Exams / SAP-C03 / Study guide / Storage performance and caching (EBS, EFS, FSx, ElastiCache, DAX)
Storage performance and caching (EBS, EFS, FSx, ElastiCache, DAX): SAP-C03 study notes and practice questions
What to know about Storage performance and caching (EBS, EFS, FSx, ElastiCache, DAX) for the SAP-C03 exam, from the Architecture Design and Implementation Built for the Cloud domain (25% of the exam guide), then 1 free question to test it. Part of the SAP-C03 study guide.
SAP-C03 describes how an application reads and writes data (block or file, shared or not, protocol, IOPS, latency) and asks for the storage and caching layer that fits.
Storage by access pattern
Need
Choose
Objects, any scale, durable
S3; S3 Express One Zone for single-digit millisecond access in one AZ
Block storage for one instance
EBS
Shared Linux file system (NFS), elastic
EFS
Windows SMB shares with Active Directory
FSx for Windows File Server
HPC or ML scratch with S3 integration
FSx for Lustre
NFS, SMB and iSCSI together, snapshots, clones, SnapMirror from NetApp
FSx for NetApp ONTAP
Low-latency NFS migrated from ZFS
FSx for OpenZFS
EBS
gp3: 3,000 IOPS and 125 MiB/s baseline included, provision up to 80,000 IOPS and 2,000 MiB/s independently of size; usually cheaper than gp2.
gp2: IOPS tied to size (3 per GiB) with burst credits.
io2 Block Express: highest IOPS, sub-millisecond latency and durability for critical databases; Multi-Attach for clustered apps.
st1 / sc1: throughput-oriented and cold HDD for large sequential reads.
Instance store is fastest but lost when the instance stops.
EFS
Regional (multi-AZ) or One Zone. Throughput modes: Elastic (scales automatically, pay per use; the default for spiky workloads), Provisioned and Bursting. Lifecycle management moves cold files to Infrequent Access and Archive classes.
Caching
ElastiCache (Valkey, Redis OSS, Memcached; serverless or node-based): lazy loading (cache-aside) or write-through, with TTLs to limit stale data. Use Valkey or Redis OSS for replication, Multi-AZ failover, persistence, sorted sets and pub/sub; Memcached for a simple multithreaded cache.
DynamoDB Accelerator (DAX): in-memory cache in front of DynamoDB with microsecond reads and no application logic changes beyond the DAX client; serves eventually consistent reads (strongly consistent reads pass through).
CloudFront caches at the edge; API Gateway caching for REST APIs; RDS read replicas are not a cache but offload reads.
How to choose
Scenario wording
Choose
“Database volume needs more IOPS without growing size”
gp3 with provisioned IOPS (io2 for highest)
“Many Linux instances share files across AZs”
EFS
“Windows file shares, DFS, AD integration”
FSx for Windows File Server
“Process S3 training data at hundreds of GB/s”
FSx for Lustre linked to S3
“Microsecond reads from DynamoDB, minimal code change”
DAX
“Cut repeated expensive SQL queries”
ElastiCache cache-aside
Common traps
EFS for Windows apps or FSx for Windows for Linux NFS clients.
Bigger gp2 volumes just to get IOPS.
DAX for strongly consistent reads or for non-DynamoDB databases.
Instance store for data that must survive a stop.
Caching without TTLs or invalidation, serving stale data indefinitely.
A car maker's engineering group keeps its design files on shared storage. About 800 Windows engineers reach the files over SMB with Active Directory permissions, and a farm of Linux render nodes must read the same files over NFS.
The group needs snapshots so engineers can recover earlier file versions, and it needs the storage to survive the loss of one Availability Zone. The IT team wants a managed service that does not require a file server to be patched.
Which solution meets these requirements?
Choose one answer.
AA. Create an Amazon FSx for NetApp ONTAP file system deployed across two Availability Zones, join it to AWS Managed Microsoft AD for SMB access, and mount the same volumes over NFS on the render nodes.
BB. Create an Amazon FSx for Windows File Server file system deployed across two Availability Zones, join it to AWS Managed Microsoft AD, and turn on shadow copies so engineers can restore earlier versions.
CC. Create an Amazon EFS file system with mount targets in each Availability Zone, and mount it on the render nodes and on the Windows engineers' workstations.
DD. Attach Amazon EBS io2 volumes with Multi-Attach to a pair of Linux file servers in one Availability Zone, and export the volumes to both groups of clients.
Show the answer and explanation
Answer: A
A. Right: Correct. FSx for NetApp ONTAP serves the same volumes over SMB with Active Directory permissions and over NFS, and it supports snapshots and Multi-AZ deployment. AWS runs the file system, so no file server needs patching.
B. Wrong: FSx for Windows File Server serves SMB and supports snapshots and Multi-AZ, but it does not offer NFS for the Linux render nodes. The render farm would have no access path.
C. Wrong: EFS serves NFS to Linux clients and does not offer SMB with Active Directory permissions. The Windows engineers would not get the access model they use today.
D. Wrong: Multi-Attach volumes are block devices that need a cluster-aware file system, and they live in one Availability Zone. The design also means managing the file servers, which the IT team wants to avoid.
Explanation
ONTAP is the FSx option that serves both SMB and NFS from one managed, Multi-AZ file system with snapshots. Windows File Server lacks NFS, EFS lacks SMB with Active Directory permissions, and EBS block volumes need a file server layer.
Want more on Storage performance and caching (EBS, EFS, FSx, ElastiCache, DAX)? The 5 full-length practice sets have 5 more Storage performance and caching (EBS, EFS, FSx, ElastiCache, DAX) questions, 375 in all, none of them in the free exam. One payment of $14.99, no subscription. See the practice sets →
Study Storage performance and caching (EBS, EFS, FSx, ElastiCache, DAX)