Essential Technology: A GCS Blog

I am pleased to announce that GCS has achieved Professional Partner status with VMware. Though a single step in our long term VMware strategy, this status enables GCS to branch into other VMware product lines beyond vSphere 4. GCS is currently piloting VMware View 4 with several clients and looks forward to deploying desktop virtualization around this platform. GCS deploys server virtualization solutions built on both VMware vSphere and Microsoft Hyper-V.

By far our clients and prospects have two SANs in mind, the Dell Equallogic and the HP Lefthand. While both are impressive SANs that are very similar, there are some critical differences.

Clustered Storage vs Cluster-able Storage: Both products offer clustered storage but only Lefthand can tout that out-of-the-box. Each Lefthand includes two chassis by default where-as Equallogic offers dual-controllers and an extremely reliable single chassis configuration. What is the actual difference in reliability? I have not found any real-world tests but if I can get two for the same price as one, I'll go that route. Advantage: Lefthand

Licensing: Both Dell and HP have listened to their clients anger at complicated licensing systems. Both the HP Lefthand and Dell Equallogic offer an all-inclusive licensing method. Unfortunately unless you buy the HP Lefthand Starter SAN Solution, you may need to be the Multi-Site/DR license in the future for real-time failover and failback. Advantage: Equallogic

Density: The HP Lefthand is essentially a ProLiant DL320s server. The Dell Equallogic is custom designed chassis. Because of this difference the Equallogic gets more spindles per U. This is certainly a consideration when you're ordering racks of SANs. It is much less of a concern when order a SAN or two. Advantage: Equallogic

Groups: EqualLogic PS series allows only two SANs in a group. Data cannot span more than these two SANs. With HP's SAN/IQ Network RAID you can span multiple SANs in a variety of configurations. This provides better performance and reliability. Advantage: Lefthand

Site to Site Replication: Out of the box both SANs offer site to site replication. Only the HP Lefthand supports synchronous data replication with automated failover and failback. The Lefthand supports multiple sites in all configurations. Advantage: Lefthand

Obviously a lot of features are excluded as they are quite similar between both products. The reason GCS chose to emphasize the HP Lefthand SAN was a significant price and value advantage. However it is not quite so apparent as when compared to other SAN vendors on the market.

Here is a helpful comparison chart for HP and Dell's iSCSI SAN lines:

This brief guide will outline the different SAN classes available to the SMB buyer.

Entry Level: If you're buying a SAN, you need a few basic features. iSCSI support and vendor certification (Microsoft, VMware, Citrix) will meet minimal needs. These will enable your virtual environment to utilize high availability features in any of those vendor's environments. HP's MSA 2012i G2 and Dell's MD3000i both meet these criteria. Expect to spend as much as $15k on these devices. The HP will let you mix and match SAS and SATA drives in a single chassis for best use of your SAN dollars. Neither of these solutions offer thin provisioning or clustered storage. What's that mean? Less efficient per GB and much greater risk of failure. If you're concerned about putting all your VMs in one basket, and you should be, then look to the Mid Level, below.

Mid Level: This is what you want, if you can afford it. Two major features enter play here: thin provisioning and clustered storage. I'll touch on both of these now with more to come soon. Thin provisioning allows for oversubscription of storage. Don't worry about it, just do it. Clustered storage is like clustered servers. Two, or more, boxes configured for failover. The HP Lefthand provides the lowest entry cost to true clustered storage. Dell's Equallogic, Compellent, NetApp, Xiotech and others each offer some unique features. Expect to spend at least $30k on this device.

Enterprise: Forget about it. You can't afford it and wouldn't fit in your server closet if you could. Vendors such as HP, EMC, IBM, NetApp and others live here.

Virtual Storage Appliance: For those organizations that may already have a large investment in internal storage in servers or direct attached storage, a VSA may be the best bet. This software solution aggregates storage across your servers into an iSCSI SAN with similiar feature benefits to a full system. You will be able to support High Availabilty and VMotion/Live Migration with this solution. Obviously since you are only buying the software the entry cost is much lower than a hardware solution. HP's Lefthand offers a VSA for VMWare (Xen and Hyper-V are coming). StorMagic has an interesting option that currently supports only VMWare as well, but Hyper-V support is coming.

A not-too-brief market overview should whet your appetite. Look for more info on the features and key differences between vendors to come soon.

GCS' Early Happy Hour was filled to capacity by IT Executives from all over Central Texas. For many it was their first look at Windows 7. Many attendees agreed that the upgrade to Windows 7 is unavoidable. Following the demo, some attendees even looked forward to it! Our discussion highlighted the improved UI, Branch Caching, Windows XP Mode and AppLocker.

In addition to Windows 7 the attendees got a look at Windows Server 2008 R2. We spent quite a bit of time talking about Hyper-V, System Center Virtual Machine Manager R2 and the new Live Migration features. Marquis covered Clustered Shared Volumes in depth, as they pertain to Live Migration. We also covered the new Active Directory Recycle Bin and Branch Caching.

We wrapped up the event with a look at VMWare's VSphere 4. The new Fault Tolerance features garnered the most interest. VSphere's Fault Tolerance allows for failover between hosts without downtime. They have set the standard here. Unfortunately most of the new features in VSphere 4 are not included in the Essentials and Essentials Plus package which is the only version that is price competitive to Hyper-V.

We'll be hosting our next even soon, focused on Cloud Computing in the SMB environment.

We're finally seeing the ProLiant G5s and virtualization-enabled PowerEdges become widely available in the used server market. Many clients are asking for our recommendations. In a virtualized environment with VMware's High Availability or Microsoft's Failover Clustering, is used hardware an acceptable option?

First let me say that I've long been opposed to used hardware in traditional environments. The hardware represents a small portion of the overall cost of the implementation. The installation and configuration, whether performed in-house or by a consultant, is wasted when the hardware has to replaced. I would rather see that service cost spread across four of five years of use. Why save a few thousand dollars to sacrifice a lot of ROI? Cash is the only valid reason. The difference between new and used gear is rarely enough to determine the fate of a project.

Now we have a highly available, virtualized system. We have architected enough capacity to easily run in the event of a single server failure. Are we more comfortable pushing the typical refresh cycles of servers? YES! If we're only refreshing from a fear of hardware failure, then used equipment should be an option considered. In two recent projects we've been able to go from 20ish servers to two servers without buying a single new server! We're reusing equipment from the organization in these cases, but that compares well to procuring used hardware.

Make sure when ordering that your hardware is identical to take advantage of the high availability features. SANs are still new enough to have limited availability on the used market. If you can knock $20k or $30k out of the upfront cost by using used servers, why not?

HP's Proliant G6 maxes out at 144GB of RAM. 16 core processors are due out in 2012. How many VMs can you cram in a quad, 16-core box with a 144GB of RAM? Enough. Arthur Cole describes trends toward both mainframe-style aggregration and grids of smaller systems. Which is right your for you? One factor in determing the best strategy is:

Additional servers reduce the overhead for virtulization failover. Much like RAID configurations, with 2 servers you must allocate 50% overhead for failover capacity on each server. Three servers only need 33% overhead and so on.

A recent project has forced me to analyze two cloud computing providers - Terremark and Rackspace. In this post I'll discuss my analysis of the two services and use a recently completed Private Cloud project to provide cost comparison.

Terremark's Enterprise Cloud service fit the needs of my client best. This is a fully virtualized environment that completely divorces the client from hardware concerns. The Enterprise Cloud enables rapid deployment of Virtual Machines. All management services from the OS up and managed by the client.  It consists of HP ProLiant servers with a 3Par SAN. Redundant firewalls and load balancers are included in every package. All equipment is shared. Virtualization is performed using VMWare ESX 3.5. Server management occurs through a web application and includes many tasks that you would find in VMWare's management tools. Terremark offers packages at fixed Ghz and GB or RAM (entry is 5Ghz and 10GB RAM). You then add storage, backup and bandwidth. There is no incremental charge per VM, only for the resources it consumes.

Rackspace's Platform Hosting takes a different approach. They supply a mix of dedicated and shared resources to create a similar environment. Firewalls, load balancers and servers are dedicated. SAN resources are shared. Rackspace allows access to the VMware infrastructure tools so all functionality is exposed. Rackspace's price plans vary based on hardware redundancy and resources required.

Some interesting points:

• Terremark offers technical support 12 hours per day. Rackspace's support is 24/7/365.
• Terremark is hardened to withstand a CAT 5 hurricane. Good thing since they are based in Miami.
• Pricing favored Terremark by 25-30% when comparing environments with similar features.
• Network configuration is limited at Terremark. There is a maximum of 4 VLANs.
• Both solutions offer use of existing licenses or their hosted licenses
• Neither were excited about voice applications due to QoS concerns.
• Rackspace does not offer a backup service in the Platform Hosting option. Backup will require a dedicated server and regular offsite downloads.
• Rackspace included 6 TB of download and unlimited upload in their base plan. They claim it was the equivalent of 20Mbps sustained for the entire month.

Honestly I'm afraid I have to end this blog post by changing the name to a braindump. I am beginning a 30 day demo with Terremark immediately. I will continue to assess the two providers and post the information here. Look for follow up posts soon.