dynamic harware patitioning
-
07 Nisan 2007 Cumartesi 12:40
i wish to ask about how dynamic hardware partitiong [which was very strongly discussed about 1,5year ago] is connected with viridian. i saw many presentations and not understand the connection point between DHP and what i can configure via Virtualization Management Console. will i have possibility to partition hardware via this console or will it be made automaticly by the system.
f.ex - if i have 4 processors and 4 virtual system - it's simple configuration. but if i have 2 processors and 4 vsystems? will i be able to exlusively attach one of the processors to one system and use the second one for the lasting 3 vsystems?
even more complexed situation is with memory and bridges [ i read it is now tested on x64 platforms], but still can not understand are those two technologies or one, and is the VMC the console for dynamic hardware partitioning as well?
Tüm Yanıtlar
-
14 Nisan 2007 Cumartesi 13:49
Hi.
I am not sure if I understand your question correctly as I am unsure about what you exactly mean by "dynamic hardware partitioning". Some x86 server hardware is capable of being partitioned from a pure hardware perspective (example includes the IBM x440/x445/460/x3950 families and the Unisys ES7000 families). These are not usually considered "dynamic partitioning" features as they typically requires a restart. Also the granularity of these partitioning schemas is very flacky (i.e. a minimum of 4 sockets). This has usually nothing to do with what Viridian will be able to do and manage. Consider these two technologies (hw partitioning and Viridian) as complementary instead.
That is: you can have an 8 socket system with 2 hardware partitions of 4 sockets each with one being used "natively" by a Windows install (with SQL for example) and the other one being "Viridianized" so that you could create more virtual partitions with a MUCH greater level of granularity (the Viridian management tools will not be able to deal with the hw based partitioning as far as I know at least).
For your doubt about how to allocate the processors on the "viridianized" partition well that is the basic of how any virtualization solution would work: oversubscribing processors. So you can have 4 sockets (i.e. 8 cores or pCPU's) in the Viridian server/partition and be able to create a number of guests whose sum of vCPU's exceed 8. You can then let the system map the workload running on the vCPU's to the pCPU's with its own scheduling algorythms or you might want to choose to pin yourself a particular Guest to a particular pCPU so that you might have Guest 1 pinned to pCPU 0 and all other Guests share the remaining 7 pCPU's (this would imply you rate Guest 1 more important than the others)...
Massimo.
-
22 Nisan 2007 Pazar 11:58
what i mean by dynamic hardware partitioning is there: http://www.microsoft.com/whdc/system/platform/server/dhp.mspx
i did not saw possibility to assign virtual processor or phisical one in virtualization management console nighter in virtual machine manager. it rather looked like the maxium numer of processor was the phisical number. do you now there will be possibility to exlusively assign hardware resources as well as create virtual ones or just assuming? it is not only about processors but network cards, and the very special treatment for the memory [NUMA?]
-
25 Nisan 2007 Çarşamba 20:54
OK so take into account we are in "speculation mode" here since we don't see anything for true but my understanding is as follow: you should really see this DP (dynamic partitioning) thing as a layered technology underneath Viridian (and Virtualization Manager). Consider Viridian as a "role" for your Windows host and, as I mentioned in my previous post, you can have a big piece of hw that can be sliced into many partitions (used to be "static partitions" now they will get dynamic with DP) and one of these partitions might have a virtualization role, the other one might have a database role etc etc. So with DP you will be able to hot-add processors (and NIC's and Memory etc) into these partitions and the "role" will pick up these new resources. So the DB parition through DP will get 2 new cpu's and 16GB of additional memory that the DB can instantly use on the fly. On the other partition (i.e. with the "virtualization role") Viridian will dynamically see 4 new cpu's and 32GB of additional memory that can use to either boost the performance of the current vm's or it can use these additional resources to create additional vm's. The point is that Virtualization Manager will be able to deal with the number of CPU's (and memory and NIC's etc etc) that sees at a certain point in time while DP is responsable for changing this amount of basic resources on-the-fly without a reboot .... something that I understand today is hard to imagine.
Massimo.
-
07 Mayıs 2007 Pazartesi 19:37
...must be satisfying for now q:
waiting for a beta1 - should be in couple of weeks (:
.png)
