DX Storage is a high-performance, massively scalable, self-managing, cost-effective, fixed content
storage software that runs on clusters of standard x86 computer hardware. DX Storage provides the
integrity of an archive with the performance of primary storage.
1.1. Components
The DX Storage infrastructure is made up of the following components.
1. Cluster of x86 computers/nodes with one or more physical volumes running DX Storage software
2. Client applications that access the cluster through HTTP
3. TCP/IP network connecting cluster nodes and clients
1.2. About this Document
1.2.1. Audience
This document is intended for people in the following roles.
1. Storage system administrators
2. Network administrators
3. Technical architects
Throughout this document, the storage system administrator and network administrator roles will be
referred to as the administrator. The administrators are normally responsible for allocating storage,
managing capacity, monitoring storage system health, replacing malfunctioning hardware, and
adding additional capacity when needed. This document along with the application guide will be
valuable to technical architects in designing scalable, highly redundant, cost effective application
storage solutions.
1.2.2. Scope
This document covers the basic steps necessary to deploy and administer a DX Storage cluster.
The reader is expected to be familiar with TCP/IP networking and have a basic knowledge of x86
hardware setup. For additional configuration details and in-depth administration information, see the
DX Object Storage Administration Guide.
For information regarding DX Storage’s HTTP interface and application integration, see the DX
Object Storage Application Guide.
Chapter 2. Network Infrastructure and
Recommendations
The following sections in this chapter discuss how to configure your network for DX Storage
clusters:
• Section 2.1, “About DX Storage in the Network”
• Section 2.2, “Setting Up the Network for DX Storage”
• Section 2.3, “Setting Up Network Services”
• Section 2.4, “Setting Up PXE Booting”
• Section 2.5, “Setting Up a Configuration File Server”
• Section 2.6, “Network Devices and Priority”
• Section 2.7, “Proxying the Admin Console”
• Section 2.8, “IGMP Snooping Support”
2.1. About DX Storage in the Network
The following topics in this section provide a high-level overview of setting up a DX Storage cluster
in your network:
• Section 2.1.1, “Sample Networks”
• Section 2.1.2, “Layer 3 Switching and Routing”
• Section 2.1.3, “Switching Hardware”
• Section 2.1.4, “Internet Deployments”
2.1.1. Sample Networks
The following figure shows a basic network where the DX Storage cluster nodes, the clients, and
the servers are all in the same subnet. While this is the easiest to setup and requires only basic
hardware, it does not offer any traffic separation between the DX Storage nodes and the rest of the
network.
The next figure shows a more sophisticated network topology that uses a router to separate traffic
between the DX Storage nodes and the rest of the network.
A router or a layer 3 switch routes network packets between subnets. A router segregates network
traffic by filtering packets based on the subnet to which they are addressed. This segregation is
important so that DX Storage nodes have predicable network bandwidth for their use and so that
their multicast and unicast traffic doesn’t interfere with computers on the rest of the network.
2.1.3. Switching Hardware
When selecting Ethernet switching hardware, consider that many client workstations have 100Mbps
network interfaces and it might not be cost effective to connect these workstations to Gigabit ports.
Additionally, the operating systems and applications running on these workstations might be unable
to effectively utilize more than 100Mbps of bandwidth.
The following figure shows a network architecture where workstations, enterprise servers, and DX
Storage nodes are isolated on appropriately sized switches.
With more sophisticated switching hardware, the network segments can be isolated as different
Virtual LANs (VLANs) on the same device. Additionally, some enterprise-class switching hardware
has routing capabilities.
You can also design the DX Storage subnet to use redundant switches in the event of a switch
failure. The following figure shows an example of DX Storage nodes connected to multiple network
switches. In the event any single component becomes unavailable, a redundant path is available so
communication is uninterrupted.
When deploying DX Storage with multiple switches, the interconnection between the switches
requires planning. To allow for full speed communications between active ports on different
switches, the network connection between switches must be faster than the individual ports. Contact
your switch provider for information about proprietary software or use a mechanism such as link
aggregation.
2.1.4. Internet Deployments
Network security is one of the top considerations during the deployment of any service on the
Internet or within an extensive enterprise WAN. In these types of deployments, put a firewall or
filtering router in front of DX Storage to control the kind of traffic and requests that are allowed to
reach the cluster nodes.
The preceding figure shows a firewall that allows requests on TCP/80. This is the default SCSP port,
but it should be changed to match the scspport value set in the node or cluster configuration file for
the DX Storage cluster if it is something other than 80.
If the firewall is sophisticated enough to examine Layer 7 (Application Layer), or the contents of
the HTTP requests, further restrictions should be made to allow only GET, HEAD, POST, DELETE
requests. If a cluster is exposed read-only to these external clients, the POST and DELETE
requests can be blocked to prevent updates to the cluster. To prevent client access to the node
status page, the firewall should deny “GET /” requests to the cluster nodes.
Administrators should block Internet access to the Admin Console port (default TCP/90) and to
the SNMP port (UDP/161). In wide-area networks, further restrictions might be desirable to restrict
access to these services to specific administrative networks or workstations.
Anytime critical devices such as firewalls are introduced into a network architecture, they should be
deployed in redundant pairs to minimize the chance of failures that cut-off all client access.
2.2. Setting Up the Network for DX Storage
Client applications must be able to initiate TCP connections with all nodes in a DX Storage cluster
using the designated access port, typically port 80. Internally, DX Storage nodes must be able to
communicate with each other using UDP, TCP, and multicast.
The following topics in this section discuss how to set up a DX Storage cluster in a standard TCP/IP
networking environment: