ASK THE EXPERTS

When implementing an iSCSI solution, what are some of the best practices in managing the network?

By Jay Kramer, VP Worldwide Marketing, iStor Networks


So often we think of only the storage component when considering the implementation of iSCSI but the network topology is of equal importance for a successful implementation. One of the incredible values of an iSCSI solution is that it can utilize a ubiquitous IP network for a consolidated shared storage implementation. How that network is set up can have a significant impact on meeting service level objectives for performance and high availability. Here are a few best practice considerations:

Consider the configuring of VLANs as a simple technique to isolate servers on the network. This creates advantages for enhanced security of server access to specific storage targets and also performance acceleration by managing the traffic more efficiently.
  • The integraStor™ family of products uniquely provides the capability to create up to 8 VLANs for optimizing the storage network.
  • Competitive systems are typically only able to support up to 2 VLANs and more limited in network management capabilities.
Consider the implementation of host-centric techniques for dynamic load balancing and fault tolerance where there is more than one path between a server and a storage array.
  • Link aggregation (channel bonding) is the technique of taking several distinct Ethernet connections and making them appear as a single link. The integraStor family of products supports Link Aggregation Groups (LAGs) for up to 8 iSCSI (Ethernet) ports. These 8 ports can be grouped in a variety of configurations based on the service level objective of the customer application(s) in the deployed environment. (e.g., four groups of 2 ports each, one group of all 8 ports, etc.).
  • IntegraStor products can also be implemented with either Multiple Connections per Session (MCS) or Multi-Path IO (MPIO). MCS provides multiple TCP/IP connections per initiator / target iSCSI session to optimize bandwidth and network resiliency. MPIO provisions multiple initiator / target iSCSI sessions per LUN in order to enable network failover resiliency at the server.
These two techniques offer the customer the following advantages:
  • Achievement of a highly resilient storage network; should a connection fail, then the data can utilize an alternative path.
  • Achievement of greater performance by balancing the workload across multiple iSCSI ports.
  • Achievement of consistent throughput at peak workload times providing a quality of service that meets the most demanding processing environments.
Consider the implementation of a TCP/IP offload engine (TOE) to accelerate the throughput of data across the network by offloading the storage arrayÕs TCP/IP overhead tasks to a separate functional processor.

  • integraStor products have a built in TCP/IP offload engine which facilitates full wire speed performance and optimizes the data movement for maximum throughput.
Consider the implementation of Jumbo Frames and Flow Control for performance optimization. Larger frames require fewer frames to transfer the same amount of data. If the servers, switches and storage in the configuration all support jumbo frames, then the host CPU utilization can be reduced and the data traffic can be accelerated through the network. Flow control helps to optimize the information transfers at peak times.

  • As a best practice, the network switch typically needs to be configured for Jumbo Frames since they usually default to a smaller frame size. When purchasing or implementing a storage network, make sure the switch has enough buffer memory and can support both Jumbo Frames and Flow Control concurrently. integraStor products support both Jumbo Frames and Flow Control and the integraStor 10G architecture can exploit these capabilities for a higher throughput end-to-end customer solution.



» Return to the front page
UPCOMING EVENTS
VMworld Europe 2008
February 26, 2008
Cannes, France

Microsoft Windows Server 2008 Launch
February 27, 2008
Los Angeles

March 4, 2008
New York, NY

March 11, 2008
Chicago, IL

March 13, 2008
San Francisco, CA

March 18
Boston, MA
Detroit, MI

March 20, 2008
Houston, TX

March 25, 2008
Washington, DC

April 1, 2008
Minneapolis, MN
Phoenix, AZ
Seattle, WA

April 8, 2008
Dallas, TX

April 29, 2008
Atlanta, GA
HELPFUL LINKS
CONTACT US
iStor Networks, Inc.
7595 Irvine Center Drive
Suite 100
Irvine, CA 92618

949-753-8999 TEL
949-753-1068 FAX

Email: info@istor.com

Web: www.istor.com