Sunday, August 30, 2009

EMC Centera and Production Archive

In a recent briefing EMC articulated the performance advances they have made with Centera and the interesting use case evolution they have witnessed. Apparently data access patterns for Centera have moved from one predominantly weighted towards the write once read occasionally of large files to one that is increasingly write, with infrequent changes and read often, increase in data requests, decrease in file size and total file population in the datastor measured in the billions. This is the data access pattern they define as dominating a production active and is characterized as mission critical real-time access to long term storage.

As positioned by EMC, production archive is a hybrid storage tier that needs to have the attributes of Tier 1 storage such as low latency, high IOP’s, optimized for small grained random access, cacheable and with extensive data protection schema’s. It also needs the attributes of a persistent data store where data may not be accessed for months, but must be available quickly when needed; robust data management for long term retention; data optimization; bandwidth centric; data retention features and immutability. All these are characteristic that are visible in Centera.

So how has Centera’s performance improved over the past 5 years?

Attribute………...……2004………………….2009:
Object/node……………..…5M…………………...…100M
R/WR Ration……………...20/80……………….….45/65
Read Performance……....50obj/sec……………..900obj/sec
Write performance…...…50obj/sec………….….650obj/sec
Access Method………..….Centera API………….XAM API
Self Healing………………..Weeks………………..…Hours

With a 5000+ install base and all the characteristic needed to support an active or production archive visible in Centera, this platform looks well positioned to continue its dominant position as a performance archive solution. However competition in this space is growing from such vendors as HDS, Content Archive Platform; IBM, DR550 and HP, ExDS 9100. Not to ignore the smaller players such as Permabit, Data center Series Model 4010; Nexsan, Assureon; Caringo, a hardware agnostic CAS software solution, CAStor and ProStor with their removable disk option, Infinivault.

No comments: