Cloud Storage: Questions to ask

Very short - just a few thoughts I don't want to lose track of. My company is partnering with cloud storage providers, so these are questions we have to ask on behalf of our customers.

1. What are their security models? 
2. How are they protecting your data? Is it replicated to multiple destinations?
3. How are they storing your data? Is it encrypted, and how are the keys protected?
4. Are they using your data for something they shouldn't?
5. How would large restores be handled?

SLA checklist:

1. Security and Privacy (see 1 & 4 above)
2. Redundancy (see 2 above)
3. Bandwidth (see 5 above)
4. Response time (more of a managed service item, as in "how long to initiate a restore request")
5. DR - service provider's ability to recover from their own disaster, and service levels should match your RTO/RPO.
6. Non-performance clauses - rebates, waive fees, termination after multiple failures
7. Access to service provider data facility
8. Certification (SAS70 or PCI)

Portable Datacenter - Cloud Datacenter

What if you want to use the cloud for a DR site? What are the security issues? Some quick thoughts:

Document your own security requirements and audit yourself (do you live up to your own standards?), and publish your security requirements to see if a “cloud” or even hosted solution will be able to meet your requirements.

I think that virtualized networking and security tools should allow one to build a self-contained virtual infrastructure that can be portable: move it from one site to another and preserve a secure perimeter, regardless of the physical infrastructure provider (another private datacenter, a managed hosting operation or a “cloud”). If that is the case, the security SLA of the cloud provider becomes less of an issue.

An example is a new technology for storage that encrypts the data written to the storage, regardless of the encryption offered by the storage provider (e.g. What if the provider key is compromised? No problem if the data is encrypted as it is written to “cloud” disk…).

I am thinking about a vision of a fully virtualized datacenter, including networking and security, that would allow it to be moved from site to site and preserve all elements no matter where it was moved to. This would be a sort of virtualized “pod” or “datacenter on wheels”. The wheels, in this case, would be the internet.

The portable (or virtualized) datacenter has its routing, switching, firewalls, IDS, load balancing, servers and storage self-contained in the “trailer”. All  the owner of this mobile datacenter would need is a place to park it (the Cloud), a connection to the internet and power.

So, if Altor (or something competitive), in combination with Cisco/Juniper virtual switches and VMware tools (vShield?), can provide that capability, does it matter (as much) about some of the detail about the Cloud hosting provider? When I deployed infrastructure building blocks to AT&T datacenters, I built my own network and firewall, and didn’t rely on AT&T to do that for me. What is the difference here? All I was concerned with in Secaucus or Watertown, with AT&T, was physical security. All datacenters meet that basic requirement, and all the tenants use the POP at the datacenter, but they all have separate self-contained networks for the applications.


Agree, disagree?

EMC Unisphere Quick Notes

EMC is announcing a new management interface for its mid-tier storage line, that sounds and looks very promising.

Unisphere is a complete integration with vCenter, along with consolidation of management of CLARiiON and Celerra. This is intended to replace Navisphere and Celerra Manager. This framework also manages RecoverPoint/SE and will include PowerPath and is linked to Replication Manager. It also integrates Analyzer and Quality of Service Manager. It is integrated into lots of support, including PowerLink, without having to log into it separately.

This is Adobe Flex-based, so pretty light weight (not Java). There is a client for Windows, and a server, like off-array Navisphere. This is helpful with geographically dispersed infrastructure It is web-enabled. You don’t have to upgrade to FLARE 30 by using the server client, which is the other way to start using Unisphere.

I will elaborate on this know when I know more.

Here are some links to whitepapers, flash demos and videos.

http://www.emc.com/products/technology/unified-storage.htm?pid=Cloudcampaign-unified_campaign-070610

http://www.emc.com/collateral/software/white-papers/h8017-unisphere-element-manager.pdf

http://www.youtube.com/watch?v=mACVdai9YwE (Unisphere demo)

http://www.youtube.com/watch?v=oKubyt6XBcI&NR=1 (vCenter plug in demo)

On another topic, quickly...

EMC is also building in compression on the CLARiiON to reduce capacity, in the 2:1 range, intended for use in production. Technically, this isn’t deduplication, but more like white space reduction. FCoE support will be included in Unified Storage, to be announced this fall.

Defining the Cloud: Litmus Test

Here is a collection of thoughts I pulled together from recent conversations with colleagues and customers, about what constitutes a "cloud solution" or whether an application is a "cloud application". I'd love to have some comments on this.

Cloud Storage
I have been working on uses cases for one aspect of the cloud, storage (besides backup). There are four that seem to make sense. Can anyone think of anything else?

1. Archival
o Geographic dispersal
o Data stored on low cost media
o Apply WORM policy for compliance
2. Web Content
o Content stored on low cost media
o Use polices to position data
o Users access data directly
3. Disaster Recovery
o Replicate Data into the Cloud
o Apply polices to geographically disperse data
o Production data can be recalled when disaster strikes
4. Dynamic Data Allocation
o Directed at dynamic content such as video and delayed live feed
o Data replicated in geo-specific clouds (i.e. NA, SA, EU, Africa)
o Rapid rollout and tear down required


Some other thoughts from a recent presentation I made:
1. Remember principles of data management
o Performance
o Capacity
o Workflow
2. Cloud Storage works when
o Performance is not an issue
o Capacity is needed
o It fits the workflow
3. Cheap storage is easy, we all can do this, but…
o Geographically replicated, efficiently managed, AND cheap?
4. A useful archive or deep storage tier
o Data mining & re-analysis done in-situ with local cloud server resources if needed
5. “Downloader pays” good model to offset costs

Is the Cloud "right" for you?
As we (data storage architects) attempt to design and implement solutions, we are trying to solve a customer’s business problems while setting them up with a foundation for “smart growth” in technology. We should be thinking about these questions:
1. Will the customer benefit from a self-provisioning architecture that allows business users to set up their own servers/applications/storage/networking, and does that make sense from a business perspective??
2. Will the customer’s business benefit from IT that grows and shrinks in performance and capacity as needed?
3. Will the business benefit from “pay for use” by the end user, as a way to track investment and IT resource utilization?
4. Does the customer need to allow access to its business from anywhere, at any time?
5. Should the customer IT infrastructure be rebuilt so that it consists of “pools” of virtualized computing, networking and storage; and does that make sense from a business perspective?

The answers to these questions are not always “Yes”. Most companies do not need to build a private cloud now, or in the immediate future, but they need to understand what these changes could mean to them, both as IT professionals and as business enablers.

By Who and When will the Cloud be Adopted?
As has been pointed out by others, adoption of cloud-like computing will likely take place within IT and at the consumer and small-business level first, although latency issues will remain for business computing in the public cloud until the technology finishes maturing and content can be made ubiquitous (cheaply pushed out to the edge where it is needed). IT staff at larger companies will want cloud-like computing in the datacenter to ease power and space crunches and make them more efficient with a self-provisioning, elastic, network-based infrastructure, but as EMC has noted, it is a bit harder to push the enterprise business applications into a cloud. That will come with time, as power, cooling and staffing resources force efficiency on small business and datacenters alike. It will enabled by the successes we can make happen now and as the “hump” of the bell curve moves closer to the elastic infrastructure and virtualized computing environment we now call the Cloud.

The Litmus Test
Ask yourself if any technology fits most or all of these requirements? If not, it is not a cloud technology.
• Elastic?
• Self provisioning?
• Pay as you go?
• Pooled resources?
• Ubiquitous network access?

Example: SharePoint
SharePoint would be a great application to have in the cloud, but the resource pooling and provisioning would be infrastructure based, and that is where the cloud technology surfaces.
A service provider could allow you to request what is needed to deploy or scale SharePoint, but what is the technology behind the portal? BMC? Surgient? Something else? What is the storage behind it? How will the content and computing be geographically dispersed so that latency won’t be an issue?

I think any web-based interface to an application can be considered “enabling”, since that meets the “ubiquitous access” criteria. What will it be used for? How will it enable self-provisioning? Is what is being done through the interface able to scale?

Ask yourself (or your customer) this:
It all comes back to the same thing…what are you moving (and why) from the old paradigm of datacenter computing and moving to the new paradigm, regardless of whether it is private, public or federated?

Virtualize Your Files with F5

The first of many postings on the gripping topic of...storage. Yes, there are a few of us out there, talking about one array vs. another, one new technology vs. another, puncturing marketing hype, blah, blah, blah.
I hope to stand out from some of the others, by virtue of the fact that I really worked as a storage and backup administrator for years, trying to figure out how expert and informed storage design and administration could help fulfill business needs and requirements. How much does that cute little modular array in the window really cost? How can I manage file systems as well as databases?
I work as a Solutions Architect for GreenPages, a consulting and integration firm based on an island in the Piscataqua River between Portsmouth, NH and Kittery, ME, possible the coolest location for a technology firm ever. Much of my work is pre-sales engineering and storage systems design (primary, backup and archive). I specialize in looking at the life-cycle of data and how to manage it from a cost/benefit point of view.
My first posting is a re-write of a recent webinar I did for GreenPages on File System Virtualization. My new stump speech is about managing file systems with the best technology I have ever seen on this, from F5's Acopia Networks ARX product line. Read below, and comments are welcomed.



How To Get A Grip On Storage Growth (and reduce costs)

Too much data! According a couple of recent surveys by IBM and other research organizations, up to 80% of data stored on hard drives is ”unstructured”: files, simply put; messy, unorganized, redundant and duplicated files. But you still have to back it up – over and over and over again.


To add insult to injury, you are running out of space on your file servers or NetApp or Celerra, and you need to buy more disk shelves and hard drives. Wait, isn’t that array coming off warranty and now you have to pay $20,000 for the next YEAR??? Someone just called and said the Windows file server that the corporate office is using is giving a warning that it is about to run out of space, but there is no downtime allowed and DON’T ARCHIVE ANYTHING. Did we forget to mention that you haven’t finished a full backup on time in a month? Budget numbers are due in the next two weeks. What will you do?

• What if you could find a product to help you get rid of the clutter and junk, reducing the use of your storage by 50%, without the end users even being aware that their old data is now on a SATA array stuffed with 1 TB drives?
• What if you could now complete your backups with hours to spare?
• What if you can finally prove that Marketing has been storing multiple copies of those huge PowerPoint presentations with wanton disregard for the impact on overall storage utilization?
• What if you could put off the storage expansion indefinitely?
• What if you could migrate the data off of that old array without any downtime, and push that boat anchor out to the loading dock and save $20,000 in your budget?

What if?


It will only get worse. According to IDC, way back in 2005, average growth rate of corporate data was 60% annually. If you need 10 TB today, you will need 16 TB next year, unless you take action. Your storage services model will NOT BE SUSTAINABLE without a new way of looking at managing storage.



File System Virtualization



We have virtualized servers, and we have virtualized storage. Now the time has come to tackle the messiest and biggest storage problem of all: cluttered, congested, sprawling file storage systems. All of the scenarios I described above are based on real life situations, with real GreenPages customers. What they all have in common, that file system virtualization can help them with, is files (lots and lots of them) being served to their end users via Windows Shares (CIFS) or NFS exports, on Windows, Linux, Unix, NetApp or EMC Celerra file servers. (Sorry, Mac and Novell users, we have to talk to you separately…).


What does “File System Virtualization” mean? It means that you separate the physical location of your files from the means to access them. Put in place a proxy to file systems, without disruption to your business. No agents, no forklift upgrades, no sweeping changes of your client drive mappings. This can be done at network speeds: the end user requests a file from his home directory, sending the request to the proxy appliance, which has mapped out the location of all of the files on every filer and server in the network. The request is terminated and a new one is created to fetch the file and return it to the end user, with no one the wiser. This is done largely in hardware, at RAM speeds. No latency.


Take it a step further. You scan those file systems, and move everything that hasn’t been touched or changed in a year off of your expensive/old/overloaded filer to some new inexpensive iSCSI storage (about $2/GB!!!), and you have freed up 65% of your storage! Back up the stale data once a month, slam shut that big backup window, call your storage vendor and tell them to cancel that order for the new storage array, or better yet, trade in the old Chevy for a new BMW and show off the latest in thin provisioning and built-in storage virtualization features, all without a “maintenance window”.
Now you have a real archive system that reduces the file system utilization without “stubs” or pointers! Migrations for replacing hardware or load balancing take place without the pain of Robocopy or Rsync.



What can do all this? What is the solution?


What can do this? F5 Acopia ARX Appliances, in combination with any number of second tier or nearline storage devices that your Friendly Neighborhood Storage Architect can recommend.

• Archive to Enterprise Vault FSA? Done.
• Archive to Data Domain? Done. (Oh, and this solutions REPLICATES your archived data to an undisclosed offsite location.)
• Tier your gluttonous Art Department’s image files to some inexpensive EMC iSCSI storage (yes, they make inexpensive storage….sheesh!)
• Trade in that old SDLT tape library for a sleek little LTO4 autoloader.

Well, this concludes the first ever Not Sexy blog. I hope this was informative and somewhat amusing. I look forward to your comments and questions.