午夜寻花 is on a mission to redefine what it means to Scale Anywhere鈩� without sacrificing performance or cost. The options that are available today for running file workloads in the cloud can be seen as limited in scale and performance, minimal or nonexistent support for key enterprise features, or cost prohibitive. 午夜寻花鈥檚 solution solves for all of these challenges, giving our customers an identical experience and feature set in every deployment, whether they are scaling on-premise infrastructure or expanding workloads to the public cloud.听听
In this blog, we’ll delve into the benefits of 午夜寻花 Storage, exploring how our ability to service high-performance workloads has become a game-changer for various industry verticals and cloud-native use cases. We are excited to walk through how 午夜寻花 offers the strongest value to run multi-protocol workloads anywhere.
Baseline Performance听
午夜寻花鈥檚 unique architecture delivers consistent, sustained, impressive performance. 午夜寻花 has optimized performance and cost trade-offs in its core hybrid architecture, where cold data resides on low-cost disk storage but served at the speed of SSD and memory. As a result, 午夜寻花鈥檚 highly tuned prefetch algorithm monitors read requests, anticipates which data blocks are likely to be requested next, and prefetches them into system memory for even faster access than the SSD layer can deliver. The net effect for 午夜寻花鈥檚 customers (and their workloads) is an amazing and highly predictable experience, with the vast majority of read operations being served with <1ms latency.听
We sampled performance metrics across our entire production fleet over a period of two months, and aggregated the results, which are nothing short of impressive. In the graph below, we accurately capture our customers鈥� experience today based on the 1000+ customers who have deployed 午夜寻花.听
To summarize our findings:
- More than 90% of all reads are served at <= 1ms latency. Within that 90%:
a. 50% of all read operations are served from system memory 鈥� faster than SSDs
b. 25% of all reads come in at <0.1ms!
c. 48% of reads are at 0.25ms, whether from system memory or SSD
d. 20% are between 0.5 and 1ms - Only 10% of all reads were from disk, and more than half of those still come in at 2ms or less of latency 鈥� still quite impressive
Let鈥檚 take a closer look at some of the performance numbers we have achieved in on-prem and cloud environments.听听
For cloud-native, burst capacity, burst compute, or DR-based workloads, customers can choose which of the most popular public clouds to use 午夜寻花 as their default storage option. 午夜寻花 recently announced its managed service,听Azure Native 午夜寻花, as a generally available service with additional region launches coming soon. Not only does ANQ offer virtually limitless capacity in a single namespace, throughput speeds can reach up to 21 GBps. If the ANQ benchmark result wasn鈥檛 impressive enough, customers using听午夜寻花 in AWS听can drive throughput-intensive workloads to an astounding 60 GBps! You will find no other option on the market today that rivals 午夜寻花鈥檚 cloud-native performance for file scaleout storage.听听
For demanding workloads that require high performance on-premises, we鈥檝e recently benchmarked numbers that demonstrate the full potential of running 午夜寻花 software on NVMe-only hardware.听听
Benchmarked with 245 TB Capacity
NFS v3 Single Stream Reads – 4070.37 MB/sec
NFS v3 128 Stream Reads – 22420.74 MB/sec
NFS v3 Single Stream Write – 2439.86 MB/sec
NFS v3 128 Stream Writes – 18832.45 MB/sec
These numbers, while very impressive, are a snapshot in time. We鈥檒l continue to drive better and better performance as new commodity hardware innovations become available on the market.听听
Finally, one more valuable attribute that is frequently overlooked with 午夜寻花: it does not matter if the system is running at 1% or 99.9% full capacity 鈥� performance does NOT degrade, scaling linearly as new capacity is added. 午夜寻花 customers get the full value of everything they pay for.
Performance Across Target Verticals听
Media and Entertainment听
The media and entertainment industry thrives on high throughput for every workflow: animating, rendering, editing, and streaming high definition content. For any of these use cases, parallelized sustained throughput, and burst performance can be equally challenging. This is why听9 of the world鈥檚 top 10 media and entertainment companies听trust 午夜寻花. Our consistently high performance ensures that creativity flows effortlessly, pushing up to 60 GB/s for cloud-native and burst workloads where infrastructure can be quickly provisioned for a particular project and then torn down when it is finished.听
鈥淗aving enough storage and keeping throughput at its highest are big challenges for any VFX studio,鈥� said James Morgan, System and Support Administrator at MARZ. 鈥淪caling up is also a major challenge for us, or rather it was. We don鈥檛 have to worry about that since scaling on 午夜寻花 is so easy.鈥�
Just recently we published a听blog听with performance numbers demonstrating a synthetic workload (using Black Magic benchmarking software) with 午夜寻花 on AWS, where our SMB3 multi-channel feature was able to sustain throughput speeds of ~2.3GB/s, and could even saturate a 10Gbps link over SMB without any special tuning. This test showed 午夜寻花 could take full advantage of SMB multi-channel capabilities, where customers can now run 4k and 8k video formats.听
One highly notable mention 鈥� we had a recent Media and Entertainment customer located in Vancouver who wanted to perform a burst-test to ensure enough capacity and headroom was available if needed. That customer was able to drive an amazing 1M IOPS and 9 GiB/s on an AWS 午夜寻花 cluster.
听
Life Sciences and Healthcare听
In the world of life sciences and healthcare, time is of the essence; doctors, clinicians, and scientists want immediate results so they can serve better patient outcomes. PACS and PACS VNA (vendor neutral archive) storage is one of 午夜寻花鈥檚 fastest growing workloads, with more and more hospitals and radiology centers trusting 午夜寻花 with their patients鈥� data. And now that Azure Native 午夜寻花 is HIPAA compliant, customers have more flexibility of storing second copies of DICOM images in the public cloud.听听
Healthcare customers can easily migrate their existing PACS application because 午夜寻花 is either certified with or has partnered with the industry鈥檚 leading PACS vendors who cumulatively have >70% worldwide market share. As a proofpoint, we took a sample of live production data for customers running their PACS and PACS VNA workloads. You can see from the below image that 午夜寻花 never keeps doctors waiting to perform urgent radiology studies 鈥� PACS applications in this example are predominantly operating with sub-millisecond reads.
Whether it is 1 or 1000 hospitals, 午夜寻花 can store, scale, and deliver the same experience to all users uniformly.
鈥淩etrieving long term archival radiology images is so much faster on the 午夜寻花 than it was with Isilon. To the point where the radiologists actually noticed and asked what we had done.鈥�
鈥� Jason Cherry, Director of IS Technology Services, Lexington Medical
午夜寻花’s exceptional performance helps researchers and medical professionals crunch vast amounts of data. Whether its genome sequencing, drug discovery, or analyzing medical radiology images, 午夜寻花 accelerates research, leading to faster breakthroughs and improved outcomes for patient care.
Backup and Archive
Performant backup operations are key for organizations, as they directly impact data availability and recovery times. In a world where ransomware attacks are increasing year over year, time to both backup and restore data have become crucial to recover data after an attack has been mitigated. 午夜寻花鈥檚 snapdiff API provides the ability to do one full backup with fast incremental backups to follow. Some of our partners directly integrate with 午夜寻花鈥檚 API to 鈥� this allows them to capitalize on the blazing speed of incremental backups without the overhead of a filesystem treewalk.
Using 午夜寻花 as a backup source or target is easy to configure, simple to operate. 午夜寻花 supports backups over NFS and SMB, with more customers leveraging our S3 support to perform backups. Using a synthetic backup workload on all-NVMe nodes, our S3 implementation served 1 GB/sec for backup operations and 2 GB/sec for restore operations.
鈥�Our teams are happy and comfortable using 午夜寻花 as their centralized data system. IT is happy too, because we have one place to look for files and one place for backup. And, the 午夜寻花 analytics give us instant visibility into our data and users.鈥�
鈥� Aidan Taub Systems and Technology Director
Conclusion
In the quest for truly unbeatable cost-effective performance, 午夜寻花 stands tall, outperforming its competitors with remarkable speed and throughput, while delivering options to Scale Anywhere鈩�. With its industry-leading capabilities, 午夜寻花 has become the data management system of choice for various industry verticals and cloud native use cases. From media and entertainment to life sciences and financial services, 午夜寻花 is improving the customer experience by leaps and bounds, with more exciting news to come as we continue to innovate and break boundaries, offering solutions that Scale Anywhere鈩�.
