PROVIDER - Cyber resilience platform

Rubrik

Rubrik folds backup, anomaly detection, and recovery orchestration into one platform and replaces per-job configuration with SLA domains. Its pitch is bringing business services back in dependency order after an attack, and licensing must be quoted.

Backup & Migration United States 2026.11.20
4.2pts

Detailed Notes

Pros 5 items

An SLA domain expresses backup frequency, retention, and replication targets in 1 policy, and edits propagate to every object attached to it
The underlying file system delivers WORM semantics, 1 write and many reads, so no account including an administrator can delete or rewrite a snapshot inside its retention window
Anomaly detection reads metadata change patterns for signs of mass encryption or deletion and maps the event timeline to 1 concrete recoverable snapshot point
Orchestrated recovery brings up virtual machines, databases, and files in dependency order, turning a stack of manual steps into 1 task
One platform covers 3 data locations - on-premises data centres, public cloud, and SaaS - with Microsoft 365 and similar data under the same retention policy

Cons 4 items

Licensing is quoted only with no published unit price, so cost estimation needs several rounds of quotes before a fair comparison
The primary delivery model is an appliance, so data centres must budget rack space and power and the pure-software or pure-cloud path is narrower
The SLA domain abstraction is high level, and teams new to it tend to create too many narrow policies, inflating copies and storage cost
There is no local data centre or support team in mainland China, so console latency and data residency need separate evaluation

Service Overview

Rubrik is a data management vendor headquartered in California, and within Backup & Migration it takes the platform route: rather than positioning as plain backup software, it bundles backup, anomaly detection, access control, and recovery orchestration into a data security platform, sold primarily as Rubrik Security Cloud.

The deployment shape is an appliance plus a cloud control plane. Backup data lands on local appliances or the customer's own cloud storage, while the console handles policy distribution, threat detection, and recovery orchestration, with metadata synchronised over an encrypted channel. The benefit is that the restore path does not depend on the public internet: even when the console is unreachable, the local appliance can still restore on its own.

Its most discussed design is the SLA domain. Traditional backup configures frequency and retention per job, and drift is inevitable at scale. An SLA domain writes "how often, how long, and where to replicate" once, objects inherit it on attachment, and editing the policy changes everything beneath it. That abstraction drops policy count from hundreds of jobs to a dozen domains and makes it far easier to explain in an audit why a given dataset is retained for a given period.

Key Strengths

  • Policy as compliance: an SLA domain holds frequency, retention, and replication targets, decoupled from objects, so audits can show the policy-to-object-to-copy mapping instead of job logs
  • Immutability as the base layer: WORM semantics live in the file system rather than an application toggle, so snapshots cannot be deleted inside retention even if administrator credentials leak
  • Detection feeds recovery: alerts do not stop at notification; suspicious events map to specific snapshots, and orchestrated recovery can restore the pre-attack state once confirmed
  • One framework across locations: on-premises, public cloud, and SaaS data follow the same policy model, removing the split where virtual machines are recoverable while mailboxes are not
  • Orchestration removes manual steps: multiple virtual machines and dependent databases restore in order, drills become repeatable tasks, and RTO assessment gains a reproducible baseline

Product Ecosystem

Platform capabilities

Capabilities stack in three layers. The first is protection: image-level snapshots, incremental forever, and cross-site replication, with copies landing on local appliances, customer cloud storage, or vendor-managed cloud storage. The second is security: immutable snapshots, role-based access, multi-factor authentication, and metadata-driven anomaly detection. The third is recovery: file, virtual machine, database, and application level restores, with orchestration executing them in dependency order.

The console adds a data risk view that gathers which objects have not been backed up successfully in the last 7 days, which snapshots are about to expire, and which objects hold sensitive data. The point is to move backup operations from "are the jobs green" to "does the data match expectations", which only works once object naming and tagging are cleaned up.

Integrations

  • Virtualization and databases: vSphere, Hyper-V, and Nutanix AHV, plus application-consistent backup for SQL Server, Oracle, and PostgreSQL
  • Cloud and SaaS: AWS, Azure, and Google Cloud object storage as copy targets, with Microsoft 365 and Salesforce data under unified retention
  • Security stack: SIEM and SOAR platforms can receive anomaly events over the API, turning "suspicious encryption detected" into a ticket or isolation action
  • Automation: a GraphQL API covers policy creation, object queries, snapshot triggers, and restore tasks, usable as a downstream step in an orchestration engine
  • Identity: integration with enterprise identity sources, splitting backup, restore, and delete rights by role

Typical scenarios

Two kinds of organizations concentrate here. One has written ransomware recovery into business continuity targets and needs the number for "how fast can core systems come back after an attack". The other has data spread across data centres, public cloud, and SaaS and wants one policy model for retention instead of three tool sets.

Limitations

  • Licensing is quote-only with no published unit price, so selection and budgeting usually require multiple rounds of quotes and a self-imposed comparison baseline
  • An appliance is the main delivery form, which means rack space and power planning, and there are fewer paths that start purely in software or purely in cloud
  • The SLA domain abstraction is high level, and teams tend to create too many granular policies early, inflating copies and storage spend
  • No local data centre or support team in mainland China, so console latency and data residency requirements need separate assessment

Use Cases

  • Enterprises with ransomware recovery as a hard target (★★★★★): immutable snapshots, anomaly detection, and orchestrated recovery close the loop, which suits finance and healthcare with explicit RTO commitments
  • Teams with data across multi-cloud and SaaS (★★★★☆): one SLA domain framework covers on-premises, cloud, and SaaS, though object naming and tagging must be cleaned up first
  • Operations teams already running virtualization (★★★★): VM-level backup and restore granularity fits existing management practice, and migration effort sits in policy modelling rather than tool switching
  • Budget-sensitive small teams (★★☆): quote-based pricing plus an appliance raises the entry cost, and lightweight software or usage-priced cloud backup is cheaper
  • Local deployment in mainland China (★★): no local nodes or Chinese-language support, and cross-border access plus compliance review stretch the implementation timeline

Pricing

Item Price Notes
Platform subscription Quoted, tiered by protected data volume or appliance scale No published unit price; request a quote through sales or a reseller, official quote governs
Cloud storage copies Billed by capacity Vendor-managed storage or your own object storage, the latter paying only storage and egress
Long-term archive Priced by capacity and retention Longer retention and more copy tiers make the unit price structure more complex
Implementation and migration One-off project quote Policy modelling, object inventory, and historical data onboarding are usually billed separately

The table shows billing dimensions only and excludes hardware expansion, network, and staffing. For total cost comparison, use cost per TB per year as a common denominator, since it absorbs appliance amortisation, copy tiers, and support together.

FAQ

How is an SLA domain different from a traditional backup job?

The direction of coupling differs. A job hard-codes object, frequency, and retention together, so many objects mean many duplicated jobs. An SLA domain extracts the policy, objects declare which domain they belong to, and edits apply in bulk. The cost is that policy design must be right the first time, or the model collapses into one domain per object.

Can an administrator delete an immutable snapshot?

Not inside the retention window. WORM semantics are enforced in the file system, delete requests are refused, and management credentials do not bypass it; outside retention, snapshots expire per policy. That is why the window must be set to the compliance requirement, since a short window provides no protection.

How noisy is anomaly detection?

It depends on baseline data. Newly onboarded objects lack behavioural history, so the first weeks produce more noise; once a few backup cycles have run, detection concentrates on encryption signatures, mass deletion, and unusual logins. Most teams record without alerting first and turn on automatic tickets after one full cycle.

How do teams in mainland China deploy it?

Typically with local appliances kept in country and the console accessed as needed, so data does not leave the jurisdiction, though cross-border console latency affects daily operations. Where residency rules apply, confirm appliance and cloud metadata locations before purchase.

Code Examples

Query SLA domains and protected objects with GraphQL

#!/usr/bin/env bash
set -euo pipefail

# Endpoint and token come from environment variables; RSC exposes GraphQL at /api/graphql
QUERY='{"query":"query($first:Int){ slaDomains(first:$first){ nodes { name id } } }","variables":{"first":20}}'

curl -s -X POST "https://${RUBRIK_HOST}/api/graphql" \
  -H "Content-Type: application/json" \
  -H "Authorization: Bearer ${RUBRIK_TOKEN}" \
  -d "${QUERY}" \
  | jq -r '.data.slaDomains.nodes[] | [.id, .name] | @tsv'

An empty result usually means a permission problem: the token role needs policy read access, not just event visibility.

Find objects without a recent snapshot

#!/usr/bin/env python3
# Flag objects with no new snapshot in the last 48 hours
import json
import os
import urllib.request
from datetime import datetime, timedelta, timezone

ENDPOINT = f"https://{os.environ['RUBRIK_HOST']}/api/graphql"
HEADERS = {
    "Content-Type": "application/json",
    "Authorization": f"Bearer {os.environ['RUBRIK_TOKEN']}",
}

QUERY = """
query($first: Int) {
  vsphereVmNewConnection(first: $first) {
    nodes { name id latestSnapshotDate }
  }
}
"""

def gql(payload: dict) -> dict:
    req = urllib.request.Request(
        ENDPOINT, data=json.dumps(payload).encode(), headers=HEADERS
    )
    with urllib.request.urlopen(req, timeout=30) as resp:
        return json.loads(resp.read())

nodes = gql({"query": QUERY, "variables": {"first": 100}})["data"]["vsphereVmNewConnection"]["nodes"]
cutoff = datetime.now(timezone.utc) - timedelta(hours=48)

for vm in nodes:
    last = vm.get("latestSnapshotDate")
    if not last:
        print("never protected:", vm["name"])
        continue
    if datetime.fromisoformat(last.replace("Z", "+00:00")) < cutoff:
        print("stale:", vm["name"], last)

Field names and pagination arguments change between platform versions, so run a __schema introspection against the official API docs before relying on them.

Community & Resources

Visit Rubrik Website → https://www.rubrik.com/
Visit Website ↗