Just a couple of years ago, as solid-state storage technologies began finding significant mainstream adoption, Taneja Group began closely following a vendor whose architectural roadmap seemed to destine them to be the pre-eminent architectural leader for scale-out, high performance, enterprise-ready, cost-effective solid-state arrays.
That vendor was Kaminario, who first entered the market with a highly resilient, scale-out architecture that promised extreme performance with more linear scalability as well as superior availability / serviceability versus other offerings we then saw on the market.
In the past couple of years, Kaminario has continued advancing their technology in both performance and features, systematically adding the mainstream features that the enterprise demands – and are that are too often missing on high performance storage systems: features like snapshots, utilization reporting, resiliency that tolerates full node failures, and more.
In turn, Kaminario recently drew the attention of Taneja Group Labs. Scale-out and enterprise-class storage management features are not easy to architect (especially not together), and we wanted to know whether Kaminario could deliver enterprise-class wrappings with all of their historic scale-out capabilities.
Deduplication took the market by storm several years ago, and backup hasn’t been the same since. With the ability to eradicate duplicate data in duplication-prone backups, deduplication made it practical to store large amounts of backup data on disk instead of tape. In short order, a number of vendors marched into the market spotlight offering products with tremendous efficiency claims, great throughput rates, and greater tolerance for the too often erratic throughput of backup jobs that was a thorn in the side for traditional tape. Today, deduplicating backup storage appliances are a common site in data centers of all types and sizes.
But deduplicating data is a tricky science. It is often not as simple as just finding matching runs of similar data. Backup applications and modifications to data can sprinkle data streams with mismatched bits and pieces, making deduplication much more challenging. The problem is worst for Virtual Tape Libraries (VTLs) that emulate traditional tape. Since they emulate tape, backup applications use all of their traditional tape formatting. Such formatting is designed to compensate for tape shortcomings and allow faster and better application access to data on tape, but it creates noise for deduplication.
The best products on the market recognize this challenge and have built “parsers” for every backup application – technology that recognizes the metadata within the backup stream and enables the backup storage appliance to read around it.
In 2012, IBM introduced a parser for IBM’s leading backup application Tivoli Storage Manager (TSM) in their ProtecTIER line of backup storage solutions. TSM has long had a reputation for a noisy tape format. That format enables richer data interaction than many competitors, but it creates enormous challenges for deduplication.
At IBM’s invitation, in November of 2012, Taneja Group Labs put ProtecTIER through the paces to evaluate whether this parser for the ProtecTIER family makes a difference. Our findings: Clearly it does; in our highly structured lab exercise, ProtecTIER looked fully poised to deliver advertised deduplication for TSM environments. In our case, we observed a reasonable 10X to 20X deduplication range for real world Microsoft Exchange data.
Over the past few years, backup has become a busy market. For the first time in many years, a new wave of energy hit this market as small innovators sprang forth to try to tackle pressing challenges around virtual server backup. The market has taken off because of a unique set of challenges and simultaneous opportunities within the virtual infrastructure – with large amounts of highly similar data, interesting APIs for automation, and a uniquely limited set of IO and processing resources, the data behind the virtual server can be captured and protected in unique new ways. As innovators in turn attacked these opportunities, backup has been fundamentally changed. In many cases, backup has been put in the hands of the virtual infrastructure administrator, made lighter weight and vastly more accessible, and has become a powerful tool for data protection and data management.
In reality, the innovations with virtual backup have leveraged the unifying layer of virtualization to tackle several key backup challenges. These challenges have been long-standing in the practice of data protection, and include ever-tightening backup windows, ever more demanding recovery point objectives (RPO or the amount of tolerable data loss when recovering), short recovery time objectives (RTO or how long it takes to complete a recovery), recovery reliability, and complexity. Specialized data protection for the virtual infrastructure has made enormous progress in tackling these challenges, and simplifying the practice of data protection to boot.
But we’ve often wondered what it would take to bring the innovation from virtual infrastructure protection to a full-fledged backup product that could tackle both physical and virtual systems. At the recent request of Dell, Taneja Group Labs had the opportunity to look at just such a product. That product is AppAssure – a set of technology that seems destined to be the future architectural anchor for the many data protection technologies in Dell’s rapidly growing product portfolio. We jumped at the chance to run AppAssure through the paces in a hands-on exercise, as we wanted to see whether AppAssure had an architecture that might be poised to change how datacenter-wide protection is typically done, perhaps by making it more agile and accessible.
The past few years have seen virtualization rapidly move into the mainstream of the data center. Today, virtualization is often the defacto standard in the data center for deployment of any application or service. This includes important operational and business systems that are the lifeblood of the business.
For mission critical systems, customers necessarily demand a broader level of services than is common among the test and development environments where virtualization often gains its foothold in the data center. It goes almost without saying that topmost in customer’s minds are issues of availability.
Availability is a spectrum of technology that offers businesses many different levels of protection – from general recoverability to uninterruptable applications. At the most fundamental level, are mechanisms that protect the data and the server beneath applications. While in the past these mechanisms have often been hardware and secondary storage systems, VMware has steadily advanced the capabilities of their vSphere virtualization offering, and it includes a long list of features – vMotion, Storage vMotion, vSphere Replication, VMware vCenter Site Recovery Manager, vSphere High Availability, and vSphere Fault Tolerance. While clearly VMware is serious about the mission critical enterprise, each of these offerings have retained a VMware-specific orientation toward protecting the “compute instance”.
The challenge is that protecting a compute instance does not go far enough. It is the application that matters, and detecting VM failures may fall short of detecting and mitigating application failures.
With this in mind, Symantec has steadily advanced a range of solutions for enhancing availability protection in the virtual infrastructure. Today this includes ApplicationHA – developed in partnership with VMware – and their gold standard offering of Veritas Cluster Server (VCS) enhanced for the virtual infrastructure. We recently turned an eye toward how these solutions enhance virtual availability in a hands-on lab exercise, conducted remotely from Taneja Group Labs in Phoenix, AZ. Our conclusion: VCS is the only HA/DR solution that can monitor and recover applications on VMware that is fully compatible with typical vSphere management practices such as vMotion, Dynamic Resource Scheduler and Site Recovery Manager, and it can make a serious difference in the availability of important applications.
Selecting a primary storage solution is undoubtedly one of the most critical decisions an IT department can make. As the foundational piece of the modern datacenter, it represents perhaps the single most important piece of IT infrastructure for businesses large, medium or small. Business critical applications will live and breathe on the performance of the selected storage system, and business data will be inevitably constrained by the capacity of that storage system.
In the mid-market, making a storage investment can be particularly daunting, as the stakes are higher, and the selection is harder. Compared to larger enterprises, in the mid-market, storage dollars are fewer and harder to come by. Precious and often limited IT staff time is spread across more systems and technologies, their core skills are often not rooted in storage, and technically vetting a storage system can be all but impossible. This makes storage a risky proposition for the small enterprise (SME) and SMB customer. We frequently hear tales of storage system purchases where I/O is not sufficient, features are missing (or require additional licenses and cost to acquire), or where architectural compromises create availability issues that regularly impact the entire business.
For several years, the developers of the StorTrends line of NAS/SAN solutions have been working hard to architect a storage system for the mid-market that puts an end to these risks and compromises. By harnessing the engineering expertise from their parent American Megatrends, Inc. (AMI) – an innovator in storage and BIOS technologies – StorTrends has been tackling the challenge of delivering abundant performance, robust reliability, and feature rich storage with the SMB and SME customer in mind. Their claim is that the StorTrends 3400i is both one of the most cost effective choices in the market, and one of the most well rounded.
In mid-2012, StorTrends caught our attention with these claims and a series of notable customer wins in a highly competitive market. To learn more, we approached StorTrends with the idea of a hands-on lab exercise, what we call a Technology Validation, to examine in more depth how StorTrends was delivering comprehensive value for customers in the mid-market space. Utilizing our proven validation methodology that included time spent at AMI headquarters in Norcross, GA, we put a set of StorTrends 3400i storage systems through the paces, with an eye toward examining several capabilities that StorTrends claims makes the 3400i one of the best value storage options in the mid-market.
Why does storage performance matter? There are few storage administrators who have not witnessed an unhappy moment where lack of storage performance brought an application to its knees. But quantifying how storage performance matters, aside from serving as insurance to avoid an application crippling moment, is a tricky proposition.
One way is in terms of how storage performance determines the efficiency of the infrastructure. When all else is equal and well tuned – hypervisors, servers, networks – then storage becomes a precious commodity that can determine just how many workloads or applications a given infrastructure can support, and therefore how efficiently it can operate at scale.
When it comes to virtualization, this workload density has serious impacts, but fortunately is even easier to assess. A number of years ago, Taneja Group started out testing workload density among competitive hypervisors, and we labeled it VM density – a measure of how many VMs similar systems can run without compromises in performance or usability. Our net findings clearly indicated how big of an impact a small difference in VM density can have. When poor VM density makes it necessary to add more servers and hypervisors, the cost of servers is costly enough, but licensing tools like vSphere suites can add up to many thousands of dollars and dwarf hardware costs. Meanwhile, with more servers/hypervisors comes more complexity and management overhead, along with operational data center costs (power, cooling, floorspace). The consequences can easily run to tens of thousands of dollars per server. At large scale, superior VM density can easily add up to hundreds of thousands or millions of dollars for a business.
We’ve long suggested that storage performance is one of the biggest contributors to VM density, but few vendors have been brave enough to step forward and declare that their own storage system can take VM density farther than anyone else’s. Well times have changed.
Approximately 6 months ago, HP made a bold promise with their 3PAR storage systems – guaranteeing that customers moving to HP 3PAR storage, from traditional legacy storage, will double the VM density of their existing server infrastructure. Better yet, HP’s promise wasn’t about more spindles, but rather the efficiency and power of their 3PAR StoreServ storage controller architecture – they made this promise even when the customer’s old storage system contained the same disk spindles. Our minds were immediately beset with questions – could a storage system replacement really double the workload density of most virtual infrastructures? We knew from experience that most customers are indeed IO constrained, we’ve had at least several hundred conversations with customers fighting virtual infrastructure performance problems where at the end of the day, the issue is storage performance. But a promise to double the density of these infrastructures is indeed aggressive.
It goes without saying that at Taneja Group, we were more than eager to put this claim to the test. We approached HP in mid-2012 with a proposal to do just that by standing up a 3PAR StoreServ storage array against a similar class, traditional architecture storage array. The test proved more interesting when HP provided us with only the smallest of their 3PAR StoreServ product line at that time (an F200) and a competitive system made up of the biggest controllers available in another popular but traditional architecture mid-range array. Moreover, that traditional system, while first entering the market in 2007/2008, could still be purchased at the time of testing, and is only on the edge of representing the more dated systems envisioned by HP’s Double Density promise.
Our Findings: After having this equipment made available for our exclusive use during several months of 2012, we have a clear and undeniable conclusion: HP 3PAR StoreServ storage is quite definitely capable of doubling VM density versus typical storage systems.