If you are studying for the Professional Cloud Architect exam, Memorystore is one of those services where the exam will not test you on operational depth, but it will absolutely test you on the trade-offs between the two engines and between the two Redis tiers. Get those straight and you will pick up easy points.
Memorystore for Redis comes in two flavors, and the difference between them is the first thing I want you to lock in.
The Basic Tier is the cheap, simple option. It is easy to spin up, the configuration surface is small, and it works well when you just need a cache and you can tolerate some downtime. The catch is that there is no high availability and no failover protection. If the underlying instance has a problem, your cache is unavailable until it recovers, and any in-memory data is gone.
The Standard Tier costs more and asks you to make more configuration decisions, but in exchange it gives you high availability with automatic failover. There is a replica ready to take over if the primary fails, so a single-node failure does not take your cache offline. That extra configuration is also a feature in disguise, because it means you have more knobs to tune for your workload.
For exam purposes, the rule I want you to internalize is simple. If a question describes a workload that needs uptime, resilience, or anything close to a production-grade SLA on the cache itself, the answer is Standard Tier. If the question is steering you toward cost-effectiveness, dev environments, or a workload where the cache is purely a performance booster and losing it is fine, Basic Tier is the right call.
The other comparison that shows up on the Professional Cloud Architect exam is Memorystore for Memcached versus Memorystore for Redis. These are both managed in-memory caches, but they are not interchangeable, and the exam will lean on a handful of specific differences.
Memcached supports automatic scaling. You can grow or shrink the cluster to match demand, which makes it a natural fit for highly elastic workloads where the cache size needs to track traffic. Redis on Memorystore does not give you that built-in elasticity, so if a question is about scaling a cache up and down dynamically, that is a strong signal toward Memcached.
High availability flips the comparison. Memcached on Memorystore does not provide built-in HA or failover. Redis, on the Standard Tier, does. So when a scenario calls for a resilient cache that can survive a node failure, Redis Standard Tier is the answer, not Memcached.
Data structures are the next axis. Memcached is a pure key-value store and that is the entire feature set. Redis supports strings, lists, sets, hashes, and sorted sets. If a question describes a leaderboard, a queue, a ranked feed, a set membership check, or anything beyond simple key lookups, Redis is what you want. If the workload is "cache this database query result by key," Memcached is sufficient and often preferred for its simplicity.
Persistence is another sharp distinction. Memcached is purely in-memory, so a restart wipes the cache. Redis can persist to disk, which means cached data can survive between sessions. For exam scenarios that mention surviving a restart or rebuilding a cache without hitting the backing database, Redis is the right pick.
On raw performance, Memcached is slightly faster for the simple key-value path because it does less. Redis carries a bit more latency due to its broader feature set, but it is still very fast in absolute terms. I would not pick a service on a speed difference alone, and the exam will not ask you to either.
For use cases, Memcached is the typical choice for session storage and database query caching where the workload is straightforward. Redis fits the more complex use cases: leaderboards, real-time analytics, pub/sub, and message queuing.
When a Professional Cloud Architect question puts Memorystore options in front of you, I want you to ask three questions in order. First, does the workload need high availability on the cache itself? If yes, Redis Standard Tier. If no, you are open to either engine. Second, does the workload need anything beyond key-value, or does it need persistence? If yes, Redis. If no, Memcached is on the table. Third, does the workload need the cache to scale up and down with traffic automatically? If yes, Memcached has the edge.
That ordering matches how the exam tends to weight these features. HA is the most common discriminator, data structures and persistence are the second most common, and automatic scaling is the tiebreaker.
My Professional Cloud Architect course covers Memorystore Redis tiers and the Memcached vs Redis comparison alongside the rest of the databases material.
