iTechGuides is reader-supported. When you buy through links on our site, we may earn an affiliate commission. As an Amazon Associate I earn from qualifying purchases. Learn more
“Why is my database slow?” If the same data is being read again and again, repeated database work may be the problem—and a cache may help. But “Should I add Redis?”, “Do I need a cache?”, and “How do I reduce repeated database reads?” are questions to answer only after identifying which reads are slow and how fresh their results must be. Caching can reduce repeated work; it will not fix every slow query, write bottleneck, missing index, or poor data model.
When a cache can help—and when it cannot
A cache stores data that can be reconstructed from a primary data source or a previous computation. It is most useful when requests repeatedly need the same data, the workload has many more reads than writes, or the results are expensive to produce—and when the application can tolerate the resulting freshness tradeoff. AWS’s caching guidance identifies these workload characteristics as potential reasons to cache.
Start by finding the slow, repeated reads rather than assuming the database itself is the culprit. A cache is a poor substitute for fixing an unindexed query, an inefficient query plan, excessive writes, or an unsuitable data model. First establish that the same data or result is requested repeatedly, and determine how stale it can safely be.
Choose a caching pattern that fits the reads
Cache-aside for selectively repeated data
With cache-aside, the application checks the cache on a read. If the value is absent, it reads from the primary database, stores the result in the cache, and returns it. This keeps the cache focused on data that users actually request, but the first request after a miss still incurs the database read and the cache-fill work. AWS describes this pattern as lazy loading.
#1 Best Overall
Write-through for data that must be refreshed on writes
With write-through, the application updates the primary store and updates the cache as part of its write path. This can make a later read more likely to find a current cached value. The tradeoff is that data may be cached even if nobody reads it, while writes add cache work and can cause churn. AWS suggests combining write-through with lazy loading when appropriate: update known cached entries on writes, while loading other values on demand. See AWS’s caching-strategy guidance and its write-through discussion.
Query-result caching for repeat SQL results
If the same expensive SQL result is requested repeatedly, query-result caching can target that work directly. AWS documents a JDBC caching plugin for selected Java queries against PostgreSQL, MySQL, or MariaDB. It requires an ElastiCache for Valkey or Redis OSS cache and the plugin’s documented dependencies; it is not a generic switch that caches every database query. Check the AWS query-caching documentation for supported setup details.
Decide where the cache should live
- Local or client-side cache: Reads avoid a remote cache network hop, and a client may be able to serve cached data during a backend disruption. Separate clients can hold duplicate values and disagree about freshness.
- Remote shared cache: Multiple clients can share entries, and cache storage can scale separately. Each access adds a network hop, so the benefit depends on the work avoided outweighing that cost.
- Local plus remote tiers: A local cache can sit in front of a shared cache, but each tier adds more freshness, invalidation, and failure behavior to manage.
Compare options on repeated-read potential, acceptable staleness, end-to-end latency (including cache hops and cold misses), memory and eviction behavior, invalidation and recovery complexity, and measured changes in database load and application tail latency. AWS’s performance guidance also emphasizes cache behavior as part of the broader data-access design.
Set freshness rules before caching
There is no universally correct time-to-live (TTL). Choose expiration based on how often the source data changes and how harmful an outdated value would be. A slowly changing reference value may tolerate a longer TTL than a frequently changing field. AWS recommends considering both change rate and staleness risk.
Rank #3
- [Color] PCB color may vary (black or green) depending on production batch. Quality and performance remain consistent across all Timetec products.
- DDR3L / DDR3 1600MHz PC3L-12800 / PC3-12800 240-Pin Unbuffered Non-ECC 1.35V / 1.5V CL11 Dual Rank 2Rx8 based 512x8
- Module Size: 16GB KIT(2x8GB Modules) Package: 2x8GB ; JEDEC standard 1.35V, this is a dual voltage piece and can operate at 1.35V or 1.5V
- For DDR3 Desktop Compatible with Intel and AMD CPU, Not for Laptop
- Guaranteed Lifetime warranty from Purchase Date and Free technical support based on United States
For values changed by your application, explicit invalidation or write-through can keep cache entries aligned with writes—but only if every relevant write path participates. A TTL can limit how long an entry survives if an invalidation is missed. AWS recommends TTLs for cache keys except those maintained through write-through. Read AWS’s expiration guidance.
Query caching has a stricter boundary when consistency matters. AWS warns: “Query caching is not recommended for queries where strong consistency is required, or for queries inside multi-statement transactions that require read-after-write consistency.” See the query-caching limitations.
Rank #4
- Store more, compute faster, and do it confidently with the proven reliability of BarraCuda internal hard drives
- Build a powerhouse gaming computer or desktop setup with a variety of capacities and form factors
- The go to SATA hard drive solution for nearly every PC application from music to video to photo editing to PC gaming
- Confidently rely on internal hard drive technology backed by 20 years of innovation; Max sustained transfer rate OD(MB/s): 190 MB/s
- Migrate and clone data from old drives with ease using our free Seagate DiscWizard software tool
Protect the database from cache misses and outages
Prevent stampedes when popular keys expire
If a popular key expires while many requests arrive together, those requests can all miss and hit the database at once. This cache stampede can concentrate refill traffic precisely when the database is already under pressure. Jittering expiration times helps avoid many keys expiring simultaneously. For hot keys, consider single-flight coordination or locking so only one request performs the refill while others wait, or use an early-refresh strategy. Redis documents lock-based and probabilistic early-refresh approaches, while AWS discusses expiration jitter.
Keep a recovery path to the source of truth
In cache-aside, the cache is an acceleration layer; the backing store remains the source of truth. Plan what the application does when entries are evicted, the cache restarts, or the cache service is unavailable. A fallback to the database preserves access only if the database can handle the resulting load; uncontrolled fallback can turn a cache outage into a database incident. AWS’s caching guidance covers cache monitoring and workload fit.
Best Value
Measure whether the cache is actually helping
Record a baseline before introducing caching, then compare under a comparable workload. Track:
- Cache hit rate and misses, including which keys or request types miss.
- Database query volume and CPU use.
- Application P95 and P99 latency, including cold-miss behavior.
- Evictions, cache memory use, and the impact of expiration or refill bursts.
AWS Well-Architected guidance gives 80% or higher as a cache-hit-rate monitoring goal. Treat that as a starting benchmark in that guidance, not a universal pass/fail threshold: a lower rate may signal an undersized cache, or that the workload does not benefit from caching. See AWS’s monitoring guidance. A hit-rate increase alone does not prove faster application responses; confirm the change in database load and P95/P99 latency on your workload.
A practical decision sequence
- Identify the hot reads: Use query and application metrics to find repeated requests and the expensive queries or results they share.
- Define correctness limits: Decide which values can be stale, for how long, and whether read-after-write or strong consistency is required.
- Pick a narrow pattern: Use cache-aside for data loaded on demand, write-through for values that must be updated along with writes, or query-result caching when the same supported SQL result is repeated.
- Plan expiration and refill: Set TTLs according to data volatility and stale-data risk; account for missed invalidations and coordinate refreshes for popular keys.
- Test failure behavior: Check how the application handles cache eviction, restart, and unavailability, including the load a database fallback would create.
- Compare against the baseline: Keep the cache only if measured database pressure or application latency improves without violating freshness requirements.
Caching is not just a database setting: the application’s data flows and consistency rules determine whether it is correct. A qualitative study of ten software projects found that application-level caching depended on project-specific details and could be handled ad hoc, making it error-prone; that finding describes those projects, not a general failure rate. Read the study.
Do these 3 things before closing this tab:
1Clear out junk files and repair common Windows errors2Fix the driver behind crashes, sound loss and screen glitches3Repair Windows errors before they cause bigger problemsQuick Recap
Product prices and availability are accurate as of the date/time indicated and are subject to change. Any price and availability information displayed on Amazon at the time of purchase will apply.
