Skip to main content
A realtime NoSQL DB for data-intensive apps.
“ScyllaDB has been a quiet, well-behaved database... We’re not having weekend-long
firefights, nor are we juggling nodes in the cluster to attempt to preserve uptime.”
- Bo Ingram, Discord
2026.2 Highlights
ScyllaDB
Introductions
Marcin Maliszkiewicz
+ Technical Lead, Senior Software Engineer
+ Dedicated to reliable systems at scale
Nadav Har'El
+ Distinguished Engineer
+ Passionate about massive-scale NoSQL
Tzach Livyatan
+ VP Strategy at ScyllaDB
+ NoSQL enthusiast
▪ NoSQL, OLTP Distributed NoSQL Database
▪ Founded by designers of KVM Hypervisor: Avi Kivity
and Dor Laor
▪ Resolves challenges of legacy NoSQL databases
• >5x higher throughput
• >20x lower latency
• >75% TCO savings
▪ DBaaS/Cloud, Enterprise
▪ Compatible with Apache Cassandra and AWS
DynamoDB
The Database Built for
Gamechangers
“ScyllaDB stands apart...It’s the rare product
that exceeds my expectations.”
– Martin Heller, InfoWorld contributing editor and reviewer
“For 99.9% of applications, ScyllaDB delivers all the
power a customer will ever need, on workloads that other
databases can’t touch – and at a fraction of the cost of
an in-memory solution.”
– Adrian Bridgewater, Forbes senior contributor
3
Seamless experiences
across content + devices
Corporate fleet
management
Real-time analytics 2,000,000 SKU -commerce
management
Video recommendation
management
Low-latency global sports
betting
Real time fraud detection for
6M daily transactions
Uber scale, mission critical
chat & messaging app
Network security threat
detection
Power ~50M X1 DVRs with
billions of reqs/day
Precision healthcare via
Edison AI
Inventory hub for retail
operations
Property listings and
updates
Cryptocurrency exchange
app
Geography-based
recommendations
Global operations- Avon,
Body Shop + more
Predictable performance for
on sale surges
GPS-based exercise
tracking
Serving dynamic live
streams at scale
Powering India's top
social media platform
Personalized
advertising to players
Distribution of game
assets in Unreal Engine
Real-time ML-driven
recommendations
Connecting millions of
people around the globe
4
+400 Gamechangers Leverage ScyllaDB
 Vector Search 
X Cloud and Tablets -
feature parity
 Reduce TCO
+ Incremental repair
+ Dictionary Compression
+ Super fast Scale-out and Scale-in
+ Improve Performance
+ New ARM Instances: I8g family
5
2026.1 Highlights recap
 DynamoDB API (Alternator)
+ DynamoDB Streams GA
+ Vector Search (ANN) for DynamoDB API
 Stability & Performance
+ Trie-indexed as default
+ Mitigate connections storm
 Cloud
+ Vector Search Integrations
+ Private Link
 Experimental Features
+ Strongly Consistent Tables
+ vNode-to-Tablet: Online migration
6
2026.2 Feature Summary
+ 2025.1 LTS (Long Term Support)
+ 2025.4
+ 2026.1 LTS
+ 2026.2 STS (Short Term Support)
https://docs.scylladb.com/stable/versioning/version-support.html
7
Supported release
Alternator
8
Alternator:
ScyllaDB's DynamoDB-compatible API
+ Native DynamoDB API
+ Superior price/cost ratio
+ Works as-is
+ No changes to table or queries
9
+ Run anywhere
+ Any cloud, VM, k8s, bare metal
+ Tables, Indexes, Streams
+ Types, drivers
DynamoDB-compatible (Alternator) Streams has graduated from experimental to GA.
It provides a stream of time-ordered item-level modifications in any Alternator table.
+ Feature parity: Consistent with DynamoDB Streams API.
+ Use cases: Replicating data between applications, triggering Lambda-style functions, or
updating caches.
10
Stream
Updates
Kinesis Data Streams
Custom Application
DynamoDB Compatible Streams
We will:
1. Run Alternator
2. Using a Python script using Amazon’s DynamoDB SDK (boto3),
+ Create a table “test_table”
+ Enable stream on this table
+ Continuously listen for new events on the stream, printing it to the output.
3. Using Amazon’s DynamoDB CLI, create a new item or modify an existing one and watch
the events being printed.
+ aws dynamodb put-item --endpoint-url "$ENDPOINT_URL" --table-name test_table --item '{"id": {"S": "hello"}, "val": {"N": "3"}}'
+ aws dynamodb put-item --endpoint-url "$ENDPOINT_URL" --table-name test_table --item '{"id": {"S": "hello"}, "val": {"N": "4"}}'
DynamoDB Compatible Streams - small demo
11
ScyllaDB extension to the DynamoDB-compatible API that enables approximate nearest
neighbor (ANN) search on numeric vectors stored as item attributes.
Alternator Vector Search
12
table = dynamodb.create_table(
TableName='my-table',
KeySchema=[{'AttributeName': 'id', 'KeyType': 'HASH'}],
AttributeDefinitions=[{'AttributeName': 'id', 'AttributeType': 'S'}],
BillingMode='PAY_PER_REQUEST',
VectorIndexes=[
{
'IndexName': 'embedding-index',
'VectorAttribute': {'AttributeName': 'embedding', 'Dimensions': 1536},
}
],
)
13
Alternator Vector Search
response = table.query(
IndexName='embedding-index',
Limit=10,
VectorSearch={
'QueryVector': {'L': [{'N': '0.1'}, {'N': '-0.3'}, {'N': '0.7'}, ...]},
},
)
Performing the ANN search:
Alternator Vector Search
14
15
DynamoDB announced last week (Aug. 5 announcement), their own version of vector search
API inside DynamoDB.
+ Similar in features to what Alternator already supports in ScyllaDB 2026.2.
+ But the specific API is different.
Amazon DynamoDB vector search
Stability
16
Improve Stability - Connections Storm Mitigation
17
Node restart can cause a connection storm, resulting in high latency of user requests.
Mitigation includes:
+ Coordinator side Connection throttling #22828
+ Unify Authentication and authorization cache JIRA: SCYLLADB-145
+ Improved password hashing #24924
+ Using a lower default service level #26411
+ Replica-Side Load Shedding. Prevents a "spiral of death" where overloaded nodes keep
falling. #21649
18
Improve Stability - Connections Storm Mitigation
19
Role Cache
Authentication Layer
Raft Write Logic CQL Read Layer
Effect of the Auth cache improve efficiency:
+ Single shard connection/disconnection scenario
Instructions/connection decreased by 7%
+ Heap Allocations/connection decreased by 55%
+ Tasks/conn decreased by 41%.
Unify Authentication and Authorization cache
20
2025.1.9 2026.2.0
Node Restart: p99 read latency
21
Node Restart: p99 write latency
2025.1.9 2026.2.0
Performance
22
https://www.scylladb.com/2026/06/30/trie-index-3x-more-throughput/
23
Trie-Based SSTable Index
Introduced in 2026.1, Trie-Based SSTable Index
(mt format) are now the default.
+ Smaller Index Footprint: Improves cache residency
and reduces or eliminates index-related I/O.
+ Lower I/O on Cache Misses: A compact, shallow
index requires fewer reads, especially for large
partitions and materialized view workloads.
+ Reduced CPU Overhead: A more efficient index
lowers CPU cost during read operations.
24
Trie-Based SSTable Index
Test case
Legacy
Throughput
Legacy P99
Latency
Trie
Throughput
Trie P99
Latency
Throughput
gain
Typical (~20% row cache) 130k ops/s 5.1 ms 170k ops/s 1.9 ms +31%
Key / Value 90k ops/s 5.2 ms 300k ops/s 3.6 ms +233%
Large Partitions 23k ops/s 7.7 ms 37k ops/s 4.6 ms +61%
Long shared clustering
key prefixes
22k ops/s 5.4 ms 38k ops/s 3.3 ms +73%
25
Trie-Based SSTable Index
ScyllaDB Cloud
26
2026.2 brings better integrations with the popular AI tool chain:
LangChain, LlamaIndex, Spring AI, Agno and more…
CREATE CUSTOM INDEX ON my_keyspace.documents (embedding)
USING 'StorageAttachedIndex'
WITH OPTIONS = {
'similarity_function': 'cosine',
'source_model': 'openai-text-embedding-3-small'
};
27
Vector Search Integration
Private Link
ScyllaDB Cloud clusters now can be exposed over Amazon Web Services Private Link. The customer creates
a private endpoint inside their VPC, and traffic flows privately to ScyllaDB Cloud through AWS backbone
networking. Without exposing neither VPC to one another.
28
Private Service Connect
Private Service Connect is a private connectivity service of Google Cloud networking that allows consumers
to access managed services (ScyllaDB Cloud Clusters) privately from inside their VPC network. The
customer creates a private endpoint inside their VPC, and traffic flows to ScyllaDB Cloud clusters through
GCP backbone networking. Without exposing the VPCs to one another.
29
+ Ready-to-use Oracle Cloud Infrastructure (OCI) images
+ ScyllaDB k8s Operator: Oracle Kubernetes Engine (OKE) support
Improved Oracle Cloud Integration
30
Experimental Features
31
High Performance string data
consistency with Raft JIRA: SCYLLADB-13
CREATE KEYSPACE my_keyspace
WITH consistency = ‘global’;
Fast, strong data consistency
https://scylladb-raft-example.netlify.app/
32
Experimental
vNode to Tablets Migration
33
CREATE KEYSPACE ks WITH
tablets = {'enabled': false};
CREATE KEYSPACE ks WITH
tablets = {'enabled': true};
$ build/dev/scylla nodetool migrate-to-tablets
start <keyspace> Prepare a keyspace for migration from vnodes to tablets
upgrade Set this node's intended storage mode to tablets
downgrade Set this node's intended storage mode back to vnodes
status <keyspace> Show migration status for a keyspace
finalize <keyspace> Finalize the migration for a keyspace
Real Time, low impact cluster migration from vNode to Tablets for production clusters
Experimental
34
vNode to Tablets Migration
$ build/dev/scylla nodetool migrate-to-tablets --help
Manage vnodes-to-tablets migration
Migrate a keyspace from vnodes to tablets.
Subcommands:
start <keyspace> Prepare a keyspace for migration from vnodes to tablets
upgrade Set this node's intended storage mode to tablets
downgrade Set this node's intended storage mode back to vnodes
status <keyspace> Show migration status for a keyspace
finalize <keyspace> Finalize the migration for a keyspace
Experimental
CQL Extensions
35
-- Add a new datacenter
ALTER KEYSPACE my_ks WITH replication = {
'class': 'NetworkTopologyStrategy',
'dc1': ['RAC1', 'RAC2'],
'dc2': ['RAC3', 'RAC4', 'RAC5']
};
-- Remove a datacenter
ALTER KEYSPACE my_ks WITH replication = {
'class': 'NetworkTopologyStrategy',
'dc1': ['RAC1', 'RACK2'],
'dc2': []
};
Rack Aware Replication
36
CQL: Per-Row Time-to-Live (TTL)
+ New, DynamoDB like TTL feature
+ Deleted (TTled) rows update CDC and Vector Search
CREATE TABLE tab (
id int PRIMARY KEY,
t text,
expiration timestamp TTL
);
https://docs.scylladb.com/manual/branch-2026.2/cql/cql-extensions.html#per-row-ttl
37
CQL: Per-Row Time-to-Live (TTL)
+ Eventual, DynamoDB like TTL feature
+ Deleted (TTled) rows update CDC and Vector Search
+ alternator_ttl_period_in_seconds (default 24h)
CREATE TABLE tab (
id int PRIMARY KEY,
t text,
expiration timestamp TTL
);
https://docs.scylladb.com/manual/branch-2026.2/cql/cql-extensions.html#per-row-ttl
38
 DynamoDB API (Alternator)
+ DynamoDB Streams GA
+ Vector Search (ANN) for DynamoDB API
 Stability & Performance
+ Trie-indexed as default
+ Mitigate connections storm
 Cloud
+ Vector Search Integrations
+ Private Link
 Experimental Features
+ Strongly Consistent Tables
+ vNode-to-Tablet: Online migration
39
2026.2 Feature Summary
Thank you
for joining us today.
@scylladb scylladb/
slack.scylladb.com
@scylladb company/scylladb/
scylladb/