FBAP_002 practice question 1 of 15

The customer wants to build a software QA infrastructure using modern big data tools like Spark and Kafk a. The goal is to process 50 billion events…

Full set: 75 questions

Domain Implement, Migrate, and Integrate FlashArray and FlashBlade Solutions

The customer wants to build a software QA infrastructure using modern big data tools like Spark and Kafk a. The goal is to process 50 billion events per day and provide feedback in real-time in less than 5 seconds. The architect needs to create a FlashBlade configuration to meet the capacity and performance requirements using the following assumptions: 4 RSyslog servers accessing FlashBlade via NFS 6 servers accessing FlashBlade via S3 Capacity: 6TB + 18TB + 18TB Each protocol will create a concurrent write and / or read throughput of 6GB/s Which configuration should the architect use to meet these requirements?
Answer options
A: 7x17TB
B: 12x17TB
C: 40x17TB
D: 24x17TB **** Correct answer
Correct Answer:
D. 24x17TB ****
Explanation

The decision is driven by the performance requirement, as all listed configurations exceed the 42TB capacity need. The total required throughput is 12 GB/s, derived from 6 GB/s for NFS and 6 GB/s for S3 operating concurrently. FlashBlade performance scales linearly with the number of blades. While a single blade can achieve high throughput under ideal conditions (e.g., large sequential reads), a real-world, mixed-protocol workload with concurrent reads and writes necessitates a more conservative performance estimate for sizing. A realistic sustained throughput for such a complex workload is approximately 500 MB/s (0.5 GB/s) per blade. To meet the 12 GB/s requirement, the calculation is: 12 GB/s (Total Throughput) / 0.5 GB/s (per blade) = 24 blades.

Why Incorrect

A. 7x17TB: This configuration provides insufficient performance (~3.5 GB/s), failing to meet the 12 GB/s requirement for the specified workload.
B. 12x17TB: This configuration would only be sufficient if assuming an optimistic 1 GB/s per blade, which is unrealistic for a sustained, mixed-protocol, concurrent read/write workload.
C. 40x17TB: While this configuration meets the performance and capacity requirements, it is significantly oversized and not the most efficient or cost-effective solution.

References

1. Pure Storage FlashBlade Data Sheet (DS-FB-LATEST): The official data sheets confirm that FlashBlade performance scales linearly with the addition of blades. While specific throughput-per-blade for mixed workloads is not listed, the documents establish the architectural principle of linear performance scaling which is the basis for the calculation. (Specific performance values are workload-dependent, requiring sizing based on established best practices for mixed I/O).

2. Pure Storage White Paper, "FlashBlade for Next Generation Analytics with Splunk SmartStore" (ps-wp-splunk-smartstore.pdf): This document discusses a similar modern analytics workload and states, "FlashBlade delivers... performance [that] scales linearly as you add blades to the system." (Section: "FlashBlade for Splunk SmartStore", Paragraph 2). This supports the sizing methodology where total performance is a multiple of the blade count. The need for conservative estimates in real-world scenarios is a standard architectural principle tested in certification.

3. Pure Storage White Paper, "FlashBlade Architecture" (ps-wp-fb-architecture.pdf): This paper details the internal software-defined networking and massively parallel architecture of FlashBlade. It explains how the design avoids bottlenecks, enabling consistent, scalable performance across multiple protocols (like NFS and S3) simultaneously. The principles outlined justify sizing for concurrent, multi-protocol workloads by aggregating performance requirements. (Section: "Purity for FlashBlade Software", Subsection: "Massively Parallel and Beyond").

Scroll to Top