UNIX Security Consultant at a retailer with 1,001-5,000 employees
Real User
Top 5
2020-06-18T10:46:49Z
Jun 18, 2020
They are ages behind PureStorage FlashBlade in performance. We recently did an evaluation of scale-out NFS/Object storage solutions, and in order to get to our target performance requirements of about 5000 NFS Ops/sec, the quoted ECS configuration had 13 nodes. It was rejected without even testing. The minimal PureStorage FlashBlade configuration of 7 nodes gave us 80000 Ops/sec, verified with VDBench (that actually is 16 times faster than needed at half the number of nodes). They refuse to go all-flash in order not to overlap with Isilon and performance suffers dramatically from running mechanical disks. Their small object performance is quite disappointing as well. It behaves decently on large objects (think PDF files of 300k+). Even Ceph with RadosGW is 10 times faster on normal dual-socket servers from HP/Dell.
Secondly, after moving to Dell, their website is a mess. It's incredibly difficult to find updates to download.
I have no problems with our application, but we have concerns about the write performance. We would also like it to be easier to scale out, to add more boxes to the system. And we want improved performance, to use a next-generation NFS service.
What is the difference between objects and files? Objects are large collections of data collecting or storing items. Files are a single potential piece within an object.
They are ages behind PureStorage FlashBlade in performance. We recently did an evaluation of scale-out NFS/Object storage solutions, and in order to get to our target performance requirements of about 5000 NFS Ops/sec, the quoted ECS configuration had 13 nodes. It was rejected without even testing. The minimal PureStorage FlashBlade configuration of 7 nodes gave us 80000 Ops/sec, verified with VDBench (that actually is 16 times faster than needed at half the number of nodes). They refuse to go all-flash in order not to overlap with Isilon and performance suffers dramatically from running mechanical disks. Their small object performance is quite disappointing as well. It behaves decently on large objects (think PDF files of 300k+). Even Ceph with RadosGW is 10 times faster on normal dual-socket servers from HP/Dell.
Secondly, after moving to Dell, their website is a mess. It's incredibly difficult to find updates to download.
Technical support needs to be improved.
I have no problems with our application, but we have concerns about the write performance. We would also like it to be easier to scale out, to add more boxes to the system. And we want improved performance, to use a next-generation NFS service.