NAS Connectivity Architecture: Software

TCP/IP is the third of the three major components supporting communications within a diversity of transport media; however, the most popular data communications protocol remains Ethernet. Support here is evolving with increasing bandwidth to support Gigabit Ethernet standards, sophisticated participation within enterprise networks with quality of service (QoS) and routing participation. However, it should be noted that one of the limiting factors of file I/O within a TCP/IP network environment is the processing of the TCP layers for communications between client and server (for example, NAS device) elements. Note that this should reflect the TCP Off-load Engine (TOE) technology and other initatives such as InfiniBand that are evolving to optimize this latency.

To understand the relationships of network software and NAS we must describe the basic concepts behind the transport software that support the transmission of data throughout a network. As we discussed in Part II, the blueprint for open networks is the OSI layers of network processing. Parts of the OSI layers have been discussed in the previous sections of this chapter in explaining how the network devices work within LANs. We left that discussion within the transport layer, or layer 4.

We will begin the software discussion at the network layer and move up through the additional layers of the OSI model, discussing transport layer, session layer, presentation layer, and application layer. Most of our time will be spent on network, transport, and session layers due to their contribution to TCP/IP processing and their effect on I/O processing within the NAS model. However, it is important that we also mention the session and presentation layers and their effect in transporting the file level protocols, such as HTTP and FTP, and the subsequent interchange between applications and presentation layers.

The OSI model represents all the required processing for making a connection within a network as well as the necessary connections to applications that occur within a client/server computing model. Figure 12-9 shows layer 3, the network layer, and the remaining layers of the model: transport, session, presentation, and application. Each represents an important aspect of processing applications within the client/server model. More importantly, they represent a focused set of activities as they relate to NAS devices, along with their integration and interchange with the NAS micro-kernel .

Figure 12-9: OSI model with layers 4 through 7

An Overview of TCP/IP Processing

We begin the following overview by taking a look at how TCP and IP work to encapsulate and transport both data and messages throughout a network. TCP/IP supports data transmissions and functions within a switched packet network. This means that IP provides the packets and associated delivery information to the network. As discussed in the network device overviews, many of the switch and router devices derive their function from IP addresses.

Although we have already referenced IP in association with TCP/IP, the function of the Internet Protocol (IP) component is critically important to the transmission of data. IP functions as the envelope for application communications within a network. To that end, it provides an address, consisting of both network and computer addresses, for delivery within the Internet networking topologies. Using the IP address provides a means of locating any computer within the universe of the Internet and local networks that is designated to receive the IP packet.

All computers running TCP/IP software have IP addresses. These are assigned when the TCP/IP software is installed and initialized . The IP assignments become part of the computers network name and specific address within the network. The addresses, being a collection of numbers , rely on processing by the network devices to evaluate the IP address and route the IP packet to the appropriate network, and subsequently to the specific computer.

NAS functions on a TCP/IP communications basis. Therefore, NAS devices run TCP/IP and consequently play by the same rules as other computers that are addressed through IP address schemes. There are several considerations to this, howeverthe most important being the placement of NAS within the network segments due to its IP addressability and transmission of I/O file content within the IP packet.

The considerations are the placement and usage of the layer 3 and layer 4 network devices which can have a positive or negative effect on an NAS device. Well look at specific performance conditions in Part VI, however its important to understand these devices as they operate with an NAS device. Figure 12-10 further illustrates our example network with NAS being fronted with both the layer 4 device, supporting UNIX applications servers, and the layer 5 device, supporting e-mail and web servers. The layer 4 devices examine the IP packet as it comes into the network segment, and then route it according to the routing tables it maintains. This allows some requests to be transmitted directly to the NAS device for I/O file processing without them being redirected from the server. Such well-performed tasks illustrate the real value of the NAS solution. Some requests, however, are more generic and are sent to the application server where theyre redirected to the NAS device. This is a function of the application and its related request.

Figure 12-10: Network operational aspects of NAS deployment

Also a function of the NAS device to support the application server I/O requests, some degree of network efficiency can be derived by processing I/O requests directly through to the NAS device. However, layer 4 network devices only route the IP address and dont provide any additional intelligence as to its routing processes.

Layer 5 network devices may be more appropriate when trying to mediate complex data access requirements regarding application servers supported by NAS devices. One way this is done is through load balancing, where layer 5 network devices working at the session and presentation level can derive information about its network traffic and route accordingly . Therefore, in some instances, NAS devices can participate in load balancing by routing all explicit or some masked inexplicit I/O file requests directly to the NAS server.

A client request for data uses TCP to encapsulate and break the request into packets and then passes them to the IP functions. IP provides addressing information in the form of headers regarding the destination of the packets. Another important distinction of the TCP processing is the error checking it provides, especially for datacentric requests.

TCP uses a common error algorithm called checksum to calculate the data within the packet. When the packet has reached its destination, the IP functions verify the correct address and pass the packet to TCP. TCP, on the receiving end, breaks the packet and reassembles the initial client request. On the receiving end, the TCP checksum error routines verify that the correct amount of data was sent. If not, TCP returns the packet to IP, which requests a retransmission from the sending client.

The effect of the IP and subsequent TCP processing places an additional workload on both the sending and receiving computer. In the case of a server, the overhead can be extremely high given the amount of TCP/IP processing required with multiuser access. Given that the majority of overhead occurs within the TCP and IP software running on a particular server, this can become critical when considering the case for NAS.

There are important points regarding NAS that need to be thought through when working with TCP/IP processing. First, is the number of concurrent users possible with the workloads. This is actually the number of client or server redirect I/O requests planned for the NAS device. Placement within a network segment must balance the capabilities and resources of the NAS device. An obvious example would be the placement of an NAS appliance device in a network segment that handles an enterprise workload. In addition, placement within a network segment is imperative given it could easily become overloaded if the workload expanded beyond the capabilities of the NAS device. This is characterized by the placement of NAS appliance devices in network segments served by hub devices which are extended by additional client PCs or servers. The shared bandwidth of these segments can quickly overwhelm a network segment with NAS workloads.

NAS Connectivity in Operation

Our previous example of a sample network illustrates NAS I/O operations as they relate to connectivity. Figure 12-8 shows how the network operates, using sample workloads that take place in the user departments as well as in the data center. The Accounting and Sales departments are good examples of NAS appliance deployment, while the more heavyweight workloads in the data center depict mid-range and enterprise NAS solutions.

Operational aspects of Figure 12-8 include the following:

• NAS Appliances   These devices are placed within network segments that contain traffic within the local users. The devices also can be reached through administrators and support personnel for problem and maintenance activity. They become problematic with data protection (for example, backup and recovery), and are scalable within their segment.

• NAS Mid-Range   Data center devices that support the Sales and Marketing departments are greater in both capacity and data access resources (that is, the number of network interfaces). These devices also support inquiries through WAN connections and, therefore, must have additional security both prior to access and as the requests are processed within the NAS device. The scalability is much greater here given the layer 3 and layer 4 switch and router processing, not to mention the bandwidth of the data center backbone. Also, the mid-range NAS appliance can be upgraded in place to achieve larger storage capacities and data access network interfaces.

• NAS Enterprise   The other data center device supports more intense workloads by housing the data tables for data mart processing of the marketing applications. Although the access is limited to marketing and executive users, the complexity of the processing and size of I/O access is much greater than what we see with both NAS Mid-Range and NAS Appliances. Again, the scalability is handled through the bandwidth of the data center backbone, the limited number of users, and layer 3 routing. However, this solution becomes problematic with backup and recovery given its relationship to database systems running on an application server. A related problem is the capability to provide an effective disaster recovery solution of this magnitude over WAN mirroring functions due to bandwidth and size considerations.

• NAS Web   This solution, although similar to the mid-range NAS devices, is singled out due to its ability to respond quickly to the Internet type of data access, as well as the volatility of storage requirements. This solution is well placed to take advantage of both layer 3 and layer 4 network devices by providing levels of workload balancing and caching throughout the web server configurations.

Network Configurations and Workloads

The previous sample network also provided a basis to illustrate NAS placement as it relates to I/O workloads. Figure 12-11 shows how the I/O workloads operate within their specific placements. Our sample workloads form an important aspect to NAS placement decisions as they relate to user departments or placement within the data center. The accounting and sales departments are good examples of simple file processing (Sfp) using NAS appliances, while the more heavyweight workloads in the data center provide examples of both complex file processing (Cfp) and network file processing (Qfp) using a depiction of the mid-range and enterprise NAS solutions.

Figure 12-11: Network workload aspects of NAS deployment

The workload aspects of Figure 12-11 include the following:

• NAS Appliances   Workload descriptions using Sfp

• NAS Mid-Range   Workload descriptions using Qfp

• NAS Enterprise   Workload descriptions using Cfp

• NAS Web   Workload descriptions using Qfp and Cfp