C1000-038 · Question #27
A customer is running a z14-702 with two zIIPs. They started noticing that the zIIP utilization during peak hours was averaging about 85% due to a new Java application. After examining RMF, they…
The correct answer is A. Adding a third General Purpose processor and a zIIP is the only solution that will eliminate both. Option A is correct because the environment faces two distinct capacity problems simultaneously: the zIIPs are saturated (85%) and the GP CPs are overwhelmed (90%+), partly from eligible work overflowing off the zIIPs. Turning on SMT can increase zIIP thread throughput, but it…
Question
A customer is running a z14-702 with two zIIPs. They started noticing that the zIIP utilization during peak hours was averaging about 85% due to a new Java application. After examining RMF, they realized that during these periods, much zIIP eligible work was overflowing to the general purpose CPs, causing them to also run over 90% for considerable periods of time. They have a limited budget for new equipment. After reading about Simultaneous Multithreading (SMT), they think they can solve their problem by turning SMT on instead of purchasing either another GP, a zAAP, or a third zIIP in order to get more throughput. What must they consider before making the decision to turn on SMT or purchasing another processor?
Options
- AAdding a third General Purpose processor and a zIIP is the only solution that will eliminate both
- BAdding a zIIP cannot be done because a 702 with two General Purpose CPs can only have two
- CIf there is a large amount of Java work, adding a zAAP would substantially reduce the zIIP
- DTurning on SMT, while providing more zIIP throughput, could affect the zIIP intensive workloads
How the community answered
(35 responses)- A43% (15)
- B20% (7)
- C6% (2)
- D31% (11)
Explanation
Option A is correct because the environment faces two distinct capacity problems simultaneously: the zIIPs are saturated (85%) and the GP CPs are overwhelmed (90%+), partly from eligible work overflowing off the zIIPs. Turning on SMT can increase zIIP thread throughput, but it adds no physical GP CP capacity - the GPs will remain bottlenecked regardless. To eliminate both bottlenecks, you need a new GP CP (to absorb the genuine GP workload) and a new zIIP (to absorb the eligible overflow), making A the only complete solution.
Why B is wrong: B incorrectly treats the specialty-processor-to-GP ratio as an absolute ceiling. The IBM rule says zIIPs cannot exceed the number of GP CPs - but adding a GP first (as A proposes) raises that ceiling, making a third zIIP possible.
Why C is wrong: zAAPs were discontinued as a separate processor type on the z13 generation; their eligible workloads (including Java) were folded into zIIP eligibility. There is no zAAP product available to add to a z14.
Why D is wrong: D correctly notes that SMT shares core resources between threads, which can reduce per-thread performance - but it only partially addresses zIIP capacity and says nothing about the GP CP overload, so it's an incomplete consideration, not the key decision factor.
Memory tip: Think "SMT multiplies threads, not processors." It's a throughput aid within existing zIIP hardware - it cannot create new GP capacity, so whenever both processor types are saturated, hardware must be addressed on both sides.
Topics
Community Discussion
No community discussion yet for this question.