---
title: Choose a Node Family
---

import { IndianRupee } from 'lucide-react';
import { Cpu, Server, Shield, Award, Database, Zap, ShoppingBag, Layers } from 'react-feather';

# Choose a Node Family

Use this guide when the create-node flow asks you to choose a Category, Instance Type, and Plan.

A node family is not the same thing as a plan. The family describes the type of compute capacity you are choosing, such as cost-effective, compute-optimized, memory-optimized, dedicated compute, SQL-ready, or GPU-backed. The plan is the exact CPU, RAM, storage, IOPS, GPU memory, CUDA, or SQL version configuration shown by the portal.

The portal is the final source for currently available families, plans, OS versions, and prices. Use this page to understand what each family is for before you pick a plan.

<div style={{marginBottom:'32px',marginTop:'8px'}}>
  <div style={{display:'flex',gap:'8px',flexWrap:'wrap'}}>
    <a href="#e1-and-e1wc" style={{display:'inline-flex',alignItems:'center',gap:'6px',padding:'6px 14px',background:'var(--ifm-color-emphasis-100)',border:'1px solid var(--ifm-color-emphasis-300)',borderRadius:'20px',fontSize:'13px',color:'var(--ifm-color-emphasis-700)',textDecoration:'none'}}><IndianRupee size={13} /> E1 / E1WC</a>
    <a href="#c3" style={{display:'inline-flex',alignItems:'center',gap:'6px',padding:'6px 14px',background:'var(--ifm-color-emphasis-100)',border:'1px solid var(--ifm-color-emphasis-300)',borderRadius:'20px',fontSize:'13px',color:'var(--ifm-color-emphasis-700)',textDecoration:'none'}}><Cpu size={13} /> C3</a>
    <a href="#m3" style={{display:'inline-flex',alignItems:'center',gap:'6px',padding:'6px 14px',background:'var(--ifm-color-emphasis-100)',border:'1px solid var(--ifm-color-emphasis-300)',borderRadius:'20px',fontSize:'13px',color:'var(--ifm-color-emphasis-700)',textDecoration:'none'}}><Server size={13} /> M3</a>
    <a href="#sdc3" style={{display:'inline-flex',alignItems:'center',gap:'6px',padding:'6px 14px',background:'var(--ifm-color-emphasis-100)',border:'1px solid var(--ifm-color-emphasis-300)',borderRadius:'20px',fontSize:'13px',color:'var(--ifm-color-emphasis-700)',textDecoration:'none'}}><Shield size={13} /> SDC3</a>
    <a href="#sdc4-and-wsdc4" style={{display:'inline-flex',alignItems:'center',gap:'6px',padding:'6px 14px',background:'var(--ifm-color-emphasis-100)',border:'1px solid var(--ifm-color-emphasis-300)',borderRadius:'20px',fontSize:'13px',color:'var(--ifm-color-emphasis-700)',textDecoration:'none'}}><Award size={13} /> SDC4 / WSDC4</a>
    <a href="#sql" style={{display:'inline-flex',alignItems:'center',gap:'6px',padding:'6px 14px',background:'var(--ifm-color-emphasis-100)',border:'1px solid var(--ifm-color-emphasis-300)',borderRadius:'20px',fontSize:'13px',color:'var(--ifm-color-emphasis-700)',textDecoration:'none'}}><Database size={13} /> SQL</a>
    <a href="#gpu-gdc-g-and-gdc3" style={{display:'inline-flex',alignItems:'center',gap:'6px',padding:'6px 14px',background:'var(--ifm-color-emphasis-100)',border:'1px solid var(--ifm-color-emphasis-300)',borderRadius:'20px',fontSize:'13px',color:'var(--ifm-color-emphasis-700)',textDecoration:'none'}}><Zap size={13} /> GPU / GDC / GDC3</a>
    <a href="#marketplace" style={{display:'inline-flex',alignItems:'center',gap:'6px',padding:'6px 14px',background:'var(--ifm-color-emphasis-100)',border:'1px solid var(--ifm-color-emphasis-300)',borderRadius:'20px',fontSize:'13px',color:'var(--ifm-color-emphasis-700)',textDecoration:'none'}}><ShoppingBag size={13} /> Marketplace</a>
    <a href="#private-cluster-plans" style={{display:'inline-flex',alignItems:'center',gap:'6px',padding:'6px 14px',background:'var(--ifm-color-emphasis-100)',border:'1px solid var(--ifm-color-emphasis-300)',borderRadius:'20px',fontSize:'13px',color:'var(--ifm-color-emphasis-700)',textDecoration:'none'}}><Layers size={13} /> Private Cluster</a>
  </div>
</div>

---

## Where This Fits in Node Creation

In the MyAccount portal, go to Compute > Nodes and start the create-node flow.

The selection order is:

1. Choose CPU, GPU, or Marketplace.
2. Choose the operating system, GPU card, or marketplace product.
3. Choose the OS version.
4. Choose the node category, such as Linux node, Windows node, or Smart Dedicated Compute.
5. Choose the instance type or family, such as E1, C3, M3, SDC3, SQL, GDC, or GDC3.
6. Choose the exact plan.
7. Configure billing, root storage, backup, network, security, SSH keys, volumes, and advanced settings.

:::note
Some portal labels still use category names such as Linux Virtual Node, Windows Virtual Node, Linux Smart Dedicated Compute, and Windows Smart Dedicated Compute. In this guide, "node" means the compute instance you create and manage in MyAccount.
:::

---

## Quick Comparison

| Family or label       | Best for                                                                                        | Key difference                                                                                        | Avoid when                                                                               |
| --------------------- | ----------------------------------------------------------------------------------------------- | ----------------------------------------------------------------------------------------------------- | ---------------------------------------------------------------------------------------- |
| E1                    | Cost-sensitive Linux development, testing, small services, and flexible root storage needs.     | Lower-cost general compute with separately configured root storage and no default public IP.          | You need the highest predictable CPU, memory, or storage I/O performance from the start. |
| E1WC                  | Cost-sensitive Windows workloads where E1-style behavior is available.                          | Windows counterpart for E1-series behavior.                                                           | You do not need Windows, or you need a dedicated compute family.                         |
| C3                    | CPU-heavy application servers, web servers, batch jobs, analytics, and build workloads.         | Compute-optimized plans.                                                                              | Your bottleneck is memory, not CPU.                                                      |
| M3                    | Databases, caches, in-memory processing, large application runtimes, and memory-heavy services. | Memory-optimized plans.                                                                               | Your workload is mostly CPU-bound and does not need extra RAM.                           |
| SDC3                  | Production workloads that need dedicated compute resources and more predictable performance.    | Dedicated compute resources with guaranteed performance.                                              | You only need short-lived development or low-cost testing capacity.                      |
| SDC4 / WSDC4          | Dedicated compute plans where this newer or Windows-specific label appears in the portal.       | Treat it as a dedicated compute family and compare the exact plan specifications shown by the portal. | The label is not visible for your selected OS, category, location, or account.           |
| SQL                   | Windows SQL Server workloads.                                                                   | SQL Server image or plan path can show SQL version information in the plan selector.                  | You only need a normal Linux or Windows server without SQL Server.                       |
| GPU / GDC / G         | GPU-backed workloads such as rendering, parallel processing, AI, and ML.                        | GPU capacity is part of the node plan.                                                                | Your application only needs CPU and RAM.                                                 |
| GDC3                  | Deep learning, LLM inference, HPC, and newer GPU-heavy workloads.                               | Next-generation dedicated GPU compute plans.                                                          | You do not need GPU memory, CUDA, or dedicated GPU capacity.                             |
| Marketplace           | Prebuilt application stacks.                                                                    | Starts from a marketplace product instead of a plain OS image.                                        | You want a clean base OS and will install everything yourself.                           |
| Private cluster plans | Workloads that must run inside a private cluster assigned to your account.                      | Capacity comes from the selected private cluster, not the standard public create-node pool.           | You do not have or need a private cluster.                                               |

---

## How to Choose

Start with the workload bottleneck:

| If your main concern is                                | Start with                  |
| ------------------------------------------------------ | --------------------------- |
| Lowest practical entry cost for development or testing | E1 or E1WC                  |
| CPU throughput                                         | C3                          |
| RAM capacity                                           | M3                          |
| Dedicated and predictable compute                      | SDC3 or SDC4/WSDC4 if shown |
| Windows SQL Server                                     | SQL                         |
| GPU memory, CUDA, AI, ML, rendering, or HPC            | GPU, GDC, G, or GDC3        |
| A ready-made application stack                         | Marketplace                 |
| Private capacity or account-specific cluster placement | Private cluster plan        |

Then compare the exact plan values in the portal:

- CPU type and CPU count
- RAM
- Disk size
- Read and write IOPS, where shown
- GPU memory and CUDA version, for GPU plans
- SQL Server version, for SQL plans
- Root storage size, for E1-series plans
- Estimated cost and billing option

:::warning
Do not choose a family only by the name. Two plans in the same family can differ significantly. Review the Summary and Estimated Cost before you launch.
:::

---

## E1 and E1WC

Choose E1 when you want a cost-effective node for development, testing, staging, smaller services, or workloads where you want to control root storage separately.

E1 and E1WC are treated as E1-series nodes in the create-node flow. They are cost-effective instances for development and testing.

Important E1-series behavior:

| Area                 | What to know                                                                                                                                                                  |
| -------------------- | ----------------------------------------------------------------------------------------------------------------------------------------------------------------------------- |
| Public IP            | E1-series nodes do not include a default public IP. Select or attach a primary public IP where supported. Add-on IPs can be attached only after the primary public IP exists. |
| Root storage         | Root storage is configured separately in the create-node flow.                                                                                                                |
| Root storage range   | The create-node flow supports 75 GB to 2400 GB.                                                                                                                               |
| Default root storage | The default root storage size is 150 GB.                                                                                                                                      |
| Storage increments   | Storage options are smaller up to 150 GB and then use 50 GB increments from 200 GB onward.                                                                                    |
| Summary              | The Summary panel shows the root storage line item separately for E1-series nodes.                                                                                            |

Use E1 for:

- Development and test environments
- Lightweight application servers
- Staging nodes
- Workloads where you want to right-size root storage at launch
- Windows workloads where E1WC is available and appropriate

Avoid E1 when:

- You need a default public IP without selecting or attaching a primary public IP.
- You need dedicated compute resources.
- Your workload is already known to be CPU-heavy or memory-heavy.
- You need a plan where disk and IOPS are bundled in the same way as non-E1 families.

:::note
If you select E1 or E1WC and the node must be reachable from the internet, plan the public IP choice before launch. Attach a primary public IP where supported after the node reaches Running, then attach Add-on IPs only if extra public addresses are needed.
:::

---

## C3

Choose C3 when CPU performance is the main constraint.

C3 is compute-optimized for CPU-intensive workloads such as web servers, batch processing, and analytics.

Use C3 for:

- Web servers with steady request volume
- API servers
- Build workers
- Batch processing
- Analytics jobs
- CPU-heavy application runtimes

Watch for:

- RAM pressure. If the application spends time swapping or requires large in-memory datasets, compare M3 before launching.
- Storage throughput. Review disk and IOPS in the plan selector when storage performance matters.
- Horizontal scaling. If you expect to add or remove nodes frequently, also review scaling and load-balancing requirements.

Avoid C3 when:

- Your workload is mostly memory-bound.
- You need dedicated compute placement rather than compute-optimized shared-family plans.
- You need GPU acceleration.

---

## M3

Choose M3 when memory is the main constraint.

M3 is memory-optimized for in-memory databases, caching, and real-time big data processing.

Use M3 for:

- Databases that need more RAM
- Caches such as Redis-like workloads
- Search or indexing workloads with large memory needs
- Data processing services that keep large working sets in memory
- Application servers where memory, not CPU, is the bottleneck

Watch for:

- CPU saturation. If memory is healthy but CPU is the bottleneck, compare C3.
- Disk and IOPS. A database can be memory-heavy and storage-sensitive at the same time, so review the plan details carefully.
- Backup and restore requirements. Production data workloads should have a backup plan before launch.

Avoid M3 when:

- The workload is simple and small enough for E1.
- CPU is the primary bottleneck.
- You need dedicated compute resources.

---

## SDC3

Choose SDC3 when you need dedicated compute resources and more predictable performance.

The create-node flow describes SDC3 as dedicated compute resources with guaranteed performance. Supported categories can include Linux Smart Dedicated Compute and Windows Smart Dedicated Compute, depending on the selected OS, location, and account.

Use SDC3 for:

- Production workloads where predictable compute matters
- Workloads sensitive to noisy-neighbor effects
- Dedicated application servers
- Dedicated database or middleware nodes
- Cases where the portal recommends or exposes Smart Dedicated Compute for your selected OS

Watch for:

- The category you selected. Dedicated compute families normally appear under Smart Dedicated Compute categories.
- Billing impact. Dedicated compute can cost more than development-focused families, so check the live Summary.
- OS and plan availability. Not every family appears for every OS, location, or account.

Avoid SDC3 when:

- You only need a short-lived dev/test node.
- You are optimizing only for lowest cost.
- You need GPU acceleration.

---

## SDC4 and WSDC4

Some plan lists include WSDC4, and existing node docs refer to WSDC4 for some node image and snapshot behavior. Depending on your selected OS, category, location, and account, the portal may show SDC4-like or WSDC4-like dedicated compute labels.

Treat SDC4 or WSDC4 as a dedicated compute choice:

- Use it when the portal shows it for a dedicated compute workload.
- Compare CPU, RAM, disk, and IOPS in the exact plan selector.
- For Windows plans, verify whether the selected image includes only Windows or also SQL Server.
- Confirm backup, image, snapshot, and encryption options in the create-node flow before launch.

Avoid documenting or assuming a specific SDC4 plan unless it is visible in your portal selection or returned by the approved API.

:::note
If the portal does not show SDC4 or WSDC4 for your selected path, choose from the families that are visible. Availability can depend on OS, category, location, account configuration, and current inventory.
:::

---

## SQL

Choose SQL when you are launching a Windows node that needs SQL Server included in the selected image or plan.

SQL appears in two ways:

- Backend OS categories include SQL Web and SQL Standard values.
- In the Windows node flow, the plan selector can show SQL version information when the selected OS category is SQL Web or SQL Standard.

Use SQL for:

- Windows applications that require Microsoft SQL Server
- SQL Server Web edition workloads, where available
- SQL Server Standard edition workloads, where available
- Cases where you want the SQL version visible during plan selection

After launch, use [Connect and use a SQL Server node](/docs/myaccount/node/connect-to-node/sql-node) to verify SQL Server locally, configure database access, and understand SQL-specific node limitations.

Check before launch:

| Question                           | Why it matters                                                                |
| ---------------------------------- | ----------------------------------------------------------------------------- |
| Is the category Windows?           | SQL Server plans are tied to Windows SQL image paths in the create-node flow. |
| Which SQL edition is selected?     | SQL Web and SQL Standard are different choices.                               |
| Which SQL version is shown?        | The plan selector can include SQL version information.                        |
| Is the selected plan large enough? | SQL workloads can be CPU-, memory-, and I/O-sensitive.                        |
| Do you need backups?               | Database workloads should have a backup and restore plan.                     |

Avoid SQL when:

- You only need a base Windows node.
- You plan to install and manage your own database stack on Linux.
- You do not need SQL Server licensing or SQL Server image support.

---

## GPU, GDC, G, and GDC3

Choose GPU-backed families when your workload needs GPU acceleration.

The create-node flow maps GPU family labels to GPU-focused descriptions:

| Label    | Best fit                                                                                       |
| -------- | ---------------------------------------------------------------------------------------------- |
| GPU      | Graphics-intensive and parallel processing tasks.                                              |
| GDC or G | Dedicated GPU compute for high-performance AI/ML workloads requiring guaranteed GPU resources. |
| GDC3     | Next-generation dedicated GPU instances for deep learning, LLM inference, and HPC workloads.   |

GPU plans can show:

- GPU card or GPU family
- GPU memory
- CUDA version
- CPU and RAM
- Storage details, where applicable

Use GPU, GDC, G, or GDC3 for:

- Deep learning training
- LLM inference
- Model serving
- GPU-accelerated data processing
- Rendering or graphics workloads
- HPC workloads that require GPU acceleration

Avoid GPU families when:

- Your workload only needs CPU and RAM.
- You do not need CUDA, GPU memory, or GPU drivers.
- A CPU family can meet the requirement at lower complexity.

:::note
GPU node creation can expose different fields than CPU node creation. Review GPU card, memory, CUDA version, and plan details in the portal before launch.
:::

---

## Marketplace

Marketplace is not a raw compute family. It is a launch path for prebuilt products or application stacks.

Use Marketplace when:

- You want a preconfigured application image.
- You want to reduce setup time.
- The marketplace product matches your intended workload.

Choose a normal OS image instead when:

- You want a clean server.
- You need full control over installation steps.
- You want to build the application stack yourself.

---

## Private Cluster Plans

If your account has private cluster access, the create-node flow can show a private cluster option before the standard node components.

Use a private cluster plan when:

- Your organization has dedicated private capacity.
- The workload must run in that private cluster.
- You need account-specific placement or isolation provided by the cluster.

Choose the standard E2E cluster path when:

- You do not have private cluster access.
- You do not need private cluster placement.
- You want to use the normal public create-node plan list.

---

## Older or Less Common Series Names

You may see or hear older or less common labels such as C, C2, M, SD, SDC, WSDC3, VGC, GDCW3, M3VPS, C3VPS, Plesk-related plans, or application-specific plan names.

Treat these carefully:

- If the label appears in the portal, use the visible plan details to decide.
- If the label appears only in old documentation, do not assume it is still generally available.
- If the label appears only in automation or account-specific lists, it may be legacy, specialized, account-specific, or tied to another workflow.
- For public API use, rely on the approved API documentation and live API responses.

When in doubt, choose from the families shown in the portal for your selected OS, category, location, and account.

---

## Practical Recommendations

| Scenario                     | Recommended starting point                             |
| ---------------------------- | ------------------------------------------------------ |
| First Linux test server      | E1                                                     |
| First Windows test server    | E1WC if shown, otherwise compare visible Windows plans |
| Production web application   | C3 or SDC3                                             |
| CPU-heavy worker             | C3                                                     |
| Memory-heavy application     | M3                                                     |
| Cache or in-memory service   | M3                                                     |
| Dedicated production compute | SDC3, SDC4, or WSDC4 if shown                          |
| Windows SQL Server           | SQL                                                    |
| AI/ML training or inference  | GDC3 or other visible GPU family                       |
| Prebuilt application stack   | Marketplace                                            |
| Private capacity workload    | Private cluster plan                                   |

---

## What to Check Before Launch

Before clicking Launch, verify:

- The selected OS and version are correct.
- The family matches the workload bottleneck.
- The exact plan has enough CPU and RAM.
- Disk and IOPS are suitable for the application.
- E1-series root storage is sized correctly.
- Public access is planned, especially for E1-series nodes.
- Security group rules allow only the required traffic.
- Backup is enabled or intentionally deferred.
- Public IP, VPC, volume, encryption, and SSH options are correct.
- The Summary and Estimated Cost match your expectation.

---

## Related Resources

| Resource                                                                           | Use it for                                                   |
| ---------------------------------------------------------------------------------- | ------------------------------------------------------------ |
| [Create a node](/docs/myaccount/node/getting-started/create-node)                  | Launch a compute node from MyAccount.                        |
| [Connect and use a SQL Server node](/docs/myaccount/node/connect-to-node/sql-node) | Access and configure a Windows SQL Server node after launch. |
| [E1 Series Nodes](/docs/myaccount/node/features/e1-series)                         | Learn E1-specific operations after launch.                   |
| [Manage Nodes](/docs/myaccount/node/manage)                                        | Manage existing nodes.                                       |
| [Node Snapshots](/docs/myaccount/node/features/snapshots)                          | Review snapshot behavior for supported node families.        |
| [Create Images](/docs/myaccount/node/features/images)                              | Save node images for future launches.                        |
| [Node Encryption](/docs/myaccount/node/features/encryption)                        | Create encrypted nodes where supported.                      |
| [Committed Nodes](/docs/myaccount/node/troubleshoot/committed-nodes)               | Understand committed-node behavior.                          |

## Next Step

After you choose a family, continue with [Create a node](/docs/myaccount/node/getting-started/create-node) and select the exact plan in the portal.

---