Vendor Voice: IP-SAN
IP-SAN: A complete Storage solution

Traditionally, IP-SAN has been used for file services. But it is actually capable of much more. by Pramod S.

There has been enough debate on the SAN vs. NAS issue. But those who have written about this have generally taken the traditional view of differentiating SAN and NAS. The aim of this article is to understand SAN and NAS from a functional point of view, and to look at IP-SAN not just as a File-server but also as a complete storage system, which can take care of other application needs too. Technologies like iSCSI, iFCP (Internet Fiber Channel Protocol) etc have been briefly discussed as these have not yet been widely accepted in the market, and the technology is also not completely mature.

Technically, a product might offer many features, but what is the point in buying something 'big' when those technical features cannot be converted to business goals? The technology that you plan to implement, should ensure the required uptime, give you sufficient performance, reduce maintenance, and be scalable. Overall, it should make a difference in the way you work, so that with the same or less effort you deliver more support to your business.

Defining SAN
A Storage Area Network may be defined as a dedicated network of servers and storage with:

• Any to any connection
• Multiple paths to all resources
• Open structure using industry standard protocols
• No nodal dependencies (it can function even if a node or two is down)
• High bandwidth and High availability
• Scales up with no performance loss

A SAN can be deployed in two ways: using Fiber Channel or the IP way. The first is called FC-SAN while the second is IP-SAN. IP-SAN can also mean the network built with iSCSI as the protocol. I will limit my discussion to NFS/CIFS as the protocols for the IP-SAN.

Let's examine the functional difference between FC-SAN and an IP-SAN.

FC-SAN
As seen in the diagram, in an FC-SAN the file I/O is between the application and the file-system, which talks to the volume manager, which in turn makes block I/O to the storage device. As we see here, the file-system and the volume manager (if any) are on the server and the storage has the RAID and the disks.

FC - SAN

Click on image for larger view

IP-SAN
The basic components of the system are the same, but here the application makes the file I/O to a file-system which is outside the server and the block I/O now happens within the storage between the volume manager and the RAID.

APPLICATIONS
Traditionally, NAS has been used for File services only and SAN for all other applications. But those were the days of File-services from servers and 10 Mbps LANs. With dedicated appliance and gigabit speeds on the wire, throughputs that can be obtained from NAS devices are really good. Applications have their own requirements or limitations in terms of the bandwidth that they use. Just because you have connected a 200 Mbps link between the server and the storage, the application server will not use the full bandwidth, it might not use half of it too—it all depends on the application requirements. An application might require about 20 Mbps bandwidth, so there's no point giving it 200 Mbps Also, the servers have to be big enough to pump so much of data in and out, a small or medium server will not be able to pump so much I/O. A detailed assessment of the I/O requirements and server capacities needs to be made before sizing any storage device.

IS PERFOMANCE EVERYTHING?
But why is performance associated with speed and power? What is overlooked is the performance the applications will require. Should one only look at storage for performance or look at other aspects like ease of use, recoverability, manageability?

Good performance can be achieved on both IP and the FC storage. Performance is definitely required when things are running fine—it should run at optimum speed as per business requirements, but performance is not everything. Performance is good as long as things are normal, but when there is a crisis or some failure, the credibility of the storage vendor comes into play, in terms of recoverability.

Traditionally, FC-SAN has been considered to be difficult and IP-SAN easy to deploy. Also FC-SAN is associated with interoperability issues, which the IP-SAN is not. Though this does exist to an extent, it is much different from what it was before. FC-SANs are coming up with management tools, which make deploying FC-SAN relatively simple and vendors on IP-SAN making their devices faster and faster so that there is little distinction on what can run on IP-SAN and on FC. But to date IP-SAN deployments are extremely simple and easy to deploy, manage and provide some exceptional recovery methods.

IP - SAN

Click on image for larger view

PERFORMANCE
Storage on IP-SAN has evolved to a large extent than what it was before. With performance getting closer to SAN storage, while still maintaining all the manageability, and recoverability features, they make an ideal choice for most requirements. Current IP-SAN technology can address 80 to 85 percent of the storage requirements. The rest is better on SAN because of specific requirements or application needs, but can still be met on IP-SAN if the investment in FC does not make business sense. There are many media companies who use IP-SAN for streaming audio/video data. There are also enough implementations of databases (OLTP & DSS) or ERP/Billing etc on IP-SAN. This is evidence that traditional approaches are things of the past.

If we look at a development environment where databases have been deployed for developing applications or small SAP implementations with 100 odd users, even an average size IP-SAN would be enough to provide the required performance. You don't even need a big IP-SAN device.

DAFS
DAFS (Direct access file system) an open protocol fore-fronted by IP-SAN vendors, has shown extremely good results on the industry standard TPC benchmarks for OLTP. The real world benchmark showed the best price/performance for a Unix-based system, the best per-processor performance for a SPARC-based system, and the best performance per CPU-MHz on SPARC ever achieved show that extremely high OLTP requirements can also be met on the IP-SAN where performance outweighs all other requirements.

Integrating NAS with SAN

Click on image for larger view

DEFINING NAS
NAS represents devices that attach themselves on the IP network and provide CIFS/NFS services. I would differentiate NAS from IP-SAN as IP-SAN being setup with a dedicated interconnect (Gigabit) between servers and NAS—when the same storage, is connected on the Non-Dedicated 10/100 Mbps Network. Performance is one of the main criteria of IP-SAN along with other features of IP storage. A typical setup showing an IP-SAN also being connected as NAS is shown below.

SUMMARY
Storage requirements need to be looked at not just from the performance point of view, but a holistic approach needs to be taken with business needs in mind. Traditional approaches of 'IP for file' and 'FC for databases' are passé. Decisions need to be made considering all the storage requirements of an enterprise (file services and database or other applications), and the infrastructure or manpower in place. The investment that needs to be made for the deployment of the new infrastructure is also an important factor.

Pramod S. is Systems Specialist Apara Enterprise solutions pramod@apara.com