Following my presentation at Cloud Expo: The Cloud Needs to Look More Like the Data Center, I contributed the following article ‘Storage: You Once Were the Weakest Link’ to Cloud Computing Journal, describing why storage challenges in the cloud had arisen from critical business concerns and kept enterprises from migrating their applications to the cloud.
One key element that is often missing in the avid discussions on what’s keeping more enterprises out of the cloud is a common recognition that storage in the cloud is simply not what it should be. It may surprise you, but to a large degree the solution to some of the cloud’s biggest concerns – such as security, compliance, control and complexity – has to do with storage.
So what’s wrong with storage in the cloud?
Unlike compute or networking in the cloud, storage in the cloud is very different from the storage enterprises use in their own data centers. If we look at cloud compute, the VMs are exactly the same as those you would provision in your own data center; networking is very similar: Internet Protocol (IP) and Virtual LANs let you connect machines to each other with reasonable performance and low latency. But storage in the cloud is not the same as that in an on-premise data center.
Enterprise storage in data centers is usually comprised of block- or file storage sold as a SAN or NAS system that offers high performance, granular control, and data management tools. However, cloud storage is almost always an entirely different type of storage: object storage. Rather than using iSCSI, NFS or CIFS protocols that are commonly used by most applications and operating systems to access storage, object storage uses HTTP as its protocol, just like browsers do. It can do two things: upload or download; put a file in, or retrieve it. For static content like website assets, video, or Facebook, object storage is absolutely great, as this type of storage is incredibly efficient for cloud-scale content. For example, Netflix is said to run on object storage. But with object storage you can’t perform any edits or changes to just a section of the file, as required in transactional applications such as databases, CRM, financial software, and collaboration suites.
However, for critical business applications and databases as they exist today (and have existed in the enterprise for decades), clouds offering only object storage are not meeting enterprise applications’ needs.
Solutions with compromise
Some clouds offer enterprise-friendly storage by employing physical SAN or NAS arrays. However, these solutions come with quite a few compromises: private storage is often not on-demand and requires a long-term commitment; performance and high availability are frequently mutually exclusive; volume sizes are often limited; and most offerings cannot deliver critical enterprise capabilities such as clustering and shared volumes. Other clouds offer block volume services, but those, too, suffer from limited volume sizes and an inability to cluster and share volumes. In addition, in order to provide file storage, the customer must install and operate third-party software, which increases complexity, VM costs, and, at times, licensing costs.
The following storage impediments are not just minor technical obstacles, but directly correlated to the top ranking public cloud hosting concerns expressed by enterprises, and keeping them out of the cloud:
Security: In addition to specific methods of physical security and encryption of data at-rest and in-flight, two key elements found in on-premise storage would go a long way toward achieving the security, privacy and control required in the cloud:
- Isolation – storage needs to be private and dedicated, where resources (drives) are never (ever!) shared among multiple users, and users know exactly where their data is stored. I’d like to envision isolation such that users are able to physically obtain their drives (and data), if needed.
- Encryption has to be enabled by customer-owned and managed keys that are not available to the provider – or anybody else – which means no one other than the customer can ever access the data.
Compliance: Companies in highly regulated industries, such as financial services and healthcare, need to comply with rigid regulations. Many of their applications and management tools run on high-performance block storage, have access to NFS or CIFS protocol or require true high availability of up to five 9s, usually achieved by clustering of shared servers (something not available in most cloud offering). Yet again, storage ends up being a cloud Achilles’ heel that makes any such deployment extremely cumbersome, costly, or simply impossible.
Control: Cloud storage usually offers users little to no hardware control. Any measure of elasticity usually refers to volume size only, and to a usually small predefined limit. The elasticity rarely, if ever, extends to performance. Common data center storage features, such as thin provisioning or remote replication, simply are not available. CIO, CTO & Developer Resources.
In addition, truly controlling how data, applications, and databases perform in the cloud means establishing consistency with the resources that IT managers already have on premise. The administrator also needs to be provided with a clear overview of the resources used and the tools to granularly manage, scale, and control everything from the physical drives’ types, to their capacity, speed, utilization, RAID configuration, IOPS, throughput and advanced caching configurations (specifically with SSD cache).
Complexity: Incompatible environments, performance bottlenecks and proprietary software stacks may necessitate re-architecting existing applications to fit the cloud storage environment – a non-starter given the typical IT team’s many other priorities.
Getting storage right in the public cloud
To help enterprises move more applications to the cloud and take advantage of its pay-as-you-go flexible model and resulting agility and spending efficiency, the public cloud needs to offer storage that enables the same experience as on-premise NAS and SAN systems. It needs to offer comparable levels of performance, control, security and compliance to reduce the risks – driven by change and complexity – associated with migrating existing applications and databases to the cloud.
Furthermore, when storage in the public cloud looks and behaves like on-premise storage, it also means enterprise data center managers and their vast knowledge, experience, and tools are applicable in the cloud as well. This delivers a real game-changing advantage for enterprise adoption. For example, newer approaches from Zadara Storage have allowed certain users to reduce total storage by costs by 50% simply by enabling NFS support – or gain an element of nimbleness in meeting client needs that was otherwise impossible by enabling clustering in the cloud.
It’s indeed possible to have it all with cloud storage – QoS, scalability, security, and control, in a pay as you go, bill-me-by-the-hour business model. Instead of having storage be the reason enterprise applications are hesitating to move to the cloud, the right enterprise cloud storage is becoming the reason behind many organizations’ vastly more competitive cost, agility and performance.
– –
Has the competence of cloud storage kept your business from migrating applications?
Let me know in the comments below, or join me on Twitter @NoamShen
