Table of Contents
- Deciding to build or buy your object storage solution?
- Choosing your hardware components
- Minimum Hardware Requirements
- Recommended Hardware Requirements
- Additional Components
- Buying a Turnkey Solution
Deciding to build or buy your object storage solution?
Once you have decided that you need to implement an object storage solution, you may consider the question of whether it would be better to build your own object storage solution starting with open source code or to buy a proven object-based storage solution such as Caringo Swarm. There are a number of considerations that come into play when making this decision.
Considerations when building your own storage solution?
- Building a solution takes a substantial amount of work, particularly on the front end (hardware sizing & purchase, integration, etc.).
- Think about industry, government, legal and business-specific requirements and if you need a custom-built solution to meet them.
- Flexibility for units of purchase and licensing may be available in an “on-demand” model.
Considerations when buying a complete storage solution?
- A ready-to-deploy solution can ease the demand on your staff.
- Complex requirements may not be met by an out-of-the-box product.
- This option requires a more structured approach to acquisition with longer lead time for ordering and deploying additional storage/
Choosing your hardware components
When it comes to choosing your hardware components for running object storage software, you need to look at a number of factors including processors (CPU), memory (RAM), storage (HDD or SSD) and networking (LAN, WAN and bonding).
Processor (CPU)
- Supports x86-64 Intel or AMD standard processors
- Single or multiple sockets supported (and multi-core)
- Recommended use of Intel or AMD processors which include Advanced Encryption Standard New Instructions (AES-NI) support:
- Used by Swarm for improved performance of Encryption at Rest (EAR)
- Most modern server processors include AES-NI support as of the year 2010
Memory (RAM)
To run object storage, there are memory-required functions related to object count, object type and data-protection scheme. If you have a larger cluster, you should plan for additional memory for the Overlay Index or other features that may require resources.
RAM Per Storage Node for Object Capacities
| GBs Server Ram | 16 | 32 | 64 | 128 |
| Storage Node RAM index slots | 268M | 536M | 1073M | 2146M |
| Immutable Objects | 268M | 536M | 1073M | 2146M |
| Mutable Objects | 134M | 268M | 536M | 1073M |
| 5:2 Erasure Coded Objects | 26M | 53M | 107M | 214M |
Storage Disks (HDD/SSD)
- Direct-attached drives
- Controllers: SAS or SATA JBOD HBAs (SAS preferred)
- “Hot plug” connector/backplane support
- Enterprise-class drives with 24×7 continuous duty cycle support
Networking (LAN/WAN, Bonding)
- Ethernet (with appropriate connector type)
- 1 Gb to 10 Gb (or higher if needed), tested all the way to 100 Gb with success
- Bonding of multiple ports supported for throughput and redundancy (Includes 802.3ad LAG/LACP if switch redundancy is required)
- Jumbo Frame support
- Intel and Broadcom are typical vendor choices
Minimum Hardware Requirements
Object storage can be deployed as virtual machines (VMware guests), appropriate for functional design and testing. The implementation must contain 3 or more nodes (chassis) in a cluster. You will need the following hardware:
- 4 GB RAM
- x86-64 Intel or AMD processors
- 100 Mbps network interface card
- 1 storage drive
Recommended Hardware Requirements
It is typically recommended to deploy on physical servers, but object storage can be deployed on virtual machines (VMware). The deployment should contain a minimum of 4 nodes (chassis) in the cluster (scale up/scale out). Memory will depend on the object count and other factors. Here are the additional recommendations for hardware:
- 8 GB or more RAM
- Multi-socket/Multi-core x86-64 Intel or AMD processors
- Gigabit Ethernet network interface(s) or 10 GbE
- 2 or more enterprise-class SAS drives
Additional Components
For deploying Caringo Swarm, here are the recommendations to keep in mind for additional components.
Swarm Platform Server
The Swarm Platform Server is used to boot, monitor and manage the Swarm storage cluster. It can be deployed as virtual machines (VMware OVA available).
| CPU | x86-64 (multi-socket and/or multi-core, usually 2 cores) |
| Memory | 8 GB RAM |
| Storage | 80 GB Minimum (more preferred for large clusters, logging) |
| Networking | 1 Gb Ethernet (typically) |
| Server Count | Single |
Swarm Elasticsearch
Elasticsearch (included in the Swarm software) is used to perform query and list operations on objects within a cluster and the object metadata (including custom metadata). Elasticsearch can be deployed as virtual machines (VMware).
| CPU | x86-64 (multi-socket and/or multi-core) |
| Memory | 64 GB RAM per 1 billion distinct objects |
| Storage | 1.5 TB required for 1 billion distinct objects |
| Networking | 1 Gb Ethernet (typically) |
| Server Count | 3-4 Minimum for redundancy and performance |
Swarm Gateway
The Swarm Gateway is used to process protocols along with authentication and authorization and can be deployed as virtual machines (VMware).
| CPU | x86-64 (multi-socket and/or multi-core) |
| Memory | 4 GB RAM Minimum |
| Storage | 4 GB Minimum |
| Networking | 1 Gb Ethernet (typically) |
| Server Count | Scale out as needed to support client bandwidth |
Buying a Turnkey Solution
There are benefits to purchasing a best-of-breed turnkey object storage solution as it can speed up deployment and eliminate much of the research and decision making that is involved in building your own implementation.
Once you have decided that you prefer to purchase a complete solution (that is, one that includes all of the hardware and software needed, you will want to make sure you pick one that complements your existing storage infrastructure and that you can easily plug into your workflows via S3, NFS, Windows or macOS. You should also look for an easy-to-use UI (user interface) and one where the metadata makes it simple to retrieve objects (files) as needed. There also may be particular features that are useful depending on the type of organization you are in. For example, Partial File Restore is valuable for those who deal with digital video workflows.
Swarm Server Appliances
Caringo offers Swarm Servers to provide a complete, pre-integrated and easy-to-deploy hardware and software solution that starts at 168TB raw with pay-as-you-grow economics. Models include the Single Server Appliance (SSA), m1000 management server, s3000 storage server and hd5000 high-density storage server. Below, you can see how easy it is to scale from 168TB to multi-petabytes.
- Rapid Flexible Deployment: Swarm Servers scale from 168 TB to 100s of Petabytes. Initial deployment can take as little as 2 hours and you can expand your storage in minutes.
- Optimize Durability or Footprint: You can mix single, standard storage and high-density appliances and select any mix of replication or erasure coding.
- Feature Rich for End Users: Swarm has built-in content management and search and secures your content for sharing and delivery.
- Hassle-free for Administrators: 1 sys admin can manage 50+ PB. Swarm features rapid recovery and up to 25x less downtime when compared to RAID.
Swarm Server Appliance Line
Minimum configuration is one Single Server Appliance (SSA). You can deploy an m1000 with s3000 or hd5000 appliances for enhanced performance and higher capacity.
SSA & s3000 storage servers
| Rack Units (RU) | 1U |
| Dimensions (W/H/D) | 17.6”W, 1.7”H, 37.4”D |
| Processors (CPU) | 2 x Intel Xeon Silver Proc |
| Storage (HDD/SSD) | 12 x 14TB Seagate Exos HDD |
| Raw Capacity | 168TB RAW |
| Networking (NIC) | 2 x 10GBase-T ports |
| Warranty | 3-year Hardware Warranty |
| Maintenance & Support | 3-year |
SSA Only*
| Storage (HDD/SSD) | 2 x 3.84TB Seagate SSD |
| Additional Software | KVM and Management Services |
hd5000 high-density storage server
| Rack Units (RU) | 4U |
| Dimensions (W/H/D) | 17.2”W, 7”H, 30.2”D |
| Processors (CPU) | 2 x Intel Xeon Silver Proc |
| Storage (HDD/SSD) | 60 x 14TB Seagate Exos HDD |
| Raw Capacity | 840TB RAW |
| Networking (NIC) | 4 x 10GBase-T ports |
| Warranty | 3-year Hardware Warranty |
| Maintenance & Support | 3-year |
m1000 management server
| Rack Units (RU) | 1U |
| Dimensions (W/H/D) | 17.2”W, 1.7”H, 23.5”D |
| Processors (CPU) | 2 x Intel Xeon Silver Proc |
| Storage (HDD/SSD) | 4 x 960GB Seagate Nytro SSD |
| Storage 2 (SSD/NVMe) | 1 x 256GB Micron NVMe SSD |
| Networking (NIC) | 4 x 10GBase-T ports |
| Warranty | 3-year Hardware Warranty |
| Maintenance & Support | 3-year |
Included Software Components
- KVM
- Host OS Software (CentOS)
- Swarm Platform Server
- Swarm Storage
- Swarm Gateway (S3)
- Elasticsearch (metadata list & query)
- Software Load Balancer (HAProxy); SSL/TLS offload happens here
- SwarmNFS (optional)