- 论坛徽章:
- 0
|
回复 #15 3150391 的帖子
透明数据移植和负载均衡:
两方面理解:
其一,举例说明,假如现有4台存储节点构成的存储集群,数据量达到90%(因为Isilon的OneFS是分布式文件系统,所以每个节点上的数据量都是90%);现加入一个新节点,原来的数据将会自动透明的重新分不到这五个节点上,以为是后台执行的,所以不会影响系统的性能。
其二,负责均衡
Isilon有个模块,叫SmartConnect,这个模块,可以提供客户端具体访问哪个存储节点,从而实现负载均衡。有四个可选的策略:roundrobin、CPU、connection、aggregation。
和GX的比较:
1) 从架构角度来看:
OneFS:64-bit journaled and fully distributed filesystem (data and metadata)
GX 10.0.X:WAFL non-distributed file system, ONTAP GX 10.0
2)从方案架构角度来看:
Isilon:File system is fully distributed across entire storage cluster. Hardware is fully symmetric cluster (Global coherency & data striped across all storage nodes). OneFS (One File System) combines the traditional storage layers of the file system, volume manager and RAID into a single software layer making for industry leading simplicity and management. Fully Distributed Cluster N-way. Each "node" contains CPU, NVRAM, memory, Gbe ports and disk. Isilon solution is built from day one as a simple all in one modular network storage cluster
NetAppGX:ONTAP GX allows for traditional NetApp controllers/storage silos (FAS 3000's, 6000's) to operate under a single global namespace. "Clustered" together via GigE network must add controllers in multiples of 2 controllers at a time. Supporting from 2-24 controllers. Solution is effectively managing FlexVol volumes and Aggregates. ONTAP GX joins these flexible volumes into a single tree to create the global namespace. File system is still the traditional non-distributed WAFL file system. Solution requires management of each traditional storage silo including each FAS head, HA and FC interconnects, ESH, RAID DP, spares, etc...
3)后端连接方式:
Isilon:千兆网或者10GB Infiniband网;NetAppGX:千兆网
4)全局命名空间:
Isilon:True single global namespace (F: Drive) regardless of size of cluster
NetAppGX:Single namespace (single mount point) achieved by pulling together traditional storage silos and file systems
5)数据透明迁移(autobalance)
Isilon:AutoBalance: Automatic content migration with addition of new capacity. Isilon automatically rebalances (restripes) all data in the background across all the nodes (including new nodes) balancing content across entire cluster
NetAppGX:Manual administrative process to manage and keep data balanced across the "cluster". Requires expertise in FlexVol and Aggregate management, understanding the application workflows and hot spots, and the intricacies of understanding the virtualized cluster that ONTAP GX has created.
6)RAID管理
Isilon:RAID protection change on the fly (during r/w or capacity scaling). RAID mgmt is extremely easy and simple
NetAppGX:Cannot change RAID on the fly during production. RAID protection change requires backup/restore. Traditional RAID configuration challenges
7)性能安全性
Isilon:Because each node in an Isilon cluster is a coherent peer performance is only nominally impacted should a node(s) go out of service or come down. With NFS failover clients are automatically distributed across the remaining nodes minimizing end user impact
NetAppGX:Uncertain perfomance degredation when you lose controllers depending on workflow and configuration.
Global Memory and NVRAM
Isilon:Each node adds 4GB of Globally Coherent cache. Linearly scales. Max cache 384GB (Globally coherent with 96 nodes tested); 49.2GB NVRAM(96 Node Cluster 512MB/Node); all via TrueScale technology.
NetAppGXedicated memory per filer head. Not globally coherent. 6 - 8 GB per single head; Dedicated NVRAM per filer head. 1GB, 512MB single controller
先说这些吧,有什么问题再继续讨论。 |
|