Storage Networks

Demystifying IP SAN

by AJ Casamento on ‎03-16-2016 04:03 PM - last edited on ‎03-18-2016 11:43 AM by ShelbyKhan (10,193 Views)

Demystifying IP SAN

To paraphrase the immortal Will Shakespeare: A Storage Area Network (SAN) by any other name would still provide a shared resource for data storage. Which of course is very little help when you’re trying to decide how an IP SAN would stack up against Network Attached Storage (NAS), a Fibre Channel SAN, or even file services such as Common Internet File System (CIFS) or Network File System (NFS) in your particular environment. Here’s some information that should help.

 

IP SAN: The Good, the Bad, and the Ugly

People frequently associate the term IP SAN with an Internet Small Computer Systems Interface (iSCSI) solution. ISCSI is an IP storage networking standard for linking data storage facilities. With over a decade of development behind it, iSCSI offers some significant advantages. It provides block storage at a lower cost than Fibre Channel, leverages existing infrastructure and administrator knowledge, and isn’t subject to distance limitations. What’s important to note is that these advantages are only in play as long as there is no chance that the network will ever need to scale, and that performance is not a requirement. That’s because the marketing hype which states that vast scale and performance are achievable when storage and application traffic share an Ethernet cable is, well, hogwash.

 

The fact is that application response time is very heavily impacted by competing network flows because of the way TCP/IP— the suite of communications protocols that connect Internet hosts—works. Imagine a car that pulls in front of an 18-wheel truck on a highway. The truck driver brakes, causing the truck to lose momentum. The truck can slowly regain that momentum, but in the time required to get back to its original speed, the truck is moving more slowly than it was—and so is the traffic behind it. The same is true for the TCP highway. Application traffic tends to be random and bursty, while storage traffic tends to be more sequential. Mixed traffic patterns cause packet loss, slowing traffic in what’s called a TCP slow start. Like the cars behind the truck, network traffic slows.

 

That’s a problem. But it doesn’t mean that an IP SAN shouldn’t be used for business-critical, high-profile, bandwidth-intensive, latency sensitive applications. What it means is that the problem isn’t the IP SAN. It’s the mixed traffic. The solution is clear. Un-mix the traffic by creating a dedicated IP SAN.

 

Dedicated IP SAN Provides Scalable High Performance

A dedicated IP SAN eliminates the huge performance impact of TCP slow starts caused by mixed traffic, as these graphics show. The first two graphs depict a fairly consistent 20% database application performance gain when the shared network disruptions are eliminated by a dedicated network. The following measurements of the same system show how eliminating the latency spikes due to TCP slow start responses to dropped packets produces consistent performance in the dedicated network.

 

A dedicated IP SAN will provide a measurable and consistent performance advantage when compared to a shared IP SAN, every time, under any circumstance. And that advantage will scale with network growth. If you’d like more proof, I refer you to the below video named "Is Application Growth Slowing You Down?" from IDC.

 

 

Meeting Business Goals with an IP SAN

Life is unpredictable. That’s as true in the data center as anywhere else. Administrators don’t know what the next big application will be, or what new business challenges will affect network performance, capacity, and scalability requirements. Ten years ago, the concept of streaming the majority of television entertainment to private homes over an IP network would have been unthinkable. Today, it’s considered normal.

 

One thing is clear. Administrators want maximum performance with maximum flexibility. That means the freedom to place any application on any platform at any time, for any reason, with greater than 90% utilization. In the IP SAN world, that requires a dedicated solution. A dedicated IP SAN can provide consistently high performance today, along with the flexibility and scalability to meet the unimaginable challenges of tomorrow.

 

If you’d like to learn more about IP SAN and design considerations, check out:

http://www.brocade.com/en/possibilities/solutions/ip-storage.html?intcmp=lp_hp_ipstorage_bn_00002

 

 

Comments
by
on ‎03-17-2016 12:04 AM

IP SAN = Is a Welcome back to FCoE Second Generation ?

by qskills
on ‎03-23-2016 03:32 PM

Isn't it the challenge to run properly common network and storage traffice over one unified fabric? Otherwise I could leave my existing FC-SAN, which does storage networking way better than IP-based solutions? Running a dedicated IP-SAN has similar high costs and management overhead as it's with LAN and FC-SAN right now.

Top Kudoed Posts
Labels