The other question is whether the subject is supported to install Oracle VM (configured according to the note http://www.oracle.com/technology/tech/virtualization/pdf/ovm-hardpart.pdf) at the nodes of Exadata.
Thank you very much for the help
I found references that Oracle VM does indeed exist but it's restricted to internal use, so I guess this is not a technical issue but a marketing policy?
I've been looking at the possibility of using an Exadata database deployment, with a front end / middle tier application, relatively small that do not justify getting an additional Exalogic. It would be logical to have this installed on the Exadata, and Oracle VM would provide the required tier partitioning.
Hopefully there are some plans for the future to release Oracle VM on Exadata.
Indeed I mixed up the two items. Sadly I hoped that there was indeed an Oracle VM for Exadata, which would have been logical / practical for a SME organization moving to the Exastack platform.
Just to see that I understand the current Exastack Engineered systems offers, I am outlining the salient points:
1. Exadata - Is designed (marketed) to be an Oracle 11g + RAC exclusively, running under Oracle Linux. CPU licensing is satisfied by physically (hardware) disabling CPU's supplied in the unit.
2. Exalogic - Is designed (marketed) for the middle-ware / applications, Using Oracle VM + Oracle Linux, and optionally the Oracle SOA Suite middleware Applications, plus third party application, and possible other host O/S supported by Oracle VM.
If an organization has a web application, based on traditional web server front tier, application server tier (JAVA / JEE), and a data tier (Oracle DB), then what are the choices to move to the Exastack platform:
1. Move the Web / application server tiers to Exalogic, and the data tier to Exadata, seems the way the engineered platform is marketed, however depending on the application design / usage, this may not justify that level of investment. e.g. If the application is application intensive, then an Exalogic with a relatively small Oracle DB instance is required, alternatively if there is a DB intesive application, then an Exadata with a relatively small web / application server tier is required. Hence perhaps would it not be logical to offer such a hybrid "Exastack" model for SME starting out, then when volumes demand it scaling horizontally:
e.g. Get Exadata for the Oracle 11g + RAC, and reserve 2 or more processing units (X3-2) for the middle-ware (Oracle VM + Oracle Linux) in the same physical rack, when volumes increase, a dedicated Exalogic is added and the middle-ware applications migrated to the Exalogic platform, freeing up (scaling up) the Exadata platform.
In addition, given the licencing restriction, can Exadata be partitioned for a production database instance/s and development instance/s on the same physical rack? Note; this might be somewhat worked around, since a HA demanding business may have a DR site with a second Exadata, which can host the development DB instances when in normal operations mode.
I feel this is more of a marketing rather than technical issue here, but if there is some flexibility in the Exastack configuration, would simplify and lower the affordability bar for SME / start-ups.
As far is i know it is not supported to install weblogic en Exadata compute nodes, at least i have never heard of anyone doing that. The only option if you want to mix Exalogic and Exadata in your rack is to go for a supercluster, based an SPARC systems, this also means that you need to run on Solaris (Solaris in intel is also an option on the compute nodes of the Exadata btw).
For applications that do not justify getting both an Exadata & Exalogic, because either do not have high volumes and/or it's start-up business, then would be logical to install on one platform, Exadata both the database tier and the middle-ware tier so it's affordable to purchase 1 Exadata rack for production + 1 Exadata rack for DR, rather then to have 1 Exadata , 1 Exalogic for production, and 1 Exadata , 1 Exalogic for DR. The latter configuration, i.e. and Exalogic rack may be added when the business volumes increase, e.g. after 1.5 to 2 years.
Having to go for a supercluster, is even less realistic, 1. do not need SPARC / Solaris, 2. Puts the costs beyond the budget of the SME / start-up.
The alternative is to forget about the "Exastack" engineered systems, as these are effectively not marketed to SME / start-ups, which would be a pity, because from a technical point of view it will make sense to consolidate.
to illustrate this, here is a possible scenario:
Get an Exadata 1/8 rack for the data tier for production + 1 Exadata 1/8 rack for DR
Purchase 3 X X3-2 SUN servers (exactly what is installed on the Exadata), and configure these with Oracle VM + Oracle Linux, and install the web / application server. If these can be physically installed into the Exadata rack, then it will be reusing the common networking fabric thus saving on having to get an additional rack + network switch etc.
I think you are correct in targeting 1/8th racks for your SMB. The cost of entry is much more acceptable than an X3-2 Quarter rack. Also, using the X3-2 standalone servers would be a good option for your OVM environment. As for placing the X3-2 standalone servers in the Exadata rack, Oracle does not allow for using the Exadata rack space for anything other than Exadata components.
Understand that, my question is for SMB can we have an Exalogic + Exadata in one rack? Rather than having / forced to get both. As business grows, then it would be justified to get the two specialised racks and split load.
Note: why is Oracle VM not used in Exadata? This would be more flexible than disabling hardware? At a very minimal cost / overhead!
Depends! SMB can cover such a big range, and this may be true for the smaller end of the scale, or departimental servers (where consolidation is not an option).
For the higher end of the scale, an Exadata could be an affordable option, e.g. Starting on low volumes with projections in 2 to 3 years for higher volumes that will make use of the Exadata throughput.
My point being, if Exadata uses Oracle VM, then a 1/8 rack could use the hard CPU licensing for the production licence terms, and still use the rest on the CPU for test Oracle VM / DB instances. Rather have to get more server units.
Note: using a 1/2 rack (4 server units) could configure 2 units for production + 1 unit for test server instances.
I guess this is not really a technical topic, but rather a marketing product bundling, which could be better suited to a forum thread under sales & marketing if there was one!