Best fit
Large Redis-style datasets where DRAM, replicas, and shards dominate cost.
EloqKV
Scale Redis-Compatible Data Without Scaling DRAM
EloqKV uses NVMe-backed storage to keep large datasets fast at much lower cost.
50x+
lower infra cost (1TB example)
<5ms
P99.99 latency on NVMe
Redis/Valkey
compatible API
Trusted by teams running data at scale
What is EloqKV?
A Redis-compatible database for large stateful workloads
EloqKV is a Redis- and Valkey-compatible database that serves datasets far larger than RAM — at cache speed, with full ACID transactions.
Datasets beyond RAMRedis/Valkey compatibleStable P99.99 latency
Common uses
Session stores, user profiles, shopping carts, feature stores, and leaderboards — where the dataset dwarfs the hot working set.
DRAM to NVMe economics
The cost changes when capacity leaves DRAM
In the 1TB / 100K QPS / HA example below, EloqKV runs at about $650/mo versus about $33,120/mo on ElastiCache — more than 50× lower infrastructure cost.
1TB Redis-style workload at 100K QPS
| Option | Monthly infrastructure | Relative cost | Data placement |
|---|---|---|---|
| EloqKV | $650/mo | 1x baseline | 2 x i4i.2xlarge, each with 8 vCPU and 1 x 1875 NVMe SSD |
| ElastiCache for Redis OSS | $33,120/mo | 51.0x vs EloqKV | 26 x cache.r7g.4xlarge (13 shards x primary+replica) |
| Redis Cloud Pro | $35,018/mo | 53.9x vs EloqKV | Official Redis calculator estimate: 82 x Large shards for 2TB effective HA memory |
Assumptions: 1TB logical dataset, 100K QPS, P99.99-under-5ms latency target, HA included, us-east-1, and 730 hours/month. EloqKV uses EC2 1-year Standard Reserved, No Upfront pricing. Excludes transfer, snapshots, support, and discounts. Pricing reviewed June 2026. Calculate savings for your workload.
Predictable tail latency
P99.99 under 5ms on NVMe
The tail-latency spikes that break other SSD-backed caches come from LSM compaction stalls and write amplification. EloqKV's storage engine drops the LSM design entirely — one disk access per read, no background compaction — so P99.99 stays flat. A 2TB dataset holds P99.99 under 5ms on a single NVMe node, work that would otherwise need a ~20-node all-RAM Redis cluster.
<5ms
P99.99 at 2TB on NVMe
1 IOP
per read — no LSM levels
2TB
stable on a single node
Beyond cache
A transactional database behind a Redis API
EloqKV is full ACID with distributed transactions, so it covers workloads a cache cannot — without giving up the Redis interface.
Distributed ACID
Cross-shard MULTI/EXEC with consistent reads and writes — not just single-key atomics.
Ditch the Duo
Retire the MySQL + Redis combo and its cache-invalidation logic. One system is the source of truth and serves it at cache speed.
Durability optional
Run cache-style for raw speed, or enable WAL-backed recovery when the workload needs durable state.
Redis & Valkey wire-compatible. Point your existing clients, libraries, and tooling at EloqKV unchanged.
How it compares
Every Redis alternative trades something away
EloqKV's goal is to trade nothing — capacity, tail latency, and transactions at once.
| Capability | EloqKVRedis API on NVMe | In-memory cacheRedis · Valkey · ElastiCache | SSD / tiered KVe.g. KVRocks |
|---|---|---|---|
| Serves datasets beyond RAM | ✓ | ✗capped by DRAM | ✓ |
| P99.99 below 5ms under load | ✓non-LSM engine | ✓all-DRAM | ✗LSM compaction jitter |
| Distributed ACID transactions | ✓cross-shard MULTI/EXEC | ✗single-shard only | ✗limited |
| Infra cost, 1TB + HA | 1× · $650/mo | ~51× | Lower DRAMbut unstable tail |
| Redis / Valkey wire API | ✓ | ✓ | ✓ |
P99.99 vs KVRocks measured on a 2TB on-disk dataset at 100K QPS. See the benchmark.
FAQ
Common Redis migration questions
What is EloqKV?
EloqKV is a Redis-compatible key-value database for large datasets that are too costly to keep entirely in DRAM. It uses NVMe-backed storage with optional persistence. Read the EloqKV docs.
How does EloqKV reduce Redis infrastructure cost?
Redis cost scales with memory and replicas. EloqKV serves large datasets from NVMe and object storage, so capacity planning starts from total dataset size and HA needs. See the cost breakdown.
How does EloqKV use NVMe without high tail latency?
EloqKV combines NVMe-aware storage with asynchronous execution to keep P99.99 latency below 5 milliseconds, measured on a 2TB dataset at 100K QPS. Results depend on hardware, key size, access pattern, peak QPS, and durability settings. Review the latency benchmark.
Does EloqKV require durable persistence for every workload?
No. Run cache-style, or enable WAL-backed persistence when recovery is required. Learn about durability.
Is EloqKV a drop-in Redis or Valkey replacement?
EloqKV is Redis and Valkey compatible, but teams should validate commands, cluster behavior, latency, and operations before migration. Check command compatibility.
How do teams migrate from Redis to EloqKV?
Mirror data, validate reads and commands, then move writes after the target deployment is caught up and monitored. Read the migration guide.
Evaluate EloqKV