Getting Thin for the Fall – Twice the Utilization at No Extra Cost!
Often I am asked what our 'virtualization strategy' is. Like it should be one product or one solution. The reality is that we all have the same strategy – to help customers mask the complexity of managing multiple instances and increase the utilization of the assets, at the same time. For Sun Storage this means – disk virtualization, tape virtualization and with our partners, fabric based virtualization.
It's worth remembering that Sun has an incredible portfolio of virtualization across Systems and Software – which combined with the Storage really provides an unparalleled solution design. It is also very Eco friendly – more in the next blog.
We have just added another part to our disk virtualization portfolio - the Sun StorageTek 9985V system. Like its bigger sibling the ST9990V , the ST9985V has the ability to virtualize multiple storage devices from all vendors and present them to the application as a single pool of resources. It also has the ability to let the application see more storage than actually exists on the physical disks within the system.
So what?
In the past, the number one priority was to ensure that the application had access to enough storage - no matter what. This meant that data center managers had to allocate extra capacity in case there was a spike in demand for the application, and to ensure that there was enough space for all the data copies needed for business continuity or to recover from a disaster. In addition, adding storage when a new application was deployed lead to lots of storage devices that had to be managed independently, that couldn't access each other's resources, and that all had a lot of overhead. There had to be a better way.
Technologies such as the virtualization and thin provisioning capabilities in the ST9985V can help alleviate this inefficiency by giving all devices the same management interface and by reducing the amount of physical capacity required to support the application. More importantly, these technologies can help data center managers move towards a tiered storage architecture where data is stored on a device according to a pre-assigned value set by the organization. Since both of these considerations change over time, it's important that the tiered architecture have the appropriate tiers (primary disk, secondary disk, virtual tape, tape archive, etc.) and you have the ability to move the data from tier to tier according to policy.
Bottom line: Thin provisioning can increase your disk utilization from under 30% to over 60%. That's more power to your bottom line not your next disk array!
Posted at 09:00AM Sep 10, 2007 by Nigel Dessau in Sun Storage | Comments[4]
How are you enabling organizations who cannot, through extensive internal purchasing processes, provision quickly enough once their TP'd storage reaches critical limits? Is Sun going to provide a utility-like model where customers can have storage capacity provisioned and true up monthly or quarterly. We've looked at this and honestly, having a horserace through procurement when you're running out of capacity is not what we want to go through.
I'll throw it out there that with the cost of disk these days the value of TP is diminishing. The complexity of moving data around via policy, which is not easy or cheap, is offset by ever decreasing raw disk prices.
Lastly, what is Sun doing in the area of interoperability? Is the long term plan, to front end "other people's storage"? How is support matrix management accomplished in this model?
Posted by Chris on September 11, 2007 at 12:11 PM MDT #
Nigel,
Thin provisioning is terrific and has been successfully adopted in many customer sites now for quite some time (e.g. 3Par). What's really exciting about this announcement is that you're combining thin provisioning with the capability to virtualize other vendors' storage arrays-- isn't this correct? But unlike most diskless virtualization appliances, if I want to pursue a true tiered storage strategy and include tier 1, I can use the internal (disk-full) device capabilities of your product.
Wikibon.org users tell me that they have these pain points:
1) Migration costs are killing them (up to $50,000 and 5+ months per array);
2) Virtualization appliances often cannot virtualize tier 1 because of performance concerns;
3) Virtualization appliances have limitations on LUN support requiring them to purchase more appliances and they can't virtualize across appliances...creating more 'islands.'
Seems to me a dead end strategy that your products address perfectly.
Am I missing something?
Posted by Dave Vellante on September 12, 2007 at 08:35 AM MDT #
Nigel,
Thin provisioning is terrific and has been successfully adopted in many customer sites now for quite some time (e.g. 3Par). What's really exciting about this announcement is that you're combining thin provisioning with the capability to virtualize other vendors' storage arrays-- isn't this correct? But unlike most diskless virtualization appliances, if I want to pursue a true tiered storage strategy and include tier 1, I can use the internal (disk-full) device capabilities of your product.
Wikibon.org users tell me that they have these pain points:
1) Migration costs are killing them (up to $50,000 and 5+ months per array);
2) Virtualization appliances often cannot virtualize tier 1 because of performance concerns;
3) Virtualization appliances have limitations on LUN support requiring them to purchase more appliances and they can't virtualize across appliances...creating more 'islands.'
Seems to me a dead end strategy that your products address perfectly.
Am I missing something?
Posted by Dave Vellante on September 13, 2007 at 09:33 AM MDT #
Mr. Dessau:
I have been a fan of AMD since the 486 days but have honestly bounced back and forth between Intel and AMD based on performance per dollar. I, like many fans, were pulling for you when you got into the quad-core market and are very happy to see the latest offerings.
With that, I offer you a "what I would do if I could make decisions at AMD"! The computing world is moving towards laptops. They are convenient and free us from the confines of our home offices. The market for home desktops is shrinking rapidly in direct correlation to sinking laptop prices. If AMD has a real opportunity to overtake Intel it is in the laptop segment and server. If I were making decisions, I’d scale back my desktop war with Intel and focus on servers and laptops. Both of these segments have one thing in common – low power requirements. If AMD can focus on producing competitive-level performance with superior power management in the server and laptop chip markets, you’ll have a rapid increase in market share. The days of the home desktop are coming to an end. In 5 years, most new computers bought will be laptops and “Netbooks” (EEE PC, Cloudbook, Intel Classmate, etc).
Home users want more battery time from their laptops. Corporate users want more energy efficient servers to reduce consumption as well as to tout Green IT initiatives.
Thanks for listening!
Eric Scarlett
Volvo IT North America
Posted by Eric Scarlett on March 28, 2008 at 07:02 AM MDT #