Amazon’s recent announcement of the AWS Storage Gateway was a surprise to some and for others just a natural extension of the online retailer’s strategy to grab all of a customer’s data and store it in its Simple Storage Service (S3) cloud.
The Amazon Web Services (AWS) Storage Gateway, introduced in January, is the company’s first foray into the on-premises cloud-storage space. The gateway is intended to be an on-ramp into the Amazon S3 and EC2 (Elastic Compute Cloud) networks for storing file data for data protection and disaster recovery. It is a way for customers to “kick the tires” of cloud-storage appliances that will get them used to working with the cloud before they graduate to a more sophisticated hardware-based appliance that meets their needs for true enterprise-class cloud storage. Implemented as software that installs as an image on a VMware ESXi 4.1 virtual machine, the gateway connects to the network via the iSCSI storage protocol, where it can attach to direct-attached storage (DAS) and network-attached storage (NAS). It supports the CIFS and NFS file protocols. The offering is in beta now, and it has the ability to cache files on-premises on the road map.
Amazon isn’t alone, however. A number of vendors are attacking the on-premises cloud-storage gateway market, among them Nasuni, Nirvanix, StorSimple and TwinStrata. Originally these companies called their products cloud-enablement appliances, thereby drawing attention to the purpose of the product: to enable organizations to put data in the service-provider cloud.
Since the cloud-storage gateways were first introduced in early 2010, the market has evolved. Rather than offering simply software- or hardware-based gateways to cloud storage, vendors are now offering a more complete storage solution, in which the gateway plays not only as an on-ramp to the cloud but also as a local on-premises repository for unstructured data — files, spreadsheets, images — and structured transactional data — databases, ERP and CRM systems. Vendors are also adding enterprise features such as deduplication and encryption to their gateway products and targeting them for not only the online backup and disaster recovery market but also for archive as well as primary storage.
Use cases for cloud storage
According to recent research from Storage Strategies NOW, an industry analyst firm focusing on storage technologies, cloud storage is being adopted by North American organizations of all sizes. Of 187 respondents, 40 percent of enterprise-size organizations (5,000 to 100,000-plus employees) are deploying cloud storage, followed by 33 percent of midsize organizations (1,000 to 4,999 employees) and 27 percent of small businesses (1 to 999 employees). Further, they are deploying cloud storage as expected for email archives, followed by backup applications and primary-storage applications (front office and database applications).
It shouldn’t be any surprise that small businesses are among the first to adopt cloud storage for backup data and then disaster recovery. These businesses are often strapped with limited budgets and a lack of full-time, skilled IT staff, so they turn to clouds built or used by the value-added resellers (VARs) or managed service providers (MSPs) that manage their networks for support and deployment. In doing so — adopting cloud storage — they can easily and affordably replace shuttling tapes off-site for disaster recovery with cloud-based services that let them implement disaster-recovery strategies.
Table 1. What storage applications are you planning to deploy or have you deployed in the storage cloud? (Select all that apply.)
|Backup||Archive||Disaster recovery||Primary data|
Source: Storage Strategies NOW, 2011
Midsize businesses are more resistant to adopting cloud storage, except for backup applications, citing in the study that they will wait until this year through 2014 to deploy public-cloud storage. For enterprise-size businesses, between 2012 and 2014, 45.3 percent plan to use the cloud for backup applications, 15.4 percent for archive purposes, 18.2 percent for disaster recovery and 22.7 percent for primary storage.
Competition to the AWS Storage Gateway
By contrast to other vendors of cloud-storage gateways, the AWS Storage Gateway looks anemic. It lacks a number of enterprise features such as encryption, where the user holds the encryption keys, deduplication and support for a local data cache. It supports only iSCSI block storage, which makes file-level restorations impossible. And it lacks scale: Volumes up to only 1 TB in size can be created, and they can’t be resized after creation. The AWS Storage Gateway also supports a maximum of only 12 volumes, in contrast to vendors such as Nirvanix that have no restriction on volume size or number of volumes per virtual machine.
In addition, the AWS Storage Gateway is implemented only as a virtual appliance. While a virtual appliance can easily support small installations or branch offices within larger organizations, enterprise users often want a physical appliance that reduces contention for resources and can support primary storage, storage bursting or online transaction processing (OLTP).
Another weakness of the AWS Storage Gateway is the cost for the system. At first blush the $125 per month charge for the gateway appears inexpensive when compared to other vendors, which offer all-inclusive, one-price-per-month or -year services. But when a user cobbles together the AWS Storage Gateway with the Amazon S3 or EC2 clouds, calculates the costs of storing data in the cloud and of getting it out (additional costs such as the transfer-out fees that have been factored into other vendor’s offerings), and installs it into a VMware ESX 4.1–only environment, the costs can be steep. Imagine a small business restoring data from the cloud at a cost of $0.12 per GB of data stored.
Vendor lock-in is also inherent with the AWS Storage Gateway. Unlike products from TwinStrata, Nasuni and StorSimple, the gateway can be used with the Amazon S3 and EC2 clouds alone. This makes it difficult, if not downright impossible, for customers to migrate data between clouds run by different vendors if they are unsatisfied with the service. Nirvanix alone, like Amazon, only supports one cloud.
There are many other differences between the AWS Storage Gateway and other vendors’ products that we won’t discuss here but that you can see in Table 2 and Table 3.
Table 2. Cloud gateway features and characteristics
|AWS Storage Gateway||Nasuni Filer||Nirvanix CloudNAS||StorSimple 5020 and 7020||TwinStrata CloudArray|
|Cost per virtual gateway appliance||$125/
|$0||$0 with Nirvanix Cloud Services||$250/
|$4,995 for virtual appliance (no monthly fee)|
|Cost per physical gateway appliance||N/A||$5,000||$0 with Nirvanix cloud services||$30,000 or $70,000 for physical appliance||$8,995 for physical appliance|
|Use cases||Backup and disaster recovery||Primary storage, backup, disaster recovery, archive||Backup, disaster recovery, archive, storage tiering, collaboration||Primary, backup, disaster recovery, archive||Primary storage expansion, backup, disaster recovery, archive|
|Cost per GB
|Supported protocols||iSCSI||CIFS, NFS||iSCSI, REST, SOAP, NFS, CIFS||iSCSI||iSCSI|
|Encryption||No control||Y, user holds keys||Up to AES 256-bit encryption support||Y, user holds keys||Y, 256-bit AES, user holds keys|
|Resizable data volumes||N||Y||Y||Y||Y|
|Dynamic resizable cache||N||Y||Y||Y||Y|
|Snapshots||Block-level||File-level with versioning||Block-level with Panzura appliance on-site||Block-level with Windows Server file system||Block-level; 168/volume|
|Limited||Unlimited||Customers can select up to nine replicas and choose which data centers they reside in||Unlimited||Up to 168/volume|
|WAN-optimized restoration||N||Y||Yes, with Riverbed Whitewater appliance on customer site||Y||Y|
|Application consistency||N||N/A (only relevant to blocks)||Yes, file-based locking with Panzura appliance on-site||Y||Y|
|Supports local cache||N||Y||Y, with Nirvanix CloudNAS, Riverbed, Panzura, TwinStrata, etc.||Y||Y|
|Deduplication||N||Y||Y, via both Panzura and Whitewater appliances||Y||Y, via compression, deduplication and local caching|
|High availability||N||Y||Y, SLA support up to five 9s||Y||Y, optional on virtual and physical appliances|
|Bandwidth controls||N||Y||Y, with Riverbed Whitewater appliance on customer site||Y||Y|
|VMware-ready certification||N||N/A||Y , and with Veeam for virtual infrastructure management and replication/
|Microsoft certification||N||N/A||Y, via partners||Y||N|
|Multisite access||Y, optional||Y (read/
|Y, Nirvanix has a single global namespace||Y||Y, for disaster recovery|
|Local storage as cloud||N||Y||Y, via hybrid and private clouds||Y||Y, NAS storage, SLAs|
|Nondisruptive software upgrades||N||N/A||N||Y||N (auto-updates, restart required)|
|Max. volume size; number of volumes per virtual machine||1 TB/
volume; 12 volumes/
|As many as you want||Unlimited volume size, number of volumes, file size and number of files||>500 TB; 512/virtual machine||384 TB/volume; 128 volumes/
appliance; 50 PB total
|Cloud service providers supported||Amazon||Amazon, Windows Azure||Nirvanix||Amazon, AT&T, Windows Azure, Nirvanix, Rackspace, HP, Google, OpenStack, EMC Atmos, Dell/Caringo, Scality||Amazon, AT&T, Nirvanix, Rackspace, HP, EMC Atmos, Peer 1, Windstream, Mezeo, Scality|
Source: Storage Strategies NOW, 2011
Table 3. Cloud service-provider costs
|Amazon S3||$0.125||$0.12||$0.01/1,000 PUT, COPY, POST or LIST requests; $0.01/10,000 GET and other requests|
|Windows Azure||$0.14||$0.12||$0.01/10,000 requests|
Source: Storage Strategies NOW, 2011
And, finally, don’t forget the AWS Storage Gateway is a beta version. Amazon has promised that a future version of its software will allow on-premise caching, where frequently accessed data will remain on local storage and only the entire data set will reside in its cloud.
The future of gateways
The cloud-storage gateway as it evolves is being assimilated into many data-protection packages and hardware-based appliances as a feature rather than as a separate product. Although many times these are basic gateways that don’t incorporate enterprise features, they are the future of cloud-storage gateways as indicated by these announcements in the recent past.
Last year EMC introduced the EMC Cloud Tiering Appliance, which integrates with its VNX storage array, to enable the tiering of inactive files into the EMC Atmos cloud. EMC also has the EMC Cloud Backup Option for its data-protection product, Networker. CommVault has been shipping a cloud-storage connector for its Simpana data information management software that allows customers to back up and archive data into Amazon S3, Microsoft Azure and Nirvanix clouds. Symantec isn’t left out: With NetBackup 7.1, the company introduced a cloud connector for the Nirvanix cloud for backup. With NetBackup 7.5, the company is extending that to AT&T, Rackspace and Amazon S3 clouds.
Appliance vendors are also implanting cloud-enabled appliances. Quantum is expected to offer cloud backup with its DXi deduplication appliances, and Dell will integrate the cloud into its storage products as another tier of storage.
Adding the cloud as another tier of storage is perhaps an unrealized or at least unspoken goal of Amazon’s. The company, which at present stores over 762 billion objects in Amazon S3 and processes over 500,000 requests per second, intends to own not a portion of your data but all of it. The AWS Storage Gateway is its first step in accomplishing this goal.
The AWS Storage Gateway doesn’t signal the death of the storage gateway for all the reasons given above, but it provides validation that the vendors of gateway products — Nasuni, TwinStrata, Nirvanix and StorSimple — are headed down the right path with feature-rich products and all-inclusive cloud services. The challenge for startups like Nasuni, TwinStrata and StorSimple is to innovate faster than Amazon on features and to continue to partner with multiple cloud-storage providers, thereby offering a choice of storage clouds other than the Amazon-only model.