Itzikr's Blog

XtremIO Content Addressable Storage (CAS)

How does that simplify Storage Management?

You might be asking yourself, what does the XtremIO Content Addressable Storage (CAS) architecture have to do with Storage Management Simplicity?

Well lets start with a quick recap of the XtremIO Content Addressable Storage (CAS) architecture. Following, I will explain why this architecture, besides providing consistent performance, in-line data services etc. also simplifies the daily tasks of the Storage Admin.

Back to the basics: Content Addressable Storage (CAS) architecture: What is it?

The classic metadata structure of traditional storage systems directly links the Logical Address of a Block to the Physical Location of the Block. In this metadata structure, every logical block written, has a physical block linked directly to it. In addition, as most traditional storage systems were architected for a spinning disk storage medium optimized for Sequential writes the address of the logical address affects the physical location that the data is stored. This can lead to an unbalanced storage array that can suffer from hot-spots as specific address space ranges may experience more performance / IOPs than other address space ranges.

The XtremIO took a different approach of managing it Meta Data structures. XtremIO was targeted from day one to support a flash/random access medium. Deduplication and Copy Data Management operations were designed in an always on, inline architecture. The result is a metadata structure that completely decouples the Logical address space address from the Physical one. This is done by leveraging a dual metadata structure:

Stage 1: Logical address to Hash fingerprint

Stage 2: Hash fingerprint to Physical address

If an identical data block is written twice, resulting in deduplication, than only the logical address to the hash needs to be registered. The physical location of the content hash value remains the same. Only the reference count of this hash value is increased by one. This is to indicate that now there are more than one logical address referencing the physical location of the content hash value.

Stage 1: Incoming data is hashed. Logical address is mapped to Content block hash.

Picture #1: All data blocks are hashed/ fingerprinted

Stage 2: Hash address space is distributed evenly to all particpating Storage Controller resources. This results in an auto-balanced system. So if, for example, a specific data base table/ volume is hammered, this will not result in any hotspots, as the data of All Volumes is Evenly Distributed to ALL SSDs.

Note: XtremIO supports a Scale-out architecture. Therefore the total hash range will always be evenly distributed between all of the Storage Controllers of the XtremIO cluster.

Picture #2: Hash range (color-coded below) distributed evenly between all Storage Controllers

Okay, so now the answer to the expected question: How does the unique content addressable storage architecture simplify management?

1. No Pools Management
   

   In XtremIO , Pool Management is not required. All volumes are serviced by All Storage Controller resources leveraging a single Storage Pool.

   So how did we simplify the life of the Storage Admin? No Pool Management required.

2. Auto balanced/ no hot spots
   

   XtremIO was designed to be auto-balanced. Therefore there is no need to optimize the system to support hot spots/ data pounding of specific logical address ranges.

   So how did we simplify the life of the Storage Admin? No need to monitor and troubleshoot hot spots in the storage system.

   

3. All Meta data stored in memory
   

   The metadata structures of the XtremIO architecture are VERY EFFICIENT. Copies are super-efficient metadata constructs. Metadata between the Source and Copy is shared. There is no copying of the Meta data structures. Therefore, there is no need to allocate and increase memory resources when additional Copies are made.

   So how did we simplify the life of the Storage Admin? There is no need to tune the system and allocate additional memory resources when Copies are made. There is no need to increase Memory resources as the number of Copies is increased.

4. Copy Data Management Operations are ALWAYS performed in memory
   

   All Copy operations are always performed in memory. So if a X-COPY operation, a Protection Copy (backup) or a Repurpose Copy (clone) is created – there is no performance impact, as none of these operations result in actual data movement. The above operations result in a super-efficient meta-data operation.

   So how did we simplify the life of the Storage Admin? The Storage Admin can create as many Copies demanded by his organization. The number of Copies is not unlimited, but the XtremIO system supports thousands of mounted and working Copies that can be leveraged for Test, Dev, Reporting and other uses cases.

   

   Bottom line: The XtremIO inherent Content Addressable Storage (CAS) architecture offloads many of the ongoing and sometimes troublesome monitoring and troubleshooting operations required by a Storage Admins of traditional Storage Architectures to optimize an All Flash Array.

   And this is before we talk about the awesome management feature offered by the XtremIO Management System and New WebUI… (see Part #2 of XtremIO Management Simplicity blog posts series).

   https://wordpress.com/post/xtremio.me/4772