Global infrastructure
Our own autonomous system, our own hardware, our own scrubbing. Compute regions across India, Singapore and the United States connected by a 500+ Gbps backbone with Tier-1 and IX peering, real-time NetFlow intelligence and always-on DDoS filtering.
Five regions, one network.
Every region runs the same stack: Compute Fabric KVM nodes with NVMe RAID, a regional pxcache collector that streams hypervisor telemetry to the core every 15 seconds, and an anycast edge that terminates HTTPS close to you. Health dots below are computed from that telemetry; IPv4 stock is read from the address pool live.
Noida, India
- Nodes
- 4 active · 4 online
- Capacity
- 368 cores · 2,896 GB RAM
- Uplink
- 100 Gbps · DDoS always-on
- Templates
- Ubuntu 22.04 LTS, Ubuntu 24.04 LTS, Windows Server 2022, Windows Server 2025, Windows Server 2019
- IPv4 stock
- 281 addresses
- Latency
- Delhi NCR < 5 ms · Mumbai ~28 ms · Singapore ~80 ms
Tier III facility in the NCR corridor; dual power feeds, N+1 cooling, direct Delhi IX reach.
Mumbai, India
- Nodes
- 9 active · 9 online
- Capacity
- 840 cores · 6,580 GB RAM
- Uplink
- 100 Gbps · DDoS always-on
- Templates
- Ubuntu 22.04 LTS, Ubuntu 24.04 LTS, Windows Server 2022, Windows Server 2025, Windows Server 2019
- IPv4 stock
- 2 addresses
- Latency
- Mumbai < 3 ms · Chennai ~22 ms · Singapore ~55 ms
Flagship region — 100 Gbps uplinks, Mumbai IX peering and the primary scrubbing centre.
Chennai, India
- Nodes
- 1 active · 1 online
- Capacity
- 104 cores · 125 GB RAM
- Uplink
- 10 Gbps · DDoS always-on
- Templates
- Windows Server 2022
- IPv4 stock
- waitlist
- Latency
- Chennai < 4 ms · Bengaluru ~12 ms · Singapore ~40 ms
Southern India coverage with submarine-cable proximity for Southeast Asian routes.
Singapore
- Nodes
- 0 active · 0 online
- Capacity
- 0 cores · 0 GB RAM
- Uplink
- 10 Gbps · DDoS always-on
- Templates
- being verified
- IPv4 stock
- waitlist
- Latency
- Singapore < 2 ms · Jakarta ~18 ms · Mumbai ~55 ms
Regional hub for ASEAN traffic; direct paths to major Singapore exchanges and Tier-1 carriers.
United States
- Nodes
- 1 active · 1 online
- Capacity
- 88 cores · 125 GB RAM
- Uplink
- 10 Gbps · DDoS always-on
- Templates
- Ubuntu 22.04 LTS, Windows Server 2022
- IPv4 stock
- 248 addresses
- Latency
- US West < 10 ms · US East ~65 ms · Mumbai ~230 ms
West-coast presence for North American users and low-latency transpacific routes to Asia.
AS152565 — JOYSVC-AS-IN
Allocated by APNIC, registered to Joy Services, India. Content network with 394 known peers (#4 in India), 5 upstreams and an 18-network customer cone (31 in total). All originated prefixes carry RPKI ROAs.
- aut-num
- AS152565
- as-name
- JOYSVC-AS-IN
- descr
- JOY SERVICES
- country
- IN
- status
- Active · allocated under APNIC
- type
- Content · Server hosting · Direct feed
- mnt-by
- MAINT-IN-JOYSVC, MAINT-IN-IRINN
- mnt-irt
- IRT-JOYSVC-IN
- IPv4 space
- 31 × /24 across 13 prefixes
- IPv6 space
- 32,773 × /48 across 6 prefixes
- AS-SETs
- Member of 20 across RIPE, APNIC, RADB and ARIN
Announced prefixes
| IPv4 | Holder |
|---|---|
| 31.42.125.0/24 | CloudOnFire |
| 38.76.192.0/20 | FOJ IP Technology Limited |
| 46.232.234.0/24 | Joy Services |
| 103.151.198.0/24 | St Broadband Cable Service Pvt Ltd |
| 103.165.11.0/24 | Orange Waves Networks Private Limited |
| 138.252.100.0/24 | JOY SERVICES |
| 138.252.101.0/24 | JOY SERVICES |
| 157.20.240.0/23 | JOY SERVICES |
| 161.248.162.0/24 | Joy Data Services |
| 161.248.163.0/24 | Joy Data Services |
| 163.61.38.0/23 | Rapid Fibernet Broadband |
| 165.101.250.0/23 | Cyflame |
| 201.7.16.0/24 | Joy Services |
| IPv6 | Holder |
|---|---|
| 2001:df4:940::/48 | JOY SERVICES |
| 2401:e920::/33 | Joy Data Services |
| 2401:e920:c7f7::/48 | Joy Data Services |
| 2401:e920:fffa::/48 | Joy Data Services |
| 2401:e920:fffb::/48 | Joy Data Services |
| 2401:e920:fffc::/48 | PT Static IP |
Upstreams
| ASN | Carrier | Family |
|---|---|---|
| AS9583 | Sify Limited (IN) | IPv4 + IPv6 |
| AS133296 | Web Werks India Pvt. Ltd. (IN) | IPv4 + IPv6 |
| AS398999 | Sucura Networks Inc (CA) | IPv4 |
| AS62255 | BiMajLink d.o.o. (SI) | IPv4 |
| AS18229 | CtrlS (IN) | IPv4 |
Downstream networks (customer cone)
Source: public BGP data for AS152565. Prefix holders reflect the registry object of each block; several are customer prefixes originated on their behalf under BYOIP agreements.
How traffic moves through Joy.
Joy NeuroMesh NetFlow → Tier-1 ISP → Last-mile partners → Joy Fabric → IX / Peering → CDN / Edge compute. Each hop is engineered for deterministic paths and predictable latency.
Anycast Edge
Nearest healthy PoP every time. Sub-second failover keeps latency low and sessions stable. Latency routing · Auto failover.
Tier-1 Peering
Direct paths via Tier-1 carriers & IXes cut hops and jitter for predictable performance. Major IX · Deterministic paths.
500+ Gbps Headroom
Backbone absorbs bursts without loss. Peaks feel flat and apps stay smooth. Loss-aware QoS · High-burst pipes.
NeuroMesh NetFlow
Real-time flow intelligence with end-to-end encrypted telemetry and early anomaly detection. Encrypted · Anomaly scoring.
Internet exchanges & transit
AS152565 peers at the major Indian exchanges in Mumbai and Delhi, in Singapore and on the US West Coast, alongside bilateral sessions with Tier-1 carriers. We prefer bilateral sessions above 1 Gbps and route-server sessions otherwise; the peering policy is open. Networks can buy IP transit at $500 / Gbps (95th percentile), DDoS-protected transit from $700 / Gbps, Virtual Connect over GRE from $200 and IX fabric ports from $150 / Gbps.
Peering & network servicesDDoS scrubbing
Every VPS, bare-metal server and transit customer sits behind always-on scrubbing. NeuroMesh NetFlow scores flow anomalies in real time, diverts affected prefixes with BGP and returns clean traffic over the backbone or GRE. Volumetric floods, amplification, SYN/ACK storms and L7 request floods are handled without customer action; WAF rules are available for HTTP origins behind the CDN.
Security & trust centerBuilt on hardware we own.
Tier III datacenters with redundant power and cooling. Compute nodes are Dell and Supermicro systems with Intel Xeon Gold and AMD EPYC processors, ECC memory, NVMe RAID and 10 / 25 / 100 G uplinks. Hypervisor APIs live on private VLANs and are reached only through the regional collector.
Compute nodes
Compute Fabric (KVM) with, host CPU passthrough, virtio-scsi single-queue with IO threads, NVMe-backed storage pools spread least-used-first.
Regional collectors
A pxcache node per region polls every hypervisor, pushes VM and node stats into Redis and executes clone / configure / boot jobs locally.
Storage
NVMe RAID for VM disks, separate backup targets for scheduled full-VM backups, and S3-compatible object storage (coming) for buckets and CDN origins.
Colocation
Per-U and full-rack colocation with metered power, cross-connects into the Joy fabric and 24/7 remote hands — see Bare Metal.
Network
AS152565 operates a 500+ Gbps backbone with Tier-1 upstreams and IX peering in Mumbai, Delhi, Singapore and the US West Coast. NeuroMesh NetFlow gives per-flow visibility for DDoS detection and traffic engineering, and every originated prefix carries an RPKI ROA.
Facilities
Tier III datacenters with redundant power and cooling. Compute nodes are Dell and Supermicro Xeon Gold and EPYC systems with NVMe RAID and 10 / 25 / 100 G uplinks, managed by the Compute Fabric and a regional collector that streams telemetry to the core every 15 seconds.
Talk to the people who run it.
Peering
peering@joy.services · open policy · send your ASN and locations
NOC
noc@joy.services · 24/7 · reachability, routing and scrubbing issues
Abuse
abuse@joy.services · include source IP, timestamps (UTC) and logs
Security
security@joy.services · responsible disclosure, acknowledged within 48 h