April 13, 2015
HPE Nimble Storage, founded in 2008, started shipping arrays in 2012. They soon debuted on the Gartner Magic Quadrant as a visionary, which was unprecedented for a new company and speaks to their technology. Within three years, they secured approximately 8,000 customers and deployed over 8,000 arrays. Their rapid growth and Magic Quadrant placement demonstrates how disruptive Nimble’s technology – which was recently acquired by HPE – is to the storage industry.
Creators of the HPE Nimble Storage file system looked at the history of storage technology including: Direct Attached Storage (DAS) or server centric storage, monolithic storage that was used in the 1990’s, and modular storage from the 2000’s. They then took a step back and created a file system design that took advantage of the benefits of past storage technologies, and redesigned it to overcome previous shortcomings. As a result, HPE Nimble invented the Cache Accelerated Sequential Layout (CASL) file system, which is a Flash Optimized Hybrid Storage System that takes advantage of the best that spinning disk and flash have to offer.
7 Reasons Nimble CASL is a Better Storage Solution Than Traditional File Systems
1. Inline Compression and Variable Block Sizes
- Traditional File Systems use fixed block sizes that don’t have the same inline compression.
- The Nimble Solution uses inline compression with variable box sizes, resulting in in 30% to 70% less flash and capacity resources and a much less costly, denser solution.
- How they do it?
Applications send data in fixed block sizes, such as 16KB, 32KB, or 64KB chunks. The Nimble CASL does not need to chop up data blocks like the other storage players do. In SQL, for instance, you can configure the application to send data in 8KB or 16KB block sizes. NetApp uses a backend block size of 4KB, and EMC uses a backend size of 8KB. When the 16KB SQL block comes to the front end processor of the storage controller, it has to catch it and chop it into smaller fixed-sized blocks on the disk in order to make a 16KB block fit in a 4KB or 8KB block. Whereas Nimble can natively catch 16KB blocks without chopping them into 4KB or 8KB block sizes, and can subsequently compress inline instantly.
2. Acceleration of Random Writes
- Traditional File Systems bolt on a flash tier to their file systems to achieve fast random writes.
- The Nimble Solution uses inline serialization of all incoming write I/O’s. This means Nimble can use traditional 7,200 RPM low cost, spinning disk hard disk drives to achieve SSD-like performance, allowing Nimble Storage to provide more cost-effective solutions.
- How they do it? Sequential layout is the process by which the frontend controller takes a bunch of random blocks coming in to the storage processor and holds them in memory. The processor then waits for more random writes and subsequently caches those as well. The storage processor then builds a long sequential stripe so that when it gets laid down to the disk, the mechanical arm within the traditional disk doesn’t have to move across the disc platter to find the different blocks of data, which would ultimately create latency, increase rotational seek times, and decrease application performance. Reducing arm movement significantly reduces rotational seek time.
3. Acceleration of Reads
- Traditional File Systems accelerate reads by frequently migrating data between flash tiers, using excess capacity, which causes unnecessary silicon wear in the flash and chews up resources.
- The Nimble Solution uses a dynamic caching strategy within the CASL file system so that although all of the data is written to the disk, the hot data is dynamically stored in flash. SQL is an example of hot data because of the frequent reads, writes, and database changes. An example of cold data is a three-month-old Exchange email, which would likely be safe to store on a traditional spinning disk.
- How they do it? Nimble’s most important differentiator with the CASL files system is how they populate the flash. Traditional storage vendors wait for a cache miss to populate the cache. For example, if a user requests something that wasn’t already in cache and then wants to pull it into cache, traditional vendors wait for the first miss, which unfortunately results in the first users not reaping the benefit of the flash. Nimble uses 22 patented algorithms that watch the data as it comes between the storage processor and the bus down to the disc and inspects it to determine how random it is, and at that moment one copy gets laid down on disc and one on flash. This allows the first person that requests the data to immediately realize the benefit of flash. Essentially, Nimble CASL evaluates how random the data is and makes a decision whether to put it in flash.
- Traditional File Systems use a copy-based snapshot methodology. The issue with copy-based snapshots is if you are doing a snapshot of a 2TB volume, you’re literally using another 2TB of space within the volume. This is horribly inefficient, chews up a ton of resources, and makes a replication/DR strategy difficult because you’re often sending large files over a WAN or Internet link every time you do a snapshot. Some companies have completely abandoned copy-based snapshots because of space/resource inefficiencies. This leaves them either without current back-up, or they have an old one they can recover, but this would be painful for the business during a real outage.
- The Nimble Solution uses a pointer-based snapshot technology that requires zero space. With CASL, some customers are doing snapshots of their SQL database every 5 minutes. Nimble offers a lot of power in terms of the granularity of the snapshots and the frequency with which a company can do snapshots, which provides incredible RTO’s.
- How they do it? A snapshot is a clone, so you’re making an exact copy; if the original is 100GB, the copy is 100GB. A good way of explaining this is using metadata as an analogy. Every bit of data that is written to the disc has properties. If you right click on a Word doc, you can view these properties (time created, size, etc.); consider this your metadata. By the same concept, a snapshot is metadata about the data that is already written on the disk. When you take a snapshot that is pointer-based, as opposed to clone based, it’s going to point back to the files for everything that was created at that moment. All the snapshot does is maintain the differences between the original snapshot and the pointer-based snapshot. Every snapshot represents the differences between the last version, so it only includes the changed data from the last time you took a snapshot. When you’re talking about replicating snapshots to another site for DR, you’re really talking about taking the original data and all the changes behind it, so you have the full and all the incrementals. If you need to get back to a point in time, you can pull the full and/or all the incrementals, up to the point needed, to recover the file. Every single snapshot when you go to restore looks like a full.
- Traditional File Systems have performance and capacity coupled, so if you need to add capacity, you may also have to address licensing and the controller heads. Plus, scaling can be disruptive and has to be done during an outage window.
- The Nimble Solution has the ability to scale-up and scale-out very efficiently, easily, and non-disruptively. Decoupling performance and capacity allows you to be very specific on how you scale. If you want to increase performance, you can just add CPUs or Flash, or if you want to increase capacity, just add arrays or disk. Another advantage is that scaling can be done during business hours.
- How they do it? Nimble accomplishes this by designing their system based on the requirements of the customer on one controller head, as opposed to two. This way, if you have to take down one of the controller heads for scaling, the environment is completely operational because it was speced to function, from a performance and capacity perspective, using one controller.
6. InfoSight Cloud-Based Management
Nimble is the only storage player that has a cloud-based controller platform. Nimble’s Infosight Engine collects millions and millions of sensor values from deployed Nimble arrays around the world, and does predictive analytics to make decisions for your SAN environment. Infosight gives you insight like, “in about 3 months you will probably have to add capacity, “ or “in 4 months you have to add additional CPU’s because the performance of your applications may start to suffer,” or “a service patch has come out, and we will automatically deploy it for you.” With analytic and data updates, Nimble is able to resolve over 90% of your technical issues through the InfoSight platform, without requiring you to pick-up the phone. Plus, if you need to call Nimble Support, the majority of the cases are resolved in 20 minutes or less.
7. Small Size
Nimble solutions are often 1/5th the size of typical SAN vendors, which equates to significantly less rack space and cooling, a cost savings over the long term. Also, denser and cleaner solutions are typically much easier to understand and manage.
How Mindsight Can Help
As we’ve demonstrated, the Nimble Storage CASL file system has many significant advantages over most competing storage players. If you’re running out of storage capacity or your critical line of business applications aren’t performing, contact a Mindsight Data Storage Solutions Architect for a consultation.
Like what you read?
Mindsight, a Chicago IT consultancy and services provider, is an extension of your team. Our culture is built on transparency and trust, and our team is made up of extraordinary people – the kinds of people you would hire. We’ve always prided ourselves in delivering the full spectrum of IT services and solutions, from design and implementation to support and management. Our highly-certified engineers and process-oriented excellence have certainly been key to our success. But what really sets us apart is our straightforward and honest approach to every conversation, whether it is for a local business or global enterprise. Our customers rely on our thought leadership, responsiveness, and dedication to solving their toughest technology challenges.