A host has 256 gigabytes of memory and runs 40 guests each allocated 8 gigabytes. What is the overcommitment ratio?
Answer: A
Forty times eight is three hundred and twenty gigabytes allocated against two hundred and fifty-six physical, a ratio of one and a quarter. Modest ratios of this kind are routine in practice. Guests rarely use their full allocation.
Please log in or register to report this question.
Three of the following are forms of virtualisation. Which is the odd one out?
Answer: B
Putting machines in a rack changes nothing about how they are addressed or used. The other three all decouple a logical resource from the physical one beneath it. Consolidation is not virtualisation. Addressing is what changes in the other three.
Please log in or register to report this question.
Network virtualisation in a cloud typically uses
Answer: C
Correct answer: overlay networks over a shared fabric
Each tenant’s network is built as an overlay - VXLAN, GENEVE or similar - whose frames are encapsulated and carried across one shared physical fabric. That gives every tenant an apparently private network with its own addressing, without a cable per tenant and without the twelve-bit VLAN limit that would otherwise cap the number of segments at about four thousand.
Please log in or register to report this question.
Assertion (A): Overlay networks allow two tenants to use the same private address range without conflict.
Reason (R): Each tenant's traffic is encapsulated and carried over the shared physical fabric.
Answer: D
Correct answer: Both A and R are true, and R is the correct explanation of A
Both are true and the reason is the mechanism. Encapsulation wraps the tenant’s frame inside an outer header carrying a tenant identifier, and only the outer header is used for forwarding on the physical fabric. The inner addresses are therefore scoped to the tenant, so two tenants can both use 10.0.0.0/8 without the fabric ever having to tell them apart by address.
Please log in or register to report this question.
Consider the following statements about virtualisation overhead.
1. Compute overhead is small with hardware assistance.
2. Input output overhead is generally larger than compute overhead.
3. There is no overhead of any kind.
Answer: A
Correct answer: 1 and 2 only
Statements 1 and 2 are correct: hardware assistance has reduced compute overhead to a few per cent, while input and output overhead remains larger because every operation may involve a trap to the hypervisor and an emulated device. Statement 3 contradicts both - there is always some overhead, and pretending otherwise is what leads to virtualised databases performing unexpectedly badly.
Please log in or register to report this question.
State whether the following is true or false.
Input output intensive workloads suffer more virtualisation overhead than compute intensive ones.
Answer: B
Correct answer: True
An I/O-bound workload crosses the hypervisor boundary on every operation, with device emulation, interrupt handling and additional copies at each crossing, so the relative cost is high. A compute-bound workload runs guest instructions directly on the processor and pays almost nothing. Paravirtualised drivers, SR-IOV and device passthrough all exist to reduce that gap - and passthrough brings its own cost, as the next item shows.
Please log in or register to report this question.
Device passthrough gives a guest
Answer: C
It removes almost all the overhead for that device but prevents live migration of the guest. Graphics and high performance networking are the usual cases. Migration is what is given up. The device belongs to one host only.
Please log in or register to report this question.
Assertion (A): A guest using device passthrough usually cannot be migrated live.
Reason (R): The physical device it depends on is not present on the target host.
Answer: D
Correct answer: Both A and R are true, and R is the correct explanation of A
Both are true and the reason is the cause. Passthrough gives the guest direct control of a physical device, so the guest holds state inside hardware that the hypervisor cannot see or copy, and the target host has a different card. Live migration therefore cannot preserve the device, which is the trade passthrough makes: near-native I/O performance in exchange for pinning the guest to one host.
Please log in or register to report this question.
