Server for storing 100 TB of data: selection and setup

Creating a server capable of efficiently storing 100 TB of data or more requires a specialized 100tb storage server with high-capacity HDD drives, organized into a fault-tolerant RAID array, which ensures reliability and high performance for tasks such as backups, media archives, and cloud storage.

Which server to choose for storing 100 TB of data?

Choosing the optimal server for storing such a significant volume of information as 100 TB is a complex task that requires considering many factors: from the type of drives to the network infrastructure. For such data volumes, it is most advisable to consider dedicated servers or specialized enterprise-grade NAS systems, rather than cloud storage, which can prove significantly more expensive in the long run with active use.

Key components to consider when choosing a large storage server:

• Drives: The main element. For 100 TB, large-capacity drives (16 TB, 18 TB, 20 TB, or even 22 TB) will be required, 6-12 of them, combined into a RAID.
• RAID controller: Hardware or software. A hardware RAID controller significantly improves performance and offloads the central processor.
• Processor (CPU): For most data storage tasks, a top-tier multi-core processor is not required. Entry-level Intel Xeon E3/E5 or AMD EPYC are optimal.
• Random Access Memory (RAM): 16 GB to 64 GB is recommended. Larger amounts of RAM can be useful for file systems like ZFS, which use caching.
• Network interface: For working with 100 TB of data, a 10 Gbit/s network port is highly desirable, especially if data access will be performed over the network by a large number of users or applications.
• Chassis: Must accommodate the necessary number of drives and provide adequate cooling.

A dedicated server provides full control over hardware and software, which is critical for scaling and fine-tuning performance.

HDD vs. SSD: why hard drives remain the foundation for large storage servers?

When choosing drives for a high capacity server of 100 TB or more, hard disk drives (HDDs) remain the unchallenged solution in terms of price per gigabyte. Despite SSDs' superiority in speed, their cost per terabyte is several times higher, making them impractical for mass data storage.

• Cost per TB: HDDs offer a significantly lower cost per terabyte. For example, an 18 TB enterprise HDD might cost around $300-400, while an SSD of similar capacity would cost thousands of dollars.
• Lifespan: Modern enterprise HDDs are designed for long-term 24/7 operation.
• Performance: For sequential read/write of large files (video, backups, archives), HDD performance in a RAID array is quite sufficient. For tasks requiring high random access speed, small SSDs can be used as a cache (e.g., with ZFS or LVM).

It is recommended to use enterprise-class HDDs (Enterprise HDDs) with CMR (Conventional Magnetic Recording) or PMR (Perpendicular Magnetic Recording) technologies and helium filling to enhance reliability and energy efficiency. They are specifically designed for continuous operation under constant load in server environments.

JBOD, RAID, or ZFS: how to build a fault-tolerant storage for a high capacity server?

The choice of data storage architecture is critical for the reliability and performance of your high capacity server. There are several main approaches:

JBOD (Just a Bunch Of Disks)

JBOD means that each disk is used as a separate device. This is the simplest option, but it provides neither fault tolerance nor performance gain. Failure of one disk leads to data loss on it. It is only suitable for temporary storage or if data is not critical.

RAID (Redundant Array of Independent Disks)

RAID combines several disks into a logical array to improve performance and/or fault tolerance. For 100 TB of data, the following levels are most relevant:

• RAID 0 (Striping): Combines disks for maximum performance but lacks fault tolerance. Failure of one disk leads to the loss of the entire array. Not recommended for critical data.
• RAID 1 (Mirroring): Duplicates data on two disks. High fault tolerance, but 50% of capacity is lost. For 100 TB, 200 TB of physical disks would be required, which is expensive.
• RAID 5 (Striping with Parity): Distributes data and parity information across all disks. Allows tolerating the failure of one disk. Efficient use of capacity. For large arrays with modern disks (16 TB+), it becomes risky due to the lengthy rebuild time.
• RAID 6 (Striping with Dual Parity): Similar to RAID 5, but with two parity blocks. Allows tolerating the failure of two disks. Ideal for large arrays (100 TB and more) with high-capacity disks, where the probability of a second disk failure during the rebuild of the first increases.
• RAID 10 (RAID 1+0): A combination of RAID 1 and RAID 0. Data is first mirrored, then the mirrors are striped. High performance and fault tolerance (can tolerate the failure of several disks, provided they are not mirrors of each other). Requires a minimum of 4 disks. Effective capacity is 50% of the total number of disks.

ZFS (Zettabyte File System)

ZFS is a file system and logical volume manager that offers exceptional fault tolerance, data self-healing, snapshots, and many other features. ZFS can replace a hardware RAID controller by functioning as a software RAID (RAID-Z1, RAID-Z2, RAID-Z3, Mirror). For 100 TB of data, ZFS with RAID-Z2 (equivalent to RAID 6) or RAID-Z3 (allows tolerating the failure of three disks) is an excellent solution, especially for those who value data integrity and advanced management capabilities.

# Example of creating a software RAID 6 on Linux with mdadm
# Assume you have 8 disks: /dev/sdb, /dev/sdc, ..., /dev/sdi
# First, clear disks of old metadata
sudo mdadm --zero-superblock /dev/sdb /dev/sdc /dev/sdd /dev/sde /dev/sdf /dev/sdg /dev/sdh /dev/sdi

# Create RAID 6 from 8 disks (6 active, 2 for parity)
sudo mdadm --create /dev/md0 --level=6 --raid-devices=6 --spare-devices=2 /dev/sdb /dev/sdc /dev/sdd /dev/sde /dev/sdf /dev/sdg /dev/sdh /dev/sdi

# Check array status
cat /proc/mdstat

# Format the array to XFS
sudo mkfs.xfs /dev/md0

# Create a mount point
sudo mkdir /mnt/storage100tb

# Mount the array
sudo mount /dev/md0 /mnt/storage100tb

# Add to fstab for automatic mounting (get the array's UUID)
# sudo blkid /dev/md0
# UUID="<your_UUID>" /mnt/storage100tb xfs defaults 0 0

How much does 1 TB of storage cost: calculations and savings with storage server rental?

Calculating the cost of storing 100 TB of data includes not only the price of the disks themselves but also many other factors. When choosing between purchasing your own server and storage server rental, rental often proves significantly more cost-effective, especially for businesses.

Purchasing your own server (on-premise):

• Capital Expenditures (CAPEX):
  • Server hardware (chassis, CPU, RAM, RAID controller, PSU): from $1500 to $5000+.
  • Drives (8x18 TB Enterprise HDD): $2400 - $3200 (at $300-400 per drive).
  • Network equipment (10GbE card, switch): from $200 to $1000+.
  • Software (OS, licenses): from $0 (Linux) to $1000+.
  • Total CAPEX: from $4100 to $10000+
• Operational Expenditures (OPEX) monthly:
  • Electricity: $50 - $150 (depends on server and tariffs).
  • Cooling: $20 - $100.
  • Internet channel: $50 - $500 (depends on speed and traffic).
  • Maintenance and support (administrator, disk replacement): from $100 to $500+.
  • Total OPEX: from $220 to $1250+ monthly.

Dedicated storage server rental (Valebyte.com):

With rental, you pay a fixed monthly amount that includes all the aforementioned OPEX, as well as equipment depreciation. You don't have to worry about initial capital expenditures, replacing failed components, electricity, or network setup.

• Valebyte offers HDD servers from $29/month for basic configurations.
• For a 100tb storage server with a powerful processor, sufficient RAM, and several high-capacity disks in RAID, the rental cost can range from $150 to $400+ per month. This includes the server, disks, electricity, cooling, network channel, and basic support.

Thus, storage server rental significantly reduces initial investments and allows for predictable budget planning, which is especially beneficial for small and medium-sized businesses, as well as startups.

Setting up a Linux server for 100 TB data storage: from formatting to mounting

After receiving a dedicated server with the necessary number of HDDs, the next step is its software configuration. We will focus on Linux, as the most popular and flexible OS for storage servers.

1. Operating System Installation

It is recommended to use server Linux distributions such as Debian, Ubuntu Server, CentOS Stream, or Rocky Linux. Install a minimal version, without a graphical interface, to save resources.

2. Creating a RAID Array

As discussed earlier, use software RAID with mdadm or ZFS. For 100 TB, RAID 6 or RAID 10 are preferred. An example command for mdadm was provided above.

3. Choosing and Formatting the File System

For large data volumes, it is recommended to use file systems optimized for this purpose. XFS is an excellent choice for large files and high performance. ext4 is also suitable, but XFS often shows better results on high-capacity arrays.

# Formatting the array to XFS
sudo mkfs.xfs -f /dev/md0

4. Mounting the File System

Create a mount point and mount the array. For automatic mounting when the server boots, add the corresponding entry to the /etc/fstab file. Use the disk's UUID for reliability.

# Creating a mount point
sudo mkdir /mnt/storage

# Mounting (temporarily)
sudo mount /dev/md0 /mnt/storage

# Getting the array's UUID
sudo blkid /dev/md0

# Example entry in /etc/fstab
# UUID="your_array_UUID" /mnt/storage xfs defaults,noatime 0 0
# The noatime option reduces write operations, improving performance.

5. Setting Access Permissions

Determine who will have access to the storage. Create user groups and configure permissions using chown and chmod.

# Example: create a group "storageusers" and give it full permissions to the directory
sudo groupadd storageusers
sudo chown -R root:storageusers /mnt/storage
sudo chmod -R 770 /mnt/storage

6. Monitoring

Configure monitoring for the RAID array and disk status. mdadm can send email notifications. Use S.M.A.R.T. to track HDD health.

# Installing smartmontools
sudo apt update && sudo apt install smartmontools (for Debian/Ubuntu)
sudo yum install smartmontools (for CentOS/Rocky)

# Checking disk status
sudo smartctl -a /dev/sdb

Valebyte Ready Solutions: Your Reliable Data Storage Server

Valebyte.com offers a wide range of dedicated servers, ideally suited for creating a 100tb storage server and larger storage solutions. Our solutions are designed to provide maximum performance, reliability, and scalability, making them an optimal choice for backups, media hosting, archives, and other tasks requiring large amounts of disk space.

We offer servers with high-capacity enterprise-class HDDs, with flexible configuration options for RAID arrays (hardware or software), as well as high-speed network connections (up to 10 Gbit/s).

Examples of Valebyte dedicated storage server configurations:

Each data storage server from Valebyte comes with a pre-installed OS of your choice (Linux distributions), as well as the option to install a hardware RAID controller for maximum performance and reliability. Our infrastructure ensures the stability and high availability of your data.

Recommendations for choosing and operating a 100-terabyte storage

Effective management of a 100tb storage server requires adherence to a number of recommendations:

1. Always use RAID (or ZFS): Never rely on JBOD for critical data. RAID 6 or RAID 10 is your minimum for such volumes.
2. Choose Enterprise HDDs: They are more expensive but significantly more reliable than ordinary consumer drives in 24/7 operation. Pay attention to CMR/PMR and helium filling.
3. Don't forget about backups: RAID provides fault tolerance but is not a substitute for backups. RAID controller failure, accidental deletion, or file system corruption can lead to data loss. Implement a 3-2-1 strategy (3 copies of data, on 2 different media, 1 of which is offline/offsite).
4. Monitoring is key to longevity: Set up notifications for RAID array status and S.M.A.R.T. parameters of disks. Regularly check logs.
5. Plan for scalability: 100 TB might seem like a lot now, but data grows quickly. Choose a server that allows adding disks or easily migrating to a more powerful solution.
6. Optimize network connection: For working with such data volumes, 1 Gbit/s might not be enough. Consider 10 Gbit/s Ethernet.
7. Test recovery: Periodically simulate a disk failure and test the recovery procedure to ensure its functionality and your readiness.

Conclusion

Choosing and configuring a server for 100 TB of data storage is a significant investment in the reliability and performance of your infrastructure. The optimal solution is to rent a specialized dedicated 100tb storage server with high-capacity HDDs in a RAID 6 or RAID 10 (or ZFS) configuration, which allows for efficient cost management and ensures high data availability. Valebyte.com offers ready-made and customizable solutions that will help you create powerful and fault-tolerant storage for any task.