Amazon DynamoDB database services for production teams.
JusDB provides Amazon DynamoDB services for scoped projects, ongoing remote DBA coverage, and production support. Our engineers work in your AWS account on key and index design, capacity planning for on-demand and provisioned tables, global tables, Streams pipelines, and migrations. 24x7 on-call is available under your service agreement.
Or request a free audit of one database instance. Already in an incident? See how incidents are handled.
- NDA before access
- Read-only by default
- Changes measured against your baseline
DynamoDB capacity and table class
Pick the setting your table uses
On-demand capacity
Default modeAWS's default mode, billed per request. A new table sustains up to 4,000 writes and 12,000 reads per second, and instantly serves up to double its previous peak.
Growth past double the previous peak within 30 minutes can throttle, so we raise warm throughput before a launch.
Plan capacity before a launchProvisioned capacity
For steady trafficYou set read and write capacity units, billed hourly whether used or not, and auto scaling moves them toward a target.
Auto scaling acts after two minutes above target, plus alarm delay, so a sharp spike can throttle first.
Review capacity mode and costStandard-IA table class
Storage-heavy tablesAn alternative to the default Standard class, with lower storage costs and higher read and write costs. Performance, durability, and availability are the same.
It can lower the total once storage costs more than half of what reads and writes cost on Standard, and a table allows two class changes per 30 days.
Table class guideDetails from the Amazon DynamoDB Developer Guide on capacity modes and table classes, checked 28 October 2026.
- Platforms
- Amazon DynamoDB in your AWS accounts, with DAX, DynamoDB Streams, AWS Lambda, and Kinesis Data Streams
- High availability
- Copies in several Availability Zones by default, global tables across Regions, and point-in-time recovery
- Incidents
- Initial-response targets are set in your service agreement. See the targets
Symptoms
Eight DynamoDB problems we are asked to fix
Each one has a cause we can measure. The scoping call starts from the symptom you see and the tables, indexes, and capacity mode involved.
Writes throttle while the table shows spare capacity because a few key values push one partition past 1,000 write units per second, a limit adaptive capacity cannot raise.
A status or date index throttles table writes because its key has few values, and DynamoDB holds back table writes when the index cannot keep up (back-pressure).
An on-demand table throttles on launch day because traffic passed double its previous peak within 30 minutes, with no warm throughput raised beforehand.
Auto scaling trails sharp spikes because it acts after two minutes above target, so provisioned tables throttle while SDK retries add requests.
A Streams consumer stalls behind one bad record because Lambda, by default, retries until the records expire, which can block the shard for up to a day.
Global tables overwrite concurrent updates because eventual consistency keeps the write with the latest timestamp for each item, so the other Region's change is lost.
Nobody has restored from point-in-time recovery so the team has not seen a restored table arrive without its auto scaling policies, alarms, Streams settings, and TTL.
Read costs are higher than traffic explains because Scans and filter expressions read items the application discards, and strongly consistent reads use twice the read units.
If one of these is costing your team time, a scoping call is the place to start.
Book a 30-min scoping call (opens Google Calendar in a new tab)Hot partitions
How a hot partition key throttles DynamoDB writes
DynamoDB places each item by a hash of its partition key, and each partition serves at most 3,000 read units and 1,000 write units per second. When writes concentrate on a few key values, their partition throttles while the table still has spare capacity.
One table, three partitions, three outcomes
Choose a state to see where the writes land.
Application writes
The order ID is the partition key, so the hash spreads writes across partitions.
Partition 1
A third of order IDs
Under limit
Partition 2
A third of order IDs
Under limit
Partition 3
A third of order IDs
Under limit
No key value draws more than a partition can serve, and burst and adaptive capacity cover short, uneven swings.
Application writes
Today's date is the partition key, so every new order lands on one partition.
Partition 1
Today's date and every new order
At the 1,000 WCU limit
Partition 2
Older dates, rarely written
Mostly idle
Partition 3
Older dates, rarely written
Mostly idle
Writes for today's date are throttled with the reason TableWriteKeyRangeThroughputExceeded while the table has capacity left, and each SDK retry hits the same partition.
Application writes
A suffix from 1 to 10, calculated from the order ID, spreads each day over ten key values.
Partition 1
Some of today's ten keys
Under limit
Partition 2
Some of today's ten keys
Under limit
Partition 3
The rest of today's keys
Under limit
A read of the whole day queries all ten keys and merges them. Sharding does not lift the partition limit: one item still takes at most 1,000 write units per second.
We design keys from your access patterns: high-cardinality partition keys, calculated shard suffixes for heavily written values, and the same skew check on index keys. Item size counts too, since a 20 KB item uses five read units per strongly consistent read.
We alarm on WriteKeyRangeThroughputThrottleEvents and keep CloudWatch Contributor Insights in throttled keys mode, so a hot key is named, not guessed. Before a launch we raise warm throughput and rehearse the peak, and a point-in-time restore, on a test table.
What we cover
Full-spectrum DynamoDB engineering
From one throttled key to a multi-Region design: CloudWatch metrics and your access patterns drive every change.
Partition key and secondary index design
Keys and indexes designed from your access patterns, so traffic spreads across partitions.
- An access-pattern table before any key is chosen
- Calculated shard suffixes for heavily written keys
- Index keys checked for skew, and projections for write and storage cost
RCU/WCU vs on-demand
Capacity mode chosen from your traffic shape and expected peaks.
- Traffic read from ConsumedReadCapacityUnits and ConsumedWriteCapacityUnits
- Auto scaling targets and limits for provisioned tables
- Maximum and warm throughput for on-demand tables
Global tables design
Replication chosen for the consistency you need, then tested by isolating a Region.
- MREC with last writer wins, or MRSC in three Regions
- One write Region per item where conflicts matter
- ReplicationLatency alarms and AWS FIS Region isolation tests
DAX caching strategy
Microsecond reads for hot keys, where eventually consistent data is acceptable.
- Item cache and query cache TTLs sized to your staleness limits
- Writes routed through DAX to keep the item cache current
- DAX or Amazon ElastiCache, decided per workload
Streams and Lambda pipelines
Pipelines where one bad record cannot stall a shard for long.
- BisectBatchOnFunctionError, MaximumRetryAttempts, and MaximumRecordAgeInSeconds set
- On-failure destinations in Amazon SQS, Amazon SNS, or Amazon S3
- IteratorAge alarms and idempotent handlers
Migration and cost audit
Moves into and out of DynamoDB, and a written review of where capacity goes.
- AWS DMS loads from relational databases or MongoDB
- Full or incremental exports to Amazon S3
- Scans, unused indexes, and table class checked against usage
Incident support
A critical (P1) page, answered within your agreement's target
JusDB keeps your recovery procedures (runbooks) written down, so an incident starts from a known plan instead of a blank page.
Initial-response targets
2 to 4 hours
Business hours
During covered business hours
15 minutes
24x7 on-call
P1 incidents
Production database down, primary unreachable, data corruption risk, or active failover event.
Severe latency spike, replication lag climbing rapidly, connection exhaustion, or degraded HA node.
Non-critical slow query regression, staging issue, backup warning, or replication check alert.
Configuration question, index advisory, planned schema review, or general operational request.
Our engineering team is in India; business-hours coverage is Mon–Fri 09:00–18:00 IST (UTC+5:30), and 24x7 on-call is available under your service agreement.
Initial-response targets, not resolution guarantees. The target for your engagement is set in your service agreement.
What happens on a P1 page
Page
You page the on-call engineer
Through the escalation path your service agreement names.
Triage
Root cause from throttling reasons and Contributor Insights
Throttled keys, index back-pressure, and stalled Streams consumers traced to the table, index, or function behind them.
Mitigate
Stabilize first, from the runbook
Raise provisioned capacity, move hot reads to eventually consistent reads, or cap a stuck consumer's retries, then fix the cause once production is steady.
After
A review after every P1
What happened, why, and the change that stops it happening again.
Before and between incidents
- Your setup, written down firstDuring onboarding we record your tables, keys, indexes, capacity modes, Streams consumers, and backups, so an incident starts with context.
- One escalation path, in writingYour service agreement names how to reach the on-call engineer, in which hours, and the AWS roles our engineers may use.
- Reviews between incidentsWe review throttling, capacity, index cost, Streams lag, and restore readiness, and report open risks with next steps.
- Service changes trackedWe track AWS changes to DynamoDB defaults, limits, and metrics, and plan the work they require.
Process
How a DynamoDB engagement runs, step by step
Written findings come before any change. Your team approves production changes, except routine actions you have pre-approved.
Scoping call
Your tables, capacity modes, Regions, and what you need covered.
Read-only review under NDA
Our engineers use an IAM role limited to the read permissions the review needs, then read your table settings, CloudWatch metrics, and Streams consumers.
A written report before any change
Each finding comes with the change we propose, how we will measure it, and a way to undo it (a rollback plan).
Changes, then handover
We test risky changes on a copy of the table, apply them in the windows you approve, and hand over updated operating instructions (runbooks).
Step one is a 30-minute scoping call.
Book a 30-min scoping call (opens Google Calendar in a new tab)Ways to work
Choose the DynamoDB service your team needs
Start with the responsibility you need covered. Each engagement has its own scope, deliverables, access requirements, and acceptance criteria.
Projects
Scoped DynamoDB projects
A defined piece of work with written findings, hands-on changes, and validation against your baseline.
- DynamoDB consultingKey design, capacity, global tables, and migration plans, scoped as one project.
- High-traffic preparednessLoad tests and capacity plans before a launch or a seasonal peak.
- Migration with AWS DMSLoads from a relational database or MongoDB into DynamoDB.
- Backup and disaster recoveryPoint-in-time recovery, restore drills, and cross-Region restores.
Ongoing
Ongoing coverage and support
JusDB monitors the databases, carries out the agreed maintenance, and responds to incidents during the coverage the proposal names. The team keeps a change record, recovery procedures, and a review after every P1.
Where your DynamoDB runs, and the tools around it
Your team and ours
What fully managed DynamoDB still leaves to your team
Managed cloud databases automate much of the infrastructure layer, but your team still owns query performance, capacity, cost, and incident decisions.
| Question | On-demand capacity | Provisioned capacity |
|---|---|---|
| Capacity planning | AWS scales with traffic. You can cap throughput and pre-warm before a peak. | Your team sets capacity units and auto scaling targets and limits. |
| Sudden spikes | Served instantly up to double the previous peak; growth past double within 30 minutes can throttle. | Short spikes can use up to 300 seconds of unused capacity (burst capacity), then auto scaling must catch up. |
| Secondary indexes | Indexes scale with traffic, but a skewed index key can still throttle table writes. | Your team sets each index's capacity, and an index short of write capacity throttles table writes. |
| Switching modes | You can switch to provisioned capacity at any time. | You can switch to on-demand up to four times in a rolling 24-hour window. |
| Runaway cost | Billed per request, so a runaway job raises the bill unless throughput is capped. | Billed for the capacity you set, used or not. |
Your team keeps control
- Set business priorities, application requirements, and release plans
- Approve production changes, except routine actions you have pre-approved
- Own your cloud accounts, access policies, and vendor relationships
- Decide when maintenance can happen and how quickly service must recover
- Move the work in-house or to another provider after your agreement's notice period, with your data returned or deleted and the transition help your agreement sets out (Terms, section 8)
JusDB handles the agreed database work
- Monitor database health, investigate alerts, and involve the right people
- Review query performance, replication, capacity, and backup health
- Carry out approved changes and update the operating instructions they affect
- Report completed work, open risks, and recommended next steps
Weighing on-demand against provisioned capacity? We compare both against your workload on a scoping call.
Book a 30-min scoping call (opens Google Calendar in a new tab)Further reading
DynamoDB guides and two comparisons
Blog guides on table design, capacity and cost, Streams, and choosing a database, plus two comparisons.
Guide
DynamoDB single-table design and indexes
Access-pattern tables, key conditions before filters, index write cost, and transactions.
Guide
DynamoDB capacity modes and cost
On-demand against provisioned capacity, table classes, index cost, and warm throughput.
Guide
DynamoDB Streams and Lambda
Stream view types, event source mappings, bisect on error, and failure destinations.
Guide
Choosing between DynamoDB and MongoDB
A selection process built on your real operations, consistency needs, and test data.
Comparison
MongoDB vs DynamoDB
Document queries against key-based access, the 400 KB item limit, and migration paths.
Comparison
DynamoDB vs Cosmos DB
Capacity units against request units, and global tables against multi-region writes.
FAQ
Common DynamoDB questions
What buyers ask about DynamoDB keys, capacity modes, global tables, caching, Streams, and migrations.
Still have a question?
Ask it on a scoping call. No production access needed.
Book a 30-min scoping call (opens Google Calendar in a new tab)What is Amazon DynamoDB and where does it fit?
JusDB recommends Amazon DynamoDB for workloads whose reads and writes are known in advance and fit key lookups and key-range queries. DynamoDB is AWS's serverless, fully managed NoSQL database. Ad hoc queries, joins, or many changing filters usually fit MongoDB or PostgreSQL better.
Partition key design: why does it matter so much?
JusDB treats the partition key as the DynamoDB decision that is hardest to change. Each partition serves at most 3,000 read and 1,000 write units per second, so a skewed key throttles while the table has spare capacity. Changing the key later means copying every item to a new table.
On-demand vs provisioned capacity: which fits?
JusDB picks the DynamoDB capacity mode from your traffic shape. On-demand, AWS's default, bills per request and instantly serves up to double a table's previous peak; growth past double within 30 minutes can throttle. Provisioned capacity bills for the units you set, used or not, and suits steady, forecastable traffic.
Global Tables: when is multi-region worth it?
JusDB recommends DynamoDB global tables when users in several Regions need local reads and writes, or you need a Region to fail over to. The default, eventually consistent mode (MREC) replicates asynchronously, typically within a second, and keeps the last write when updates collide. Strong consistency (MRSC) needs exactly three Regions and adds write latency.
DAX caching: when does it pay off?
JusDB recommends DynamoDB Accelerator (DAX) for read-heavy workloads that reread the same keys and accept eventually consistent data. DAX cuts those reads from milliseconds to microseconds, but it does not cache strongly consistent reads, and writes that bypass it leave cached items stale until their TTL expires.
DynamoDB Streams + Lambda: what is the pattern?
JusDB builds DynamoDB Streams pipelines in which an AWS Lambda function passes each item change to search, analytics, or other services. A stream keeps changes for up to 24 hours, in order per item. By default, one failing record can block its shard for up to a day, so we set MaximumRetryAttempts, BisectBatchOnFunctionError, and an on-failure destination.
When do you migrate from DynamoDB to MongoDB Atlas?
JusDB recommends moving from DynamoDB to MongoDB Atlas when keys and a few indexes can no longer serve the product's queries, such as ad hoc filters and aggregation pipelines. Data leaves DynamoDB through an export to Amazon S3, since AWS DMS does not support DynamoDB as a source. The application rewrite is usually the larger cost.
Next step
Talk to a DynamoDB engineer about your databases
Tell us what is slow, fragile, or due to change. The first 30-minute call covers your databases, current ownership, and required coverage. No production access is needed. Or request a free audit of one database instance, subject to availability: mutual NDA before read-only access, a prioritized report within one week after access is ready, and no obligation.
NDA before access, read-only by default, and written scope first
Compare with other databases
If you are weighing alternatives, these are the databases most often compared with this one.
MongoDB
A document database with ad hoc queries and secondary indexes, compared with DynamoDB when access patterns are still changing.
Azure Cosmos DB
The Azure counterpart, with serverless and provisioned capacity and global distribution. Multi-cloud teams compare capacity models and consistency levels.
ScyllaDB
ScyllaDB's Alternator API is compatible with DynamoDB, so teams evaluate it for DynamoDB-style workloads on their own clusters or other clouds.
Aerospike
A key-value database with its index in memory and data on flash, compared with DynamoDB for steady, high-throughput workloads where cost at scale matters.