly optimized (and for simple cases such as above, it is automagic nowadays), your 'cost' per search is either not at all or negligibly dependent on 'n', so halving it and splitting the load between two boxes makes no sense.

    from an implementation point of view, you would have to write a sql parser and then figure out a way to actually split the dataset based on the query. it may look easy in the case of " select * from pr0n where description like '%yo mama%;' ", but try doing it in a query with a bunch of joins and aggregates and you will be asking for a vacation at abu ghraib like it was a caribbean club med resort, in comparison to the pain of implementing this. given what i said above about the benefits of doing this in the first place, it would be nothing more than technological masturbation.

    paul
    * Rusko Enterprises LLC - Upgrade to 100% uptime today!
    * Premium NYC collocation and custom dedicated servers
    call 1-877-MY-RUSKO or paul [at] rusko.us

    dedicated servers, collocation, load balanced and high availability clusters

  30. #30
    Join Date
    Nov 2002
    Posts
    2,780
    And if you want Database clustering, don't use mysql, use Orcale on Sun clusters.

  31. #31
    Originally posted by Mfjp
    OH, I didn't actually read thru the whole thread.
    what is this, slashdot? RTFA! <g>


    I wasn't referring to any clustering. But if it was for HA setups, I know there are some third party software out there that would allow a standby, active node setup. It isn't load balanced, but it would be high availability.
    more or less, yes. this is what i have been maintaining all along =]

    paul
    * Rusko Enterprises LLC - Upgrade to 100% uptime today!
    * Premium NYC collocation and custom dedicated servers
    call 1-877-MY-RUSKO or paul [at] rusko.us

    dedicated servers, collocation, load balanced and high availability clusters

  32. #32
    Originally posted by Mfjp
    And if you want Database clustering, don't use mysql, use Orcale on Sun clusters.
    erm, what does sun have to do with this? especially since suns scale by adding more cpus and ram into the same machine (classic *nix bigiron, see ibm's power and hp's pa-risc) as opposed to clustering small cheap boxes (ie scaling up as opposed to scaling out)?

    paul
    * Rusko Enterprises LLC - Upgrade to 100% uptime today!
    * Premium NYC collocation and custom dedicated servers
    call 1-877-MY-RUSKO or paul [at] rusko.us

    dedicated servers, collocation, load balanced and high availability clusters

  33. #33
    Originally posted by Mfjp
    And if you want Database clustering, don't use mysql, use Orcale on Sun clusters.
    Tired to stay quiet. No, Oracle on Sun, as good as it's, is not the only solution. We can have a completely functional HA solution with XA-compliant relational database on Linux and as "free" as MySQL based.

    Peter.

    P.S. anybody who wants to argue this point must be prepared to stay technical and knowledgeable in this area. Thank you.
    Open Solution, Inc
    http://opensolution-us.com

  34. #34
    Hi guys,

    Thnx for giving some much ideas.

    basically i have gothrough some doc as well as Dell Cluster with PowerVault guide.

    So i have come out with this design:


    ................../server2 ----------\
    ...............LAN.....|..............SCSI Cable
    .............../..........|.....................\
    server1--..........LAN....................PowerVault SCSI Share Storage
    ...............\..........|...................../
    ..............LAN......|.............SCSI Cable
    ..................\server3 -----------/

    remark:
    server1 - Firewall and Balancing
    Dell 1750, xeon 2.4, 1GB ram, 2 x 36 SCSI 15K RPM, RAID 1

    server2 & serve3 - Application Server active/active implementation
    Dell 1750, dual 2.4, 2GB Ram, SCSI cable connect to Powervault

    PowerVault will be 2 PERC4/DC connector for server2,3 and
    (4 x 36GB SCSI 15K RPM for RAID 5, 1 x 36GB for HotSpare.

    The server1,2,3 is a LAN, server 1,2,3 having 4 NIC.


    My Concern:

    1) Assume server2,3 totaly don't have SCSI hardrive their own, they boot from the PowerVault SCSI Share Storage, so we going to partition PowerVault to 3 partition:
    Partition A for server2,
    Partition B for server3,
    Partition C for filesharing,

    server2 and server3 will have their own Apache ?, Mysql,etc..
    And some of the application data may put into the Partition C for BOTH of them access. Will this do ?

    2)If server1 also have to connect via HBA/SCSI cable to External Storage, then we have to get the Dell EMC which is costly! So is out of my consideration.

    3)If using PowerVault 22xS ,a split-bus module, two EMMs ,network adapters,PERC cards,SCSI hard drives,HBA drivers, any additional peripheral components needed. We going to use LINUX, will do be able to run in this Cluster System

    4)How about just get 2 x Dell Poweredge with RAID 5 and no more PowerVault...?

    anyway thx ..

    -qaz
    ,,Cluster..HBA..RAID..DUALCHANEL..U320..10/15KRPM...@$#fZzz
    {GOIN2MAD DOTCOM}

  35. #35
    Listen, seriously. This is not a typical situation. You are trying to use us and I refuse to help you in that. If you need a solution, just ask for sure few people on this forum can help you. But, I think they expect to be paid for that.

    Peter.
    Open Solution, Inc
    http://opensolution-us.com

  36. #36
    Hi Pete,

    ok. i already ask from Dell Support for this solutions via phone, what they giving is the PowerVault Share Storage Solutions.

    And all together cost me around USD20K. I mean if this is the correct solutions, i m ok with Dell pricing, what i m afraid of is getting a wrong solution nad wrong hardware

    Thats why i keep asking for advise. If anyone willing to share their experience, thanks for them. This the main purpose of the forum as well as the internet. "Sharing" .
    Just like Jay from Ethr.net, again many thanks for his advise. He is a nice guy.

    To who not willing to share with ppl, u can just keep silent, no one is forcing you, right?

    -qaz
    m i doing something wrong ?

  37. #37
    there is a difference between asking for advice and expecting to be spoonfed with the basics. if you go to math class and don't know multiplication, expect the professor to throw you out because you really should know multiplication before you take the calculus course.

    we asked you to go off and spend a few weeks doing research, trying to understand the underlying concepts and in general, to do your homework.

    here is a link that will help you ask questions better, resulting in people being more willing to help you: http://www.catb.org/~esr/faqs/smart-questions.html

    paul
    * Rusko Enterprises LLC - Upgrade to 100% uptime today!
    * Premium NYC collocation and custom dedicated servers
    call 1-877-MY-RUSKO or paul [at] rusko.us

    dedicated servers, collocation, load balanced and high availability clusters

  38. #38
    Originally posted by pnorilsk
    Tired to stay quiet. No, Oracle on Sun, as good as it's, is not the only solution. We can have a completely functional HA solution with XA-compliant relational database on Linux and as "free" as MySQL based.
    noone is arguing against that, at least not i. the issue at hand is that we are talking mysql and mysql is not xa-compliant. the freeness of it is not the issue, hosting service providers are locked into mysql because the clients' applications require it.

    with that said, i would be more than glad to hear about an example implementation of what you've mentioned.

    paul
    * Rusko Enterprises LLC - Upgrade to 100% uptime today!
    * Premium NYC collocation and custom dedicated servers
    call 1-877-MY-RUSKO or paul [at] rusko.us

    dedicated servers, collocation, load balanced and high availability clusters

  39. #39
    Originally posted by rusko
    noone is arguing against that, at least not i. the issue at hand is that we are talking mysql and mysql is not xa-compliant. the freeness of it is not the issue, hosting service providers are locked into mysql because the clients' applications require it.
    So, take another one. You mentioned it in your previous note. I am trying to say that Oracle is not the only one XA-compliant database around.

    Originally posted by rusko

    with that said, i would be more than glad to hear about an example implementation of what you've mentioned.

    paul
    And I may be glad to do it, but I expect to be asked nicely. Then, I must protect some proprietary (even mine) information - I am sure you understand it. I indicated that one of my solution is sitting in the JPMorgan Private Client Group - backend is Sybase more than 1TB connected to three SMP nodes. I also participated in the design and development of first NFS type cluster later mutated into HACMP in the AIX world.

    Peter.
    Open Solution, Inc
    http://opensolution-us.com

  40. #40
    Originally posted by qazakax
    Hi guys,

    Thnx for giving some much ideas.

    basically i have gothrough some doc as well as Dell Cluster with PowerVault guide.

    So i have come out with this design:


    ................../server2 ----------\
    ...............LAN.....|..............SCSI Cable
    .............../..........|.....................\
    server1--..........LAN....................PowerVault SCSI Share Storage
    ...............\..........|...................../
    ..............LAN......|.............SCSI Cable
    ..................\server3 -----------/


    Here is a problem we tried to avoid. The very rudimentary knowledge you were able to get with Paul's and mine help led you to believe that you can do it. But your note demonstrates opposite. You have few choices and it's up to you. But, I am in complete agreement with Paul, you need serious help. And I am sure these 20K will carry you far in this forum.

    Peter Kinev.
    Open Solution, Inc
    http://opensolution-us.com

Page 1 of 2 12 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
  •