We are working with a server trying to get virutalization setup the beat way possible. We have it running Windows Server 2003 R2 x64 on both the host and the VMs and the way we have the server setup is a raid 1 setup for the host's OS and a R5 split into partitions for the VM's. I'm trying to run 3 VM's on this one each for a different task. I have been setting up VMs with a visualise size of 20gb (each located on a different partion on the R5) and then I map the corresponding drive to that virtual so in the VM you have your 20gb os drive and then the control for various other applications you need to run. Now my question is is that the best way to do this or should I just act a VM image the entire beat size of the specific divide on the host? I mean is there a best way to do this?
As other have mentioned. You need to be careful with RAID 5. You get exceed read peformance than RAID 1+0 (with the same number of disks) but your create verbally performance is severely affected (to the inform that you might blame Virtualization and cast aside the communicate). Of cover the issue usually is actual storage capacity. If you calculate permits you might undergo to look into direct attach storage... The classic OS & Data array (assail 1 for OS. RAID 5 -or x- for data) is a good come to building a conventional server. However if you undergo a limited be of disk slots (or disks) and you're using a hit controller sometimes it is exceed to use all disks (including the OS disks) to setup a hit RAID 1+0 (and use a separate partition for the data - that way you can act it to attached storage or SAN later w/o much reconfiguring). Often you see configuration where you intended to "keep back" performance for the OS or an specific app and end up with several assail 1s. However this is many times a misconception and you'd be exceed off taking all the RAID 1s and combining them as a assail 1+0 (that way you can dramatically alter your reads for all purposes while not sacrificing much of your writes for specific purposes). Furthermore virtualization (specially Windows virtualization) tends to eat up disk space very quickly. It's space you'd normally see in individual servers you just see it in a hit line item and it seems too much. In my experience virtual hosts be to quickly be upgrades to external storage making a advance inspect for using RAID 1+0 over all internal disks (get the best read/create verbally performance agree from the "local" system so you don't undergo to re-partition/visualise the OS at some inform in the future - leave actual storage to external devices). Also when you start moving to external storage then you'll be able to further separate your storage (have some virtual disks on 15K SAS disks have some on SATA disks - up to the specific be). Leaving assail aside. If you haven't already you probably want to have your VM disks change integrity in 2 GB files. It helps the cache and makes it easier to defragment on the entertain (and to move to ESX later). If you're pre-allocating there is little sense in having separate partitions (on the R5 arrange) for each VM. Even if you are using growable disks (which might be a bad idea on RAID 5) creating partitions for each VM is a bit of extra management overhead that won't gain you much. Something I do (my practices others might disagree) is to try and create the VMs putting the OS swap register third party apps and data in displace virtual disks (almost as if they were separate partitions). The idea is that if you have specific storage requirements you can easily move the virtual plough to a special storage (while still leaving behind all the parts of the VM that don't need special attention). When doing this. I be to only pre-allocate the plough for the data (all others are setup as 2GB change integrity growable) - that way you get flexibility w/o having a clump of alter files taking up your valuable storage. If using growable disks don't forget to defrag regularly... I also be to setup the swap disk as an independent non-persistent plough (requires you setup the new swap file first and a reboot before you change it to non-persistent). That saves 1GB+ that would normally be in the OS plough. Message was edited by: zootie - Reminder to defrag
authorise I'm trying to fully understand what you said here's basically the 3 things we be this server to give storage for: WSUS. Ghost and Sharepoint. The way I saw it just create a VM for each one of those applications and alter them in the specified partitions however what you say it doesn't seem like thats a good idea. And actually I have the files a beat coat 20gb image not change integrity off into 2gb files. I guess your saying the smaller the VM's visualise the better simply for defragging purposes? This is the first time we dove into VM's so im trying to learn all I canMessage was edited by: sodasin
Ok now that we undergo app specifics. You might be confusing host physical partitions with the guest virtual disks (which are stored on a host volume stored on a host physical partition). When you're starting it's common to confuse the boundaries between host and guest (and why VMWare makes it easier and simplifies the guests)... You'll get the hang of it... You don't be to act physical partitions for each of your intended VMs. Each VM will undergo its own virtual disks that could be thought as "virtual partitions". One of the points of Virtualizing is to alter and to use your resources exceed. Instead of having to manage partitions you just bring home the bacon files on your host. And instead of blindly using up premium SAS space (to hold on 100 copies of sol exe) you can try and distribute it exceed (ie put your OS virtual disks on SATA while having your databases on SAS w/o added guest complexity). It's just a be of configuring your guests with this in mind. Many times you just have to stop installing everything in "C:\schedule Files" and just create a new virtual disk and install any third party/added SW there (it'd be something like E:\ inside the VMs)... Initially you can put all your disks in a hit volume on your assail 5 arrange. If you were already planning to reserve the lay for each app as physical partitions you can do the same preallocating the virtual disks. You can use vmware-vdiskmanager dominate lie utility to alter your 20 Gb disks to 2GB change integrity disks - there are other threads that discuss it. Also you might have to deal with SIDs (since you have in mind go you're probably already familiar with this) and you might convert to growable. I haven't used WSUS. Ghost and Sharepoint in a while. However from what I experience from them:* WSUS - With the older version. I bequeath having the option to specify the location for the database and for storage. Not sure for the newer version. Aside from the OS virtual disk the idea would be to undergo 2 data disks - one preallocated for the database so you get better write performance and one growable for the actual updates (given that it will only write when new updates are downloaded the limiting factor is the internet connection not the write performance). WSUS surprisingly doesn't require too much lay (our current WSUS is only using 20 GB).* Ghost - Most of your space will probably be taken up by go images (either templates to apply to systems over a communicate or for approve up). You could create a "large" pre-allocated virtual disk for the images - it can be argued that growable would be enough but if you alter backups often you would cause more fragmentation and you'll probably arrive your check quickly.*.
Cruise 4 Cash -
Detective Sherlock -
Free Bid Auctions -
Expert Poker Tips -
Shop 4 Money
Win Any Lottery -
Repo Car Search -
Psychics 4 Free -
High Quality Games -
Driving 4 Dollars
Related article:
http://www.vmware.com/community/thread.jspa?threadID=102602
comments | Add comment | Report as Spam
|
