Storage

NAS – Advantages, Limitations & Recommendations for Ethernet Storage over TCP/IP Networks

NAS is the short form of Network Attached Storage. Of course, we are talking about Ethernet, TCP/IP Networks and the storage systems that use them to store data. In this article, we will see why the NAS over TCP/IP Networks has gained a lot of popularity and acceptance in the recent years and matured as a reliable storage technology like SAN (Storage Area Networks). We will also see some best practices and recommendations for setting up a TCP/IP network to make NAS more efficient.

What are NAS, SAN and DAS?

DAS – Direct Attached Storage : This refers to all forms of storage used by individual computers to store data like internal hard disks, USB hard disks, CD/ DVD’s, USB flash drives, etc. This is the cheapest and the most accessible form of storage available. But the storage space in DAS systems is quite limited and they are not meant for multi-user purposes.

SAN – Storage Area Network : A storage area network is a separate network that uses its own protocols (read Fiber Channel), SAN Switches, HBA adapters in each server, SAN Controllers and Disks to store data. This is primarily meant for huge enterprises and data centers where the volume of data to be stored is huge and mostly consists of transaction intensive data-bases, ERP, CRM systems etc. The SAN system stores data in blocks (rather than files) which enables a lot of storage optimization techniques to be applied more easily. Yes, they are very expensive to create and maintain.

NAS – Network Attached Storage : A NAS storage appliance is a computing device that can be attached anywhere on the network, primarily to store files. NAS connects the storage to the TCP/IP network and enables users to access storage resources using both CIFS (Microsoft based Common Internet File System) and NFS (Unix based Network File System) protocols over the familiar TCP/IP networks. The NAS appliance has its own IP address and hence can be accessed by users directly over the network to store files as well as retrieve them when required. NAS generally uses multiple disks to store data and provides file level storage and access to data.

Advantages of Network Attached Storage (NAS, compared to SAN):

  • NAS systems stores data as files (predominantly) and support both CIFS and NFS protocols. They can be accessed easily over the commonly used TCP/IP Ethernet based networks and support multiple users connecting to it simultaneously.
  • Entry level NAS systems are quite inexpensive (and the costs keep coming down with mass adoption) – they can be purchased for capacities as low as 1 or 2 TB with just two disks. This enables them to be deployed with Small and Medium Business (SMB) networks easily. However, they can scale up very well, there are NAS appliances that support multiple array of disks giving a total storage capacity in excess of 1000 TB. You can add storage by plugging a new appliance or adding disks (if possible in a particular model) to existing NAS appliance.
  • A NAS device (based on the model) may support one or more RAID levels to make sure that individual disk failures do not result in loss of data.
  • A NAS appliance comes with a GUI based web based management console and hence can be centrally accessed and administered from remote locations over the TCP/IP networks including Internet/ VPN / Leased Lines etc.
  • Some entry level NAS systems come bundled with backup software which can be very useful for backing up critical data and retrieving them when required.
  • NAS appliances are connected to the Ethernet network. Hence servers accessing them can also be connected to the Ethernet network. So, unlike SAN systems, there is no need for expensive HBA adapters or specialized switches for storage or specialized skills required to set up and maintain the NAS systems. With NAS, its simple and easy.
  • Ethernet networks are scaling up to support higher throughputs – Currently 1 GE and 10GE throughputs are possible and very soon 40 GE and 100 GE throughputs would be supported through a single link. NAS systems are also capable of supporting such high throughputs as they use the Ethernet based networks and TCP/IP protocol to transport data. Also, the interoperability of various vendor devices on the TCP/IP network is already proven.
  • The management tools required to manage the Ethernet network are well established and hence no separate training is required for setting up and maintaining a separate network for storage unlike SAN systems.
  • Storage Network convergence is happening all around the world and the convergence is happening over, no wonder, the IP Network! Standards like iSCSI, FCoE etc enable SAN and NAS to coexist on the same IP network, protecting existing SAN investments. Advancements in the TCP/IP networks like Lossless Network standard and Pausing transmission from individual devices to avoid congestion etc, would make TCP/IP network as reliable as Fiber Channel network in the near future.

Dis-advantages of Network Attached Storage (NAS, compared to SAN):

  • Transaction intensive databases, ERP, CRM systems and such high performance oriented data are better off when stored in SAN (Storage Area Network) than NAS as the former creates a network that has low latencies, reliable, lossless and faster.Also, for large, heterogeneous block data transfers SAN might be more appropriate.
  • At the end of the day, NAS appliances are going to share the network with their computing counterparts and hence the NAS solution consumes more bandwidth from the TCP/IP network. Also, the performance of the remotely hosted NAS will depend upon the amount of bandwidth available for Wide Area Networks and again the bandwidth is shared with computing devices. So, WAN optimization needs to be performed for deploying NAS solutions remotely in limited bandwidth scenarios.
  • Ethernet is a lossy environment, which means packet drops and network congestion are inevitable. So, the performance and architecture of the IP networks are very important for effective high volume NAS solution implementation at least till the lossless Ethernet framework is implemented.
  • For techniques like Continuous Data Protection, taking frequent Disk Snapshots for backup etc, block level storage with techniques like Data De-duplication as available with SAN might be more efficient than NAS.
  • Sometimes, the IP network might get congested if operations like huge data back up is done during business hours. That is not the case with SAN systems as it creates a separate network for storage and hence backup/restore operations can be done at any time of the day without affecting the normal functioning of the IP network.

Network Recommendations and Best Practices for Ethernet Storage over TCP/IP Networks:

  • Use Manageable Switches throughout the network – This is important for a lot of reasons, click on the link to know more. And besides, the price difference between unmanaged and managed switches is coming down fast.
  • The Network architecture should adhere to some type of established topology – Star (preferred), Extended Star or Ring. Try not to connect switches arbitrarily throughout the network as trouble shooting such a network would become very tough.
  • Use a Network Management System or tools like Cacti/ sFlow/ Netflow etc, to have a complete view of the performance/ congestion in the network.
  • Implement VLAN’s throughout the network to logically segment the network according to the various departments/ functions etc, for both restricting the broadcast domain as well as restricting the network access.
  • In large networks, use VLAN trunking where ever necessary to allow multiple VLAN’s to travel over a single high speed link (10 GE link, for example).
  • Also use Link Aggregation techniques to combine multiple links to form a single high speed logical link and use Spanning Tree Protocol (Or RSTP) to avoid loops in the network (which might cause congestion) as well as provide redundant paths to compensate for cable cuts or link failures.
  • Use Jumbo Frames (9000 byte payload) instead of the normal Ethernet Frame (1500 byte payload) where ever possible to communicate with the storage device – This method reduces the number of packets that need to be processed resulting in higher efficiency and higher effective throughput.
  • Configure priority groups using IEEE 802.1q, thereby creating end to end QoS settings for various forms of traffic (including storage) traversing through the network. This would give more control over parameters like latency and bandwidth reservation which is necessary for high performance storage networks.

excITingIP.com

In case you have any points to add to this topic, please add them using the comments section below. In case you have any questions, you could contact us using the contact form. You could also subscribe with your email address in the box that says ‘Get email updates when new articles are published’ to keep updated on the various network technologies.