Home » Raritan Blog
June 29, 2010
Many IT professionals battle with the issue of gaining convenient “single pane of glass” access to multi-vendor and multi-technology data center servers, equipment and devices. Not only do an overwhelming majority of data centers include servers from different vendors, but they also require an array of methods to access and control them. Virtualized servers, embedded service processors (ESP), servers with KVM ports, intelligent PDU’s, and serial access routers all require frequent access and maintenance.
To meet this need, centralized management systems must to be able to support access to virtual machines and ESP-ready servers — in addition to the traditional KVM access to which we’re all accustomed.
Many administrators are implementing a hybrid approach, while others continue to expand on their KVM footprint because of its inherent performance, manageability and security features. Cost is also an obvious consideration.
Performance: ESP management features can be less “clean” and less convenient than a full-featured KVM platform. KVM solutions provide optimal video quality at any bandwidth, a broad choice of remote clients, and robust features like video scaling, full-screen mode, and Absolute Mouse Synchronization. There are also more choices when it comes to building redundancy and failover functions into the solution.
Manageability: KVM switches typically provide “plug and play” deployment features. And configuring a single KVM switch takes much less time than configuring 64 embedded processors. Although server vendors claim automatic ESP discovery, IP address ranges must be provided to the management interface. Also, each ESP physical connection requires network setup, user privileges, SNMP scripting, etc. Comparatively, a heterogeneous management system — if designed properly — can be set up quickly, provides thorough reporting capabilities and delivers centralized user rights management and authorization.
Security: Centralized authentication management is critical; not only for efficient user management — but also to ensure that the right people have access to the right equipment. Built-in ESPs are often lacking in this area.
Costs: ESPs often incur incremental cost for advanced features. They may be advertised as free, but embedded solutions may not include rights to advanced features; licenses for those are additional — as much as $500 per port. The average KVM over IP solution comes in at around $200 per server. Also, only one IP address is needed per KVM switch, while each ESP typically uses up another one on its own. The average TCO for a network port in a data center is $200. With the use of a 32 or 64 port KVM switch, a 32 to 64-to1 reduction can be made in IP addresses, network ports and cable runs.
Local Port: If there is a network failure, KVM solutions typically provide local access; embedded solutions do not.
Of course, some homework is needed to help determine if traditional KVM or ESPs are the better match for your organization. Whichever you choose, your management system should be able to accommodate both.
Dorothy Ochs
June 23, 2010
July 16, 2010
Hilton San Francisco
333 O’Farrell Street,
San Francisco, CA
Herman Chan will present “Intelligent Rack PDU’s – Key Learnings from Customer Case Studies.” Up to 90% of the power to mission critical facilities is consumed by powering and cooling equipment at the rack. In this session, Raritan will provide insights on how customers are utilizing intelligent rack PDU’s to better understand their power utilization, make better use of power and cooling capacity, and drive energy efficiency initiatives to reduce carbon footprint and save money.
June 21, 2010
One of the most important considerations in designing a Raritan enterprise datacenter management system is the architecture of the CommandCenter systems. This discussion provides an overview of how CommandCenter works and how it can be effectively deployed.
CommandCenter Architecture Overview
CommandCenter provides a single point of user access for all devices and systems under its management. The CommandCenter appliance is based on a hardened and secured Linux kernel. The CommandCenter appliance has dual hard drives in a RAID 1 configuration, dual 10/100/1000 NICS, and available dual power supplies. It is a fault tolerant appliance that is not prone to Windows/MS issues.
The primary function of CommandCenter is to provide centralized access to all datacenter IT resources. It also provides centralized user authentication, authorization, auditing, and reporting. Users connect to the CommandCenter and are authenticated and granted access based on local CommandCenter policies. User authentication can be Local, Active Directory, LDAP, TACACS, Radius, Radius w/RSA, or any combination of these. Once users are authenticated, their session is redirected and is sent directly to the desired KVM/Serial switch or datacenter device. This keeps traffic local between the user PC and the KVM/Serial switch or datacenter device. This architecture means CommandCenter systems are very robust, very scalable, and require minimal LAN/WAN bandwidth.
Herman Chan
June 21, 2010
Saw some interesting news today in an article by Katie Fehrenbacher on the earth2tech.com site, “5 Green Data Center Startups You Need To Know”. What caught my eye was the excerpt on a company by the name of SeaMicro™. They seem to be the first commercially available server solution utilizing low-cost, low-power, high volume CPUs, for server farm and cloud computing applications. The SM10000 product takes density to a new level with support for 512 Intel Atom® low power processors in a 10U package. WOW! That’s a lot of processors. That’s over 2000 processors in a full rack! The SeaMicro website suggests that this product was designed to replace 40 1 RU dual-socket servers which typically take a full rack but require only 2KW, which they suggest uses 1/4 the power and takes 1/4 the space.
When we talk at trade shows and user groups, we often cite that it’s not all about processors but rather the efficiency of the servers, e.g. how much output or workload do you get per unit input or watt. And I often bring up the analogy of miles per gallon and having hybrid cars in your data center or SUV’s in your data center. Both get you from point A to point B but one does it in a much more efficient way. Since there is a huge cost to leave servers idle, a data center operator’s goal should be to get more utilization of their servers and introduce virtualization or if I use the car analogy, pack more passengers into the car to get from point A to point B. We also regularly reference spec.org’s spec power benchmarks, to attempt to get a baseline or common way to understand server efficiency by make/model/configuration. As I was searching the site, I found that a 1U, dual-socket, 6 core per socket PowerEdge R610 system from Dell can provide 2938 server-side java operations per watt. I wonder what the SeaMicro system can do? Kudos to the SeaMicro team for thinking outside the box, or in this case, in a new 10U box. Look forward to seeing some cost and performance benchmarking information to compare against what already exists in the industry.
June 21, 2010
Raritan booth #671
June 27 - July 1, 2010
Mandalay Bay
Las Vegas, NV