The FlashArray product line powered by Purity 6.1 operating environment with integration of Compuverde software for file services, now relies on replication capability from Komprise Elastic Data Migration (EDM). EDM is known for its migration, copy engine and tiering functions across NAS and S3 storages wherever they reside on-premises or in-the-cloud.
The replication feature works in asynchronous mode coupled with snapshot triggered by Komprise and is fully automatic with some consistent guarantee at destination. But we detect a few limitations:
- it supports only FlashArray at the destination,
- asynchronous means controlled delay but data are exposed to some risk, in other words what happened between to all data created or modified since the last run as snapshot doesn't exist yet and globally everything is one the same chassis?
- following the first master copy, subsequents work at the file level with a directory granularity so for just a few blocks modified, the entire file is again copy to the destination,
and a few questions as details are missing:
- we think N-1 is supported, at least we hope because 1-1 would be a poor design, but we don't know for 1-N or cascading mode,
- the copy must evolve to a block mode to become consistent faster and the possibility to run copy sessions more often,
- and nothing is specified about parallelism between jobs, sessions.... if they use nConnect...
More globally, we don't understand why Pure Storage considers Komprise for FlashArray File Services replication but not for FlashBlade. And it's even a surprise as Komprise works with Pure Storage for NAS migration projects. In other words why the copy engine used by EDM for NAS migration is not used for FlashBlade replication?
As Komprise is also strong in file analytics, Pure Storage should be interested by the analytics part of EDM to learn from its deployed FlashBlade product but also from competition also deployed on same environment. Again not from a Komprise positioning but from a Pure Storage point of view.
Does it mean something for Komprise future?
1 commentaires:
Thanks for covering Komprise’s partnership with Pure Storage and Asynchronous Replication for FlashArray File Services! You posed some great questions in your post – responses are below.
Last week I posted an “under the hood” look at Komprise Asynchronous Replications (https://www.komprise.com/blog/under-the-hood-komprise-asynchronous-data-replication-for-pure-storage/). It provides more details about how it works.
Also, you can join Komprise and Pure Storage in an upcoming webinar where we’ll discuss our partnership and review the Asynchronous Replication solution. The webinar is on Wed 10 Mar 2021 at 9:00 AM Pacific time. You and your viewers can register here: https://www.komprise.com/resource/faster-recovery-file-replication-for-pure-flasharray-with-komprise/
Now, to respond to your questions:
> But we detect a few limitations:
* it supports only FlashArray at the destination,
Currently, Komprise Asynchronous Replication does support only FlashArray Files as source and destination, per Pure Storage's requirements. Komprise plans to support other storage systems for replication, to enable our customers to protect the wide breadth of storage they use.
* asynchronous means controlled delay but data are exposed to some risk, in other words what happened between to all data created or modified since the last run as snapshot doesn't exist yet and globally everything is one the same chassis?
The RPO (recovery point objective) is controlled by the replication schedule, which is set by the user. When using any asynchronous (periodic) replication technique, data changed after a snapshot is replicated won’t be protected until the next snapshot and replication iteration occurs. For example, the documentation for Dell EMC SyncIQ data replication says: “If a disaster occurs on the primary cluster, any modifications to data that were made after the last successful replication job started are not reflected on the secondary cluster. When a client connects to the secondary cluster, their data appears as it was when the last successful replication job was started.”
Also, Komprise recommends that each FlashArray Files be protected by replicating it to a second FlashArray Files chassis, ideally located in a different location to afford safety by geographic distance.
* we think N-1 is supported, at least we hope because 1-1 would be a poor design, but we don't know for 1-N or cascading mode,
N-1 (fan-in), 1-N (fanout), and cascade (A-B-C) are all supported configurations for Komprise Asynchronous Replication.
* the copy must evolve to a block mode to become consistent faster and the possibility to run copy sessions more often,
With its fast parallelism and incremental snapshots, Komprise Asynchronous Replication provides good performance for file replication with the current solution, especially for the workloads being targeted.
* and nothing is specified about parallelism between jobs, sessions.... if they use nConnect...
Users can configure many replications that will run in parallel. Replications are powered by the Komprise Elastic Data Migration engine, which maximizes the use of available resources by exploiting parallelism at multiple levels: shares and volumes, directories, files, and threads, to maximize performance. For more information about Elastic Data Migration, see the Komprise white paper “How to Accelerate NAS and Cloud Data Migrations” at https://www.komprise.com/resource/how-to-accelerate-nas-and-cloud-data-migrations/.
> More globally, we don't understand why Pure Storage considers Komprise for FlashArray File Services replication but not for FlashBlade… In other words why the copy engine used by EDM for NAS migration is not used for FlashBlade replication?
The reason is that FlashBlade has native file and object replication capabilities, while FlashArray File Services does not. Beyond replication, Pure Storage also partners with Komprise to use the high performance, reliable, and cost-efficient Elastic Data Migration for migrating file data onto FlashBlade.
Post a Comment