Advertisement
If you have a new account but are having problems posting or verifying your account, please email us on hello@boards.ie for help. Thanks :)
Hello all! Please ensure that you are posting a new thread or question in the appropriate forum. The Feedback forum is overwhelmed with questions that are having to be moved elsewhere. If you need help to verify your account contact hello@boards.ie

dedicated link or MPLs

Options
  • 24-02-2011 11:32am
    #1
    Faolchu
    Registered Users Posts: 1,740 ✭✭✭


    Guys been asked by the boss to look into a dedicated link between two sites.

    Basically we have an office in Dublin city and a data centre in North Dublin, we replicate our data over night on our current connection, 4mb fibre with a VPN between them but its taking for ever.

    She suggested a dedicated link. if i have a dedicated link i would also need to keep my existing fibre so our internet, email traffic and VPN connections to our main office and other suppliers can still come in correct?

    Would MPLs be a better solution? I'm not too sure exactly what that is though.

    anyone got any suggestions as to providers for these services?


Comments

  • Options
    #2
    Faolchu
    Registered Users Posts: 1,740 ✭✭✭Faolchu


    iRock

    Currently we have a 4 meg fibre connection with Magnet.

    we have an office in Dublin city centre and a data centre in north dublin. We replicate all our servers out to the data centre at night for DR purposes. we connect to the data centre using a VPN tunnel between both sites, this VPN is over the internet.

    We then have a second VPN going out to our head office which is overseas

    we have a third VPN to a third party that we need to connect to to conduct business.

    on top of that some users have VPN connectivityfor working from home, but that's nothing major because any change to our connectivity would simple mean changing the connection details on their PCs.

    i've been asked me to look into a dedicated link to the data centre with a possible increase in bandwidth to go with it. the purpose is to achieve faster replication and more stable connectivity

    I'm thinking I have two options.

    1) simple do as she asks, get a dedicated link, 10 meg or something to reduce the replication times etc of our data. this requires changing one VPN. but this also means two bills for the company, one for this link and one for the existing fibre which users would continue to use for internet browsing, connecting to head office and email etc

    or

    2) replace all out connectivity into one bundle giving us a dedicated link to our data centre, access to the Internet for email and browsing and remote workers etc while also allowing us to connect to our head office and the third party via VPN while possible ruducing costs because we only need one solution to meet our needs and not two


  • Options
    #3
    Static M.e.
    Registered Users Posts: 3,088 ✭✭✭Static M.e.


    .. OK, first off ask what the budget is that will decide a whole lot.

    If it was me, I would ask magnet if the 4 MB line was burstable = Can they increase it to 10 MB for a day or even better a week which would allow you to test how long the backups take over different speeds.

    If the line wasn't burstable, I would ask them for a quote for 10 MB \ 20 MB and 50 MB. Also ask Smart Telecom and UPC, who both do really good prices at the moment.

    I don't see any point in getting a dedicated line from your office to the datacentre to do DR \ Backups etc which is only going to be used at night. Get a much larger line into your office which everyone can use during the day and which the backups can use at night.

    IF your backups are taking too long during the night, I would question how you are doing those backups and what software you are using.


  • Options
    #4
    FSL
    Registered Users Posts: 1,456 ✭✭✭FSL


    The limiting factor is the upload speed on your line. presumably the 4Mb is the download speed.


  • Options
    #5
    bigpaddy2004
    Registered Users Posts: 1,078 ✭✭✭bigpaddy2004


    Have you considered a wireless link? Do both sites have LOS to each other?


  • Options
    #6
    Faolchu
    Registered Users Posts: 1,740 ✭✭✭Faolchu


    Static M.e. wrote: »
    .. OK, first off ask what the budget is that will decide a whole lot..
    I've to come back with options including the various costs and they will look at them all, no budget has been set as yet.
    Static M.e. wrote: »
    ..
    If it was me, I would ask magnet if the 4 MB line was burstable = Can they increase it to 10 MB for a day or even better a week which would allow you to test how long the backups take over different speeds...

    the problem with this option is that with the link in the DR site only being 5MB we wouldnt see a huge improvement unless we changed that too. Though i'm sure they will work with us easy enough if i ask and bump the link for 5-10 days or something like that


    Static M.e. wrote: »
    ..
    I don't see any point in getting a dedicated line from your office to the datacentre to do DR \ Backups etc which is only going to be used at night.

    i agree to me its a wasted cost, money i'd rather spend on improving the servers or replacing ancient kit, even now we pay our data centre provider for every MB of bandwidth allocated to us even though in theory we only use it 10-11 hours a day.

    Static M.e. wrote: »
    ..
    IF your backups are taking too long during the night, I would question how you are doing those backups and what software you are using.

    its not backups its replication of virtual servers one of which has a database of over 10 GB so once theres a single change that whole database file gets replicated again on the next scheduled run. we also have off site backups running and these take a fraction of the time.

    FSL wrote: »
    The limiting factor is the upload speed on your line. presumably the 4Mb is the download speed.

    its actually 4 MB up and down and the DR site is 5mb up and down.
    bigpaddy2004 wrote: »
    Have you considered a wireless link? Do both sites have LOS to each other?

    no we're in the south city centre the data centre out by the airport plent if obstructions in the way like the Siptu building and others. a lot of the buildings around us are taller than us


  • Advertisement
  • Options
    #7
    Static M.e.
    Registered Users Posts: 3,088 ✭✭✭Static M.e.


    its not backups its replication of virtual servers one of which has a database of over 10 GB so once theres a single change that whole database file gets replicated again on the next scheduled run. we also have off site backups running and these take a fraction of the time.

    Have you looked at any of the Depup appliances outthere which can also do your backups? DataDomain from EMC have come down in price a lot recently.

    Also, with EsXpress and Im pretty sure Vaeem do dedupe too. Dedupe allows replicate only the bits that have been changed so for you, each time 1mb of the database has been changed it replicates that instead of the whole 10GB.
    Actually you could also look at vRanger and vReplicator from Quest too.

    Just a thought


  • Options
    #8
    Faolchu
    Registered Users Posts: 1,740 ✭✭✭Faolchu


    Static M.e. wrote: »
    Actually you could also look at vRanger and vReplicator from Quest too.


    we're using vRanger 5.0 to do the replications, the server in question that takes the longest is huge though, 170GB which contains AD, File & Print, Exchange & DNS so there would be a good few changes each day. it took 666 :eek: minutes last night. I know this config is bad but the ESX box isn't the most powerful, i tried taking file and print from it last week and it ground the whole site to a halt because the CPU isn't able to run 4 VMs and replicate at the same time, it also cant hyper thread and only has room for one CPU on the motherboard

    our other two servers are being done in around the 60-70 minute time (50-60 GB ) frame and thats with 2 gb database on one of them that creats at least local SQL backup daily as well as makes a copy of itself into a testing database over night. so again it has a large number of changes.

    Spoke to magnet they will increase out link to 8 mb for 30 days, and the guys in the DR site have agreed the same so that will at lease prove that by increasing the link we can increase the troughput and reduce replication times. if not then theres some other bottleneck and i'll seriously cry LOL


  • Options
    #9
    Static M.e.
    Registered Users Posts: 3,088 ✭✭✭Static M.e.


    Couple of ideas + Questions

    What are you doing for backups are you using this for your full backup solution?

    Do you plan on running everything from the Data Center if your Building\Office goes down or do you plan on restoring from the Data Center to the office if there's a problem?
    Faolchu wrote: »
    we're using vRanger 5.0 to do the replications, the server in question that takes the longest is huge though, 170GB which contains AD, File & Print, Exchange & DNS so there would be a good few changes each day. it took 666 minutes last night.

    Hmmm you could try doing a backup to another server before you do the backup to the dr site..which might speed up the backup but..

    Faolchu wrote: »
    I know this config is bad but the ESX box isn't the most powerful, i tried taking file and print from it last week and it ground the whole site to a halt because the CPU isn't able to run 4 VMs and replicate at the same time, it also cant hyper thread and only has room for one CPU on the motherboard

    I think the problem here is more your full configuration. What I would suggest is that you break your main server out into its roles

    (AD + DHCP + DNS + Print) Server -- ~30 GB should cover this > 1 Hour backup..
    File Server
    Exchange Server

    Why not get another server and spread the load?

    If the problem here is the lack of CPU / RAM, then I could lend you another server to try out the changes if you needed it (Dell PE 2850) or buy a new one\secondhand one. I really think you should break that server out, too many eggs and all that, plus it will make things a lot easier to manage.

    Faolchu wrote: »
    our other two servers are being done in around the 60-70 minute time (50-60 GB ) frame and thats with 2 gb database on one of them that creats at least local SQL backup daily as well as makes a copy of itself into a testing database over night. so again it has a large number of changes.

    If you make the changes outlines above you might not need to up your bandwidth but its worth testing anyway


  • Options
    #10
    Faolchu
    Registered Users Posts: 1,740 ✭✭✭Faolchu


    Static M.e. wrote: »
    What are you doing for backups are you using this for your full backup solution?
    No we have backups as well as the replications. we replicate the VMS out twice at night because the SQL server runs some backup process and management process at midnight so we replicate that just after our production day and then again 12 hours later to get the automated processes over. the system states, data and databases are also backed up off site independantly of the replications nightly.

    Static M.e. wrote: »
    Do you plan on running everything from the Data Center if your Building\Office goes down or do you plan on restoring from the Data Center to the office if there's a problem?
    If we go down everyone relocates to the data center, we lease rack and desk space (4 PCS). on the last test (mid November) I had everyone up and running and had verified the data was consistant with our end of day from the day before within two hours of arriving on site. basically the replication gets us back to between 17:30 and 19:30 the previous day, our processing tends to end around 16:30,

    we can if needed then do restores from our online backups to the DR site for the likes of exchange databases to get newer emails as the exchange backup happens later than the replication. Same with our network shares, these are backed up online every 2 hours so again if someone happened to be working late any documents would be more or less current. though obvioulsy the restore of these backups changes our Recovery time

    Static M.e. wrote: »
    I think the problem here is more your full configuration. What I would suggest is that you break your main server out into its roles
    agree 100%, the problem was that when i started i inherited kit, there was no IT department before me, everything was outsourced and the company basically kept stuff ticking over and not looking to the future, they hosed everything except that one server whioch contains DC< File& Print and exchange this ran in house on Win2K, i already tried to take the file and print off the largest server and because of limitations of the CPU on our ESX host it ground to a halt when replicating.
    Static M.e. wrote: »
    Why not get another server and spread the load?
    .... I really think you should break that server out, too many eggs and all that, plus it will make things a lot easier to manage.
    Already working on a proposal for that for the middle of the year to get teh expenditure for a bettwe server, I plan on having 2 ESX hosts and replicate from one to another on our LAN a couple of times a day then replicate from teh second one off site nightly. that way if the production ESX goes off line I can just power up the most recent replica (probably only an hour or two old), if the building goes then i can at least go back to end of the previous day

    alternativly have two esx hosts with shared storage, this will help balance teh load for RAM and CPU etc and give redundancy on a hardware level which is something i currently do not have.

    our ISP has agreed to double our bandwidth for little or nothing for 30 days, and the data centre have agreed a free 30 day upgrade to 8Meg. the problem here is that if the upgrade on the DR becomes permenant the cost is an additional 3-4K a year


  • Options
    #11
    Static M.e.
    Registered Users Posts: 3,088 ✭✭✭Static M.e.


    Hmm you do seem to have covered off a number of angles.

    The only other thing I can think off is to take your File Server off the replication cycle and move it to an online backup service like KeepITSafe.* You could then have your File backups running during the day doing incremental's and the replication to the DR Site would be xGB less, which might remove the need to upgrade your line.

    You could also look at SQL mirroring with the transaction log moving across every hour.. Im not sure if you can do this with Exchange...Anyway, the idea is that you have your data moving across to the DR site at a steady steam all day rather than one big push every night.

    .. Im running out of ideas now!!


    * I dont work for KeepITsafe in any way


  • Advertisement
  • Options
    #12
    Faolchu
    Registered Users Posts: 1,740 ✭✭✭Faolchu


    Static M.e. wrote: »
    Hmm you do seem to have covered off a number of angles.
    yeah when I started this whole project and brought all our servers in house the guy i worked with didnt trust replication as a solution becuase virtualisation was new to him, he wanted the old backup tapes brought into the mix but i convinced him otherwise with the online backups being at least 50% cheaper than media based ones and it gives is a fall back in the event of a failed replication just before a disaster, even a disaster during a replication is no big deal because we can restore the system state from backup and the place our data files on top of that, granted the recovery time is greatly increased but it gicves us a second chance.

    I've worked for some clowns taht document a DR plan for their auditors and a) never test it or worse still b) doesnt actually exist. I've seen one place state they have a hot site but the contract to that site ended 2 years previously and they never updated the plan or entered into another agreement with another data center.

    Static M.e. wrote: »
    The only other thing I can think off is to take your File Server off the replication cycle and move it to an online backup service like KeepITSafe.* .
    with it being part of the same server as exchange and the DC currently its a little difficult to do that. but like i said the splitting of functions is on the cards once i get approval for the expenditure. out Audit in 2010 highlighted this and they expect it done by year end. I'm aiming for the end of June


    Static M.e. wrote: »
    You could also look at SQL mirroring with the transaction log moving across every hour..
    .
    we have 2 SQL databases both around 2GB but the amount of actual transactions is minimal so we actually have our database set to a simple recovery mode.
    Static M.e. wrote: »
    Anyway, the idea is that you have your data moving across to the DR site at a steady steam all day rather than one big push every night.
    the business is comfortable with a loss of 24 hours of data in the event of a disaster but I'm not comfortable with that. I'd much rather have replications out every 3-4 hours because in that 24 hour period that they are comfortable with a critical email could be lost or something. maybe i'm too paranoid

    Static M.e. wrote: »
    I dont work for KeepITsafe in any way
    are you standing over my shoulder listen to my phone calls? we use KIS and was just on to them about a failed backup LOL


Advertisement