Comments (0)

By admin, August 30, 2011 1:50 pm

I came across an interesting thread about 3PAR’s unique SAN capability especially about its Chunklets, again it’s in Chinese. Those who participated in that thread seemed to be the SAN elite in their fields.

To my surprise, I’ve learnt that 3PAR is still using the old PCI-X, NO SAS drives. Well heard they are going to release the SAS version shortly in Q4 2011 and there is a new series V after T & F.

Interesting enough another SMB SAN/NAS product named Drobo somehow can achieve the same unique feature as 3PAR as Drobo’s Beyond RAID technology can also form a RAID using different size, different RPM drives (ie, mix SSD/FC/SATA), What’s more Drobo doesn’t use any hot-spare for redundancy as all the drives can devote part of the capacity for the hot-spare purpose, this is another similarity to 3PAR. Well, the price difference between the two is of course beyond numbers.

 1,中端存储高端化：3PAR定为于高端存储，但是它的架构有点类似中端存储的架构，有控制器的概念，最多可以有8个控制器（S400最多有4个控制器，S800可以有8个），控制器之间由一个网状背板连接起来，每个控制器有两个INTEL的CPU负责控制指令，还有一个3PAR自己控制器的ASIC完成数据移动。普通的中端存储，一个LUN只能属于一个控制器，用户必须将LUN人工分布到两个控制器上，而3PAR每个LUN实际上是由多个控制器并行处理的。

这个怎么说那？
个人感觉本身可能打算用来代替EVA的，但是目前在hp的产品线看来他是介于xp和eva之间，当然他的满配跟hp的xp也差不多少了，更何况他更适合于云，另外当年oracle又说他跟asm结合能够如何如何的提高性能。

毕竟他是hp自己的东东，相对于可能更愿意推3par，其实3par2个系列，F和T，而跟在后面的数字就是支持的控制器的最大数，F200最大支持2个，F400最大支持4个，T400最大支持4个，当然T800最大支持8个，

至于他的控制器的芯片感觉有点类似于LSI，intel的cpu只是负责控制信息，真正做raid计算的是ASIC，所以如果和其他厂家的控制器全intel inside来pk说3par的cpu不行，可就有点外行了

当然Q4听说要出新3par就将采用双ASIC控制器，并支持SAS硬盘，将pci-x改为pci-e
另外说双A（不是ALUA哦），这点不能说其他厂家的终端存储都最多做到ALUA，其实hds的AMS也是真正的双A，至于eva和cx都是ALUA，ds5000就只能是A/P了

2,非常有特点的盘盒： 3PRA插盘的方式非常独特，和其他存储都完全不同，在4U高度的盘盒中，有10个盘包，每个盘包可以纵向插入4块盘，相当于其他存储一块盘的位置插入了四块，所以3PAR的盘的密度非常高。当然这也有问题，比如在换盘时，每次需要将整个盘包拔出，或者盘包自身的背板损坏，可能造成数据丢失等等。不过3PAR已经考虑了这些情况，并有解决方案。

其实3par有3种盘笼，DC2, DC4（T-Class ） and DC3（F-Class）DC2, DC4累就是上面写的，但是dc3是16个个drive bay，0-15就不是4个硬盘装在一起了。

3，虚拟卷管理：第一层：每个盘都被划分为256M的小块(chunklet)，由于每个盘盒都和两个控制器相连，所以这些存储块都有两个访问通道。第二层：将这些存储块，基于RAID类型和存储块的位置组成逻辑盘(LD)。第三层：将一个或多个LD映射成为了一个虚拟卷(VV)，并最终以LUN的形式输出给主机(VLUN)。这样可以将IO分散到系统所有的磁盘，光纤连接和控制器上，而不象某些存储，必须要借助主机的LVM才能实现。如果我们使用文件系统或者ASM的话，将会很方便。另外一个特点就是管理非常方便，只需要告诉系统我要多大的LUN，其他系统会自动完成

这点感觉有点类似于eva（其实eva）或者v7000，xiv，当然他们又不完全相同，XIV的Pool是个逻辑的，他并没有将disk物理隔离（譬如EVA中的Disk Group，eva有rss的概念），对于外部的IO quest由所有的Disk提供，只是在空间，也就是Size方面逻辑的划分pool

对于国内存储市场来说，3PAR 是不折不扣的后来者。也是个相对陌生的存储产品，以至于其竞争对手的人员甚至都不知道这家公司已经杀入中国市场。

3PAR 在 1999 年成立，几个创始人主要出自 Sun ，前身叫作 3PARdata ， 2008 年上市。要知道在存储技术领域竞争还是比较激烈的，EMC / HDS 等控制着高端存储的主要市场，3PAR 能突破技术壁垒并最后成功上市，没两把刷子那是绝对做不到的。

InSpire 硬件结构

3PAR 背板采用全网状的连接结构，每个控制器节点之间高速直连。因为是全网状的，所以基本上一个链路坏掉只影响直连的两个节点的通信，对其它节点无影响。每个控制器节点内置一块硬盘，用于操作系统安装。控制器节点最多可以扩展到 8 个，是 3PAR 存储最核心的组件。

相比之下，HDS 架构采用全光线交换方式（Universal Star Network），而 EMC 是采用直连矩阵方式(新一代产品采用虚拟矩阵架构–Virtual Matrix ，其实已经放弃了直连矩阵架构了)。这些连接方式的孰优孰劣历来是厂商攻击竞争对手的着眼点，能否最大限度发挥性能是用户最需要关心的。

3PAR 针对 I/O 指令和数据移动使用不同的计算芯片。I/O 指令(元数据/控制Cache)用 Intel 的芯片，而 数据移动/Cache 则使用专门设计的 ASIC 芯片来完成。

因为有专门的硬件 ASIC 芯片用于 RAID 5 XOR 校验，3PAR 号称有了其第三代 ASIC 芯片，实现的 RAID 5 是业界最快的，甚至 SATA 盘也能有不错的性能表现。(从 Oracle 公司测试的数据来看，和 RAID 10 速度的确相差无几。)

InForm 操作系统软件与虚拟化

3PAR 的操作系统叫 InForm，最初就是面向层次化的设计。与其他存储不同的是，3PAR 所有磁盘被分成 256MB 统一大小的小盘(Chunklet)，可以根据需要用多个 Chunklet 组成 RAIDlet(逻辑磁盘)。因为这个独特的设计方式，3PAR 是可以很容易做到不同容量的磁盘混用，同一个 RAID 组里都可以有不同大小、不同转速的磁盘混用，这是其他存储做不到的。而且，所有的磁盘都可以利用，因为Hotspare Chunklet 以更小的单位分散在不同的磁盘上，也不再需要单独留热备盘。空间利用率可以更充分一些。　

多说一句，有这个冗余机制，3PAR 更换磁盘也是与众不同：直接抽磁盘盒子(一个盒子可是四块磁盘啊)，我当初看到 3PAR 技术人员这么操作真是着实吓了一跳。

因为固定大小的 Chunklet 的存在，可以将 I/O 更为均匀的分散到多个磁盘上。

对于熟悉Oracle 的朋友来说，会发现这和 ASM 的思想非常接近。因而也可以和 Oracle 数据库进行无缝集成：

因为软件做得非常具有易用性，日常管理与维护远远没有其他高端存储那么复杂，新增磁盘这种事情，都是一行命令之后底层自动处理。其实在 Thin Provisioning 方面 3PAR 也是很值得一说的，比一些厂商的伪 Thin Provisioning 具体多了。限于篇幅，不赘述。

3PAR 在美国有很多金融证券行业的客户，也有 Web 2.0 行业的客户–MySpace 。在保证 I/O 响应在 10ms 以内的前提下，3PAR 的 IOPS 能力非常优异(这才是卖点，不难理解其客户多集中在证券、金融领域)。虽然有些厂商号称能得到更高的 IOPS ，但那是在 I/O 响应时间很差的情况下的数据。要说明的是，现在随着一些存储厂商在高端服务器上也支持 SSD ，未来几年如何还要再看。

前两年 3PAR 推行所谓 Utility Storage(功用存储) 理念，现在貌似改成敏捷存储了。说实话，我觉得敏捷存储真的挺适合的，3PAR 命令行批量创建 LUN 真的很让人感觉舒服。当然，也在宣传云存储和绿色存储的理念，那是题外话了。

3PAR 原来只做中高端市场，只有 T 这一个系列，现在也开始关注中低端市场了，推出了 F 系列的产品。软硬件体系基本没变，倒是没仔细看过。

当然3PAR也有几个不怎么样的地方:

1.不支持sas盘，大家都用sas，你不用，是不是有点另类？
2.换盘麻烦，不像xp，eva那样直接就拔盘（当然只是T系列），虽然这点对于用户和销售来说不是问题，但是对于工程师来说还是费劲。估计要多干10分钟的活
3.pcx-133的总线
4.虽然说背板不是个什么问题，但不管怎么说还是个单点
5.后端依然是JBOD
5.硬件感觉不如软件那么NB
6.lun的大小调整需要额外的license，不像eva，这都是基本的免费功能
7.3PAR 的技术优势在机械硬盘时期，但未来将是SSD 的时代，追求高 IOPS ，更小 I/O 响应时间，如果全用ssd即可
8.不支持raid3

1.3par相比其高端存储配置时间和复杂度大大降低，抛开xp配置的复杂度不说，只谈时间，每次做dci，格式化后，我们就可以去睡觉了。

2.相比其他中端双控存储，
他们很难做性能方面的扩展，也难以提高的更安全的保障，某些硬件故障的时候，它这个性能可能会大幅的下降，即使是某些高端的存储，如果某一个cache 板挂掉，性能可不是降低一点点哦？将近达到90%以上了。

3.自动分层
他把不同类型的硬盘，如SSD、FC或者SATA，划分为256MB Chunklets，这些Chunklets被自动的分组，构成vv，以前是针对整个卷进行RAID划分，而现在，对子卷进行RAID划分。实现了分层的存储，从而让一个卷可以跨SSD、FC和STAT等不同的类型的硬盘。

通过对vv中不同类型硬盘IOPS的监控，就可以实现基于策略的数据自动迁移，自动在不同硬盘底层平衡工作负载，如用SSD来满足应用对IOPS要求。系统帮他决定哪个应用需要IOPS比较高，自动把这部分数据迁到SSD的硬盘上，

不像其他存储，redolog你要不放在SSD上，要不放在FC盘或者SATA上，如果我很多的文件都要求性能很高就费劲了，都要放在ssd上，但是3par就不用了，你可以把他们都放在池子里面，性能不够加ssd，容量不够加sata，其他的，管他那，唉，做3par的cpu挺累的，真不容易，我感觉未来应该就是这样，对用户越来越透明，不过以后如果全部ssd，这个功能就可以88啦,关于自动分层吹牛的东西咱不聊，网上讲的很多，也很细，不过我看很多其他厂家也有这技术了

Comments (0)

By admin, August 27, 2011 9:41 pm

Today is another one of my lucky day for vSphere treasure hunt. I was able to locate three really useful ones again.

VMWare Compliance Checker for vSphere
Not much to say here, simply get it and run through your infrastructure, the assessment report is mainly related to the default settings causing security holes on ESX hosts and recommendation how to fix them.

VMTurbo

If you have already got Veeam’s Monitor (Free Version), then this is the one that you shouldn’t omit. It’s a compliment to Veeam monitoring tool where you can use VMTurbo Community Version (Monitor+Reporter) to see those details are only available for the Veeam paid version. Furthermore, it clearly displays a lot of very useful information such as IOPS breakdown of individual VMs.

Except the recommendation part is a bit too aggressive. For example, if a VM has been idled for 1 day, then VMTurbo would recommend to reduce the Ram size, but in reality the size is required during the user session. Another example is VMTurbo always suggest to reduce the size of those already Thin-Provisioned disk as it think they are taking too much wasted storage space.

The Report Feature is probably the Best I’ve ever seen and the most easiest one to setup, there is no more complicated steps like in Veeam Reporter (setup MSSQL Reporting Service along is a pain) as VMTurbo itself is a virtual appliance.

Overall speaking, it’s a very good product, way better than Xangati IMOP and has a great potential to match Veeam’s product. I still haven’t got time to play the Capacity Planning feature, last time I ran VMKernal’s Capacity View (Free Version), the recommendation was simply a joke as it said there are only 2 VM we can add while in reality there is a couple of hundreds of Ram that’s free to use, hopefully VMTurbo can give me some insight for real capacity growth prediction.

AVG Anti-Virus ISO
Simply use it to boot the VM, scan and clean the whole volume or designated directories. It also allows you to update the virus definition file in real-time. Just to make sure the disk type is not paravirtual as it won’t recognize this new disk type.

Update Aug-31

Well, after extensively testing VMTurbo for 3 days, I must admit it’s probably the most comprehensive monitoring and reporting product I’ve ever seen. I love the features that I couldn’t find in Veeam’s free edition and VMTurbo probably got every scenario combination you can think of, like those vCPU/MEM/IOPS/Network, etc.

Of course there is some drawback, one of the biggest drawback is to manage the product itself is too complicated, somehow, I don’t get the same easy to use feeling as Veeam. I understand both have tree on left pane and details on the right, but somehow it just takes me a few more clicks to find out what I need on VMTurbo, it would be perfect if they can improve this aspect. It actually did happen to me that I got lost after 10 clicks for in-depth information of what I was looking for.

Another thing is I still haven’t got the Planning working out correctly, I tested on a 2 hosts cluster with even load (probably 53%:47%), the result strongly recommend to shift all the load from host 1 to host 2, so I got a 2%:98%, isn’t this suppose to be evenly balanced? (ie, close to 50%:50%)

Finally, I understood the concept and calculation how to derive the available HA Slot today. ESX 4.1 did a great job in simplifying the whole calculating thing and present the available slot number automatically. This HA Slot Availability number is crucial in estimating the future capacity growth. The other thing is I’ve learnt  instead using Reservation, Use Limit instead for your SLA as this will not change your default HA Slot. So simply create a Resource (ie, Jail) Pool and assign CPU & MEM Limit (for Network, use VLAN with capped speed, for IOPS use SIOC) to it and place those abusive VMs into the “Jail”, Btw…there is no chance they can ever “Jail Break” under vSphere.

Comments (0)

By admin, August 23, 2011 11:02 pm

相傳位於中國大陸廣東省有一間山寨廠專門製作一些市面上沒有的豪華汽車手版﹐集中以Rolls Royce﹑Bentley和Daimler這三間英國著名的廠家為主。

之前有車模朋友告之深圳華強商業城也不少散貨﹐而且價格廉宜﹐大概人民幣200至300左右。後來也陸陸續續地出現在國際拍賣網站(居然沒給禁)﹐香港一些玩具和車模店見過傳說中它們的身影﹐但總絕對欠缺了些什麼﹐這當然不是指正宗版權的問題﹐而是指比例正確﹑車型準確的感覺﹐雖然說是樹脂膠製作的手版﹐但做工尤其是蝕刻片方面來的比較粗糙﹐如Rolls Royce的女神比例過大﹐內飾極之馬虎等等﹐所以至今一台也沒有進。

山寨廠手版其款式之多﹐猜想老闆肯定是個識貨之人﹐行家是也﹗八九成是個Rolls Royce﹑Bentley和Daimler的超級發燒友﹐要不然如何懂得推出那些最稀有﹑最值錢的款式﹗

因為自己一直都對英國皇室的專用車極為感興趣。尤其是今年六月份英皇室大婚一系列的專用車看的人們是目不暇精﹕從開始接載皇妃出場的1902年State Landau Wedding Carriage﹐到襯家坐的1977年Rolls-Royce Phantom VI﹐接著是女王乘座的2002年Bentley Golden Jubilee(今天的主角)﹐直至最後兩位新人同時出現的1965年Aston Martin DB6 Volante。每一台都充滿了豐富的歷史故事﹐尤其是最後那台Aston Martin DB6 Volante曾經是威廉王子父母黛妃和查理斯曾經在30年前一齊乘坐過的﹐真令人感嘆世事變幻無測﹐珍惜眼前人。

回到正題﹐自己一直都對兩年前沒能收到那台1比18 Minichamps出的女皇車耿耿于懷﹐因為當年的內飾做工實在對不起近千元的售價﹐但現今的價格居然仍能升高了近倍﹐太離譜了﹗最近曾經有機會用港幣750的低價收進此車﹐但看到實物後慘不忍睹﹐全車已經長滿了大大小小的氣泡﹐賣家又不肯減價﹐所以最終只能放棄﹐怪不得贏得此車的第一手買家也放棄了。

後來想收1比43的女皇車﹐但最近價格也彪升近5百﹐性價比是同樣的低﹐而且竟然跟18的一樣沒有車頂的女王徽章盾牌﹐看樣子Minichamps同樣沒拿到皇室的准許﹐沒有女王盾牌就如同沒了靈魂。這樣一來﹐國產山寨擁有女王徽章盾牌的版本頓時引起了我極大的興趣。加上價格比當初回落了不少(有近半的折扣)﹐當然不能錯過如此“珍貴”的機會(後有解說)。最近碰巧又看到MC論壇的quattro_360關於英女皇專用車手版的貼子更加心動了﹐馬上去淘寶買了回來﹐這次聽店舖老闆建議不用順風了﹐因為順風總會打開包裹檢查而不時弄壞車模。

跟店主討價還價用極合理的價格拿到手之後﹐它的精緻真的有點震撼了我的審美觀﹐把我之前對國產山寨的片面看法一掃而空﹗尤其是那些電鍍配件的精緻度﹑挑空的散熱孔﹐三盞尾燈的再現﹑車尾的天線﹑令人讚嘆的賓利標誌(雖然真車並不是用B標誌﹐看這裡)還有就是此車外型的準確程度﹗不能不提的是山寨老闆還挺幽默的﹐跟大家開了個隱藏的玩笑﹐就是後座小巧的LCD上竟然顯示了一台鮮黃色調的美國平民超跑Corvette C5﹐這點我倒是挺欣賞的﹐有才有才﹗﹗﹗

在收集關於此手版的過程中﹐發現原來它的變種極多﹐至少不下6款﹐有這款單車紅地毯的﹑單車紅地毯的有限量編號的(說明限量肯定是假的﹐至少編號是隨便印的)﹑黃色底座的版本(聽說做工最差)﹑單車紅地毯有女王人偶的(人偶竟然做得跟女巫一樣醜惡﹐看樣子一定是最難賣的一款)﹑ 雙車紅地毯有女王人偶的﹑單雙車紅地毯有女王人偶分別又有限量編號的。這麼說山寨廠家肯定積壓了好多存貨才被逼推出如此多變化的變種來進行促銷。

很高興等了兩年沒白費功夫﹐終于收到了台價廉物美的完美品。這下終于可以跟之前的那台英國皇室同樣顏色(皇家深啡色)的Daimler DS420皇太后版本配對了﹐現在就差TrueScale今年稍後會推出的Rolls Royce Phantom V Canberra – HM The Queen 1960了﹗

1/43 Bentley Royal Golden Jubilee, 2002 HM Queen

Comments (11)

By admin, August 23, 2011 2:52 pm

Dell has released the updated version of its iSCSI flagship Equallogic PS6100/PS4100 series today. Why do I say it’s an updated version instead of a complete new generation? It’s because how Equallogic naming convention after their products, previously there are PS50, PS100, PS3500, PS5000, PS6000, if it’s a new generation, then it would be PS7000 series. After going through different documents, release notes, blog articles, I confirmed my guess is right. Best of all, you can mix and match any generation of Equallogic box in your SAN, I simple love this!

The followings are some of the major changes:

1. 4GB Cache vs 2GB Cache, probably won’t do much in terms of IOPS performance gain, I knew there are other vendors like EMC/Hitachi products using hugh cache up to 32GB/64GB, but real data shows the amount of cache isn’t the decision factor for IOPS.

2. The look of the new PS6100/PS4100: front and back looks almost identical to Dell’s SMB storage Powervault series, especially the back. It seemed to me that Dell has finally decided to use the same OEM hardware to save cost since the acquisition of Equallogic back in 2007. Especially, I don’t like the cheap look of the controller module on the back, it’s too Powervault MD1200/MD1220/MD3200 alike and the design of the new PS6100/PS4100 drive tray looks really ugly to me. Well, I am a person with design background, so I have a particular requirement for the Look! “Look is Everything!” once said by Agassi!

3. Dell claimed there is 60% more IOPS with PS6100, well, it’s kind of misleading as there are 24 drives in PS6100 vs 16 drives in PS6000. Simple math shows the extra 8 drives is only 50% more spindle if not counting the spare drivers. Also it’s interesting that if you use 24 x 2.5″ 10K drives on PS6100X, the IOPS performance is probably going to be similar to 16 x 3.5″ 15K drives PS6000XV. Of course if there is major upgrade in PS6100 controller hardware, then the story is different, just like PS5000XV and PS6000XV, controller module hardware is vastly different, the performance is also different even with the same number of spindles and disk RPM.

4. All PS6100 models has reduced its size from 3U to 2U with 24 x 2.5″ 15K/10K drives, it does save some rack space and power (probably not as there are more drives per 2U). Except for PS6100XV 3.5″ model which has been increased from 3U in PS6000XV to 4U, that extra 1U is for the additional 8 drives.

5. IMOP, the best products is PS6100XS which really combines the Tier 0 (7 400GB SSD) with Tier 1 (17 x 15K drives), this is the one I will purchase in our next upgrade cycle.

6. There is a Dedicated Management port on all PS6100XV, it’s a plus for some deployment scenarios.

7. On the VMware side, Dell said EqualLogic’s firmware 5.1 has “thin provisioning awareness” for VMware vSphere. As a result, the integration can save on recovery time and manage data loss risks. Dell has focused on VMware first given that 90 percent of its storage customers use virtualization and 80 percent of those customers are on VMware.

Over the third and fourth quarters, Dell and VMware will launch:

• EqualLogic Host Integration Tools for VMware 3.1. (Great!)
• Compellent storage replication adapters for VMware’s vCenter Site Recovery Manager 5.
• And Dell PowerVault integration with vSphere 5. (Huh? PV is becoming the low cost solution VMware SAN for 2-4 hosts deployment)

Finally, I wonder what does “Vertical port failover” mean in the PS6100 product PDF?

In fact, as usual I am more interested to know the hardware architecture of the new PS6100 series. One thing for sure PS6100 storage processor is  4-cores vs 2-cores in PS4100.

Can someone post a photo of the controller module please? What chipset does it use now?

Luckily, Jonathan of virtualizationbuster  solved part of the mystery, thank you very much!

-SAS 6GB Backplane – Finally. That’s all I can say. Will we see a big difference? With 24 drives SAS 6GB + SSD’s SAS 6GB is definitely needed. There have been independent review sites of other storage hardware showing that we are getting pretty close to maxing SAS 6GB out, especially with 24 SSD’s….I will be talking to the engineering team at EQL about the numbers on this in the future, I am curious.

-Cache to Flash Controllers – Bye-Bye battery backed cache, hello Flash Memory Based controllers. All I have to say is it’s about time! (this goes for ALL storage vendors- -hello? Its 2011!). Note*- In my review of the Dell R510, the PERC H700 was refreshed in the beginning of the year to support 1GB NV Cache, which is similar to what EQL using, and it rocks!

-2U Form Factor – Well, you can’t beat consolidation right? The more space we save in a rack the better + more drives in each array = more performance per U, per rack, per aisle. I have been a HUGE proponent of the 2.5in drive form factor for many reasons outside the room for this post. I have been running 2.5in server drives from Dell for about 4 years now and they have been rock solid.

-24 Drives vs 16 Drives- Well, you can’t beat more spindles, not much to say here but if you are a shop that thinks they NEED 15k drives you really need to take a look at the 2.5in 10k drives, their performance is steller, yes its slightly slower than its 15 3.5in counterpart, but performance is pretty close, especially with 24 of these.

Comments (0)

By admin, August 22, 2011 1:50 pm

A Good Example About Veeam’s Retention

Re: Job Set for 14 Mount Points – 46 exist
Posted: Thu Jan 20, 2011 4:57 amby tsightler

Logically your requirement to keep 14 restore points can’t be met until there are at least 14 more restore points from the last VBK.

In other words, you have a full backup, then 46 incrementals, then a full backup. If Veeam deletes the first full backup, and 33 incrementals, you’ll be left with 13 incrementals and a full backup, except those 13 incrementals will be worthless since the full backup on which they were based would be deleted. Setting the number of retention points for incremental backups sets the minimum number that will be retained. The maximum number might vary based on the interval that you run full backups (either active or synthetic).

This has been explained in another thread, but here it is again for simplicity. Let’s say you run backups every day, and you want to keep 14 restore points, and you run a full every week. After one week you get this:

Week 1: F…I…I…I…I…I…I

So that’s one full, and 6 incrementals, for a total of 7 restore points, now the second week:

Week 1: F…I…I…I…I…I…I
Week 2: F…I…I…I…I…I…I

So that’s two fulls and 12 incrementals for a total of 14 restore points, now it’s day 15, and you run another full and end up with this:

Week 1: F…I…I…I…I…I…I
Week 2: F…I…I…I…I…I…I
Week 3: F

You’ve told veeam to keep 14 restore points, but it can’t delete the full backup from week one, because that would invalidate all of the incremental backups from week 1 and leave you with the following:

Week 2: F…I…I…I…I…I…I
Week 3: F

That’s only 8 restore points, and you’ve told Veeam to keep 14. Veeam will not delete the first weeks backups until the last backup of Week 3 because at that point you’d have 21 backups:

Week 1: F…I…I…I…I…I…I
Week 2: F…I…I…I…I…I…I
Week 3: F…I…I…I…I…I…I

Thus if Veeam deletes Week 1 I get:

Week 2: F…I…I…I…I…I…I
Week 3: F…I…I…I…I…I…I

14 backups, which meets the requirement.

If you want Veeam to keep the exact number of restore points you can use reverse incrementals since in that case the oldest backup can always be deleted, otherwise, it’s no different than tape backup retention has been forever, the ratio of full to incremental backups will determine the number of backups that have to be kept to meet a minimum retention period.

Comments (0)

By admin, August 20, 2011 11:47 pm

今天經過雜誌店﹐發現MAGBOOK又出版了本以法拉利為主題的專刊﹐其實此類由著名英國汽車雜誌OCTANE發行的法記專刊早已出過了。簡單的來說﹐就是把很多期關於法拉利的內容再次集中成為一本專刊。因為已經擁有類似的專刊﹐放棄之際﹐細看之下原來內容有些不同﹐就是這本專刊多了另外一間姊妹雜誌社EVO的參與﹐而且價格比普通的一期OCTANE還要便宜﹐大喜之下﹐立刻買下﹗後來上OCTANE網站﹐發現居然沒有此書的蹤影﹐真奇怪﹗

喜歡法拉利的朋友﹐這本精彩的專刊絕對不能錯過。從五十年代賽場上的火紅250F講起﹐到六十年代優雅的250/275和DINO系列﹐然後去到七十年代科幻的DAYTONA/308和512BB﹐之後進入顛峰的八十年代328/TESTAROSSA/288 GTO/F40﹐接著就是九十年代的F50和348/F355系列﹐最後當然就是現今的360/430/ENZO/599/458/FF﹗

整本專刊就像一本濃縮的法拉利聖經﹐性價比絕對超值﹐我相信日後肯定會成為車模迷們爭相尋覓的另類經典收藏品之一﹗

Ferrari Supercars from the Publishers of EVO & Octane

Comments (0)

By admin, August 20, 2011 1:55 am

I came across an interesting article today that a new storage startup Nutanix can do what Google does for a living, using cheap hardware to simulate a big pool of distributed storage, of course it’s software based raid, something like the hardware based (actually it’s still software based + vendor specific designed array box) NetApps Network Raid, 3PAR and Compellent’s Cluster File System or even the newly launched vSphere Storage Appliance (purely software based). The concept is very nice, but at the end IOPS is still linked to how many spindles you have.

Yes it may be true that Nutanix’s storage has no boundary, but in my own opinion, only the next Google or Amazon alike company may need this kind of technology to drive down its operational cost, for most SMB (which Nutanix is aiming for), a low end SAN/NAS or even a free iSCSI target like MS iSCSI Target/Windstar shall do the job beautifully for their vSphere hosts. If Nutanix really targets for SMB, then they will definitely need to reduce the price and start to use really cost effective DIY servers as the selling point.

Finally, it’s interesting to know Nutanix is in fact founded by ex-employee of Google who probably the brain behind GFS, no wonder!

2013-05-05

Nutanix新款伺服器，可讓多臺伺服器肚內的儲存硬碟，共組成統一的叢集儲存槽，企業不用額外建置SAN儲存架構

伺服器新創廠商Nutanix在今年初由逸盈科技代理來臺，推出新款伺服器，可透過Nutanix Distributed File System（NDFS）技術，讓多臺伺服器肚內的儲存硬碟，共組成統一的叢集儲存槽，企業不用額外建置SAN儲存架構，伺服器節點也能共享儲存空間。

Nutanix區域銷售經理PK Lim表示，Nutanix產品設計源自於Google、Facebook、Amazon、Twitter等大型資料中心設計概念。

過去，企業資料中心架構，伺服器運算與儲存裝置都需要單獨採購，在伺服器運算設備以外，需要再額外架設SAN儲存環境來共享大量磁碟儲存空間。

但大型的資料中心，例如Google、Facebook、Amazon等，為了加快資料存取的I/O速度，讓伺服器運算的資料不需再儲存至外部硬碟，因此就設計了Google File System技術，讓伺服器肚內的SSD或儲存硬碟可以讓多臺伺服器共享，就可以取代企業傳統的SAN架構。

Nutanix產品取經Google等大型資料中心的儲存架構概念，設計出Nutanix系列伺服器，單臺機器採2U高度，機架式設計，每臺伺服器內含4個運算節點，多臺伺服器可透過網路埠串接，並共用每臺伺服器內的儲存空間。

Nutanix新型態伺服器採2U機架高度，內含4個運算節點，由逸盈科技代理來臺。

目前Nutanix產品依照系統規格可區分為NX-2000、NX-3000等系列，高階XN-3050最高可內裝3.2TB SATA介面的SSD，每個節點採2路設計，配置Intel Xeon E5-2670系列處理器。因此，一臺2U高度的XN-3050系列裝滿4節點後最高可達8顆處理器、64核心。每個節點提供2個10GbE和2個1GbE網路埠。企業可以依照建置規模，來選擇伺服器內裝的運算節點數量和儲存大小等。

Nutanix也和VMware合作，推出在Nutanix伺服器內預載vSphere 5.1虛擬化平臺的版本，目前已在臺上市。

Comments (0)

By admin, August 18, 2011 12:12 pm

Just found out the place where I stayed for two years during my teenage is the heart of the world’s famous British motor sports event ”Goodwood Motorsport Festival”, not to mention it’s also the head quarter of Rolls Royce. How come I didn’t know anything about it where I was there?

Comments (2)

By admin, August 17, 2011 9:21 pm

記得很多年前在英國讀中學的時候﹐曾經在倫敦的Bond Street見過一輛顏色很特別的Rolls Royce﹐黑色的上半身佩搭著銀色的下半身顯得格外的高雅﹐就如傳統英式結婚男禮服般同出一轍。

令人最記憶猶新的還是那位穿著整齊黑色制服戴帽子的司機伯伯﹐還有就是座位上那位令人印象深刻戴著高禮帽﹑滿頭銀髮西裝筆挺的老紳士﹐手裡居然還拿著根18世紀的黑色拐杖。

兩人的打扮完全跟當時九十年代初的倫敦時尚氣氛格格不入﹐這樣的裝束就如同從電影裡19世紀走出來的貴族一樣令人注目﹐這也令當時的我體會了什麼是維多利亞時代遺留下來的那份優越的英倫Class味道。

多年後﹐才知道當年的那台Rolls Royce型號叫做Phantom VI﹐好像只生產了三百多台﹐所以相當珍貴。記得香港港督曾經也有一輛﹐但在1997年7月1號香港回歸的當晚就跟隨英國艦隊回去了祖家。

自己很喜歡和欣賞這類六七十年代的英倫豪華汽車﹐那種獨特的氣派是現今Phantom款式所沒能提供的﹐你如果在特定的地方如倫敦市中小或者是古舊建築物前看到真車﹐就會馬上感受到那份與眾不同的大英帝國氣派。

要知道1比18的車模世界裡幾乎沒有這些這六七十年代的英倫豪華汽車﹐1比43的更寥寥可數﹐至于那些質量參次不齊的大陸山寨廠手版﹑加上沒有正式生產牌照當然也不能算數。

好消息終于來了﹐今年3月底TrueScale公佈了一項令所有Rolls Royce粉絲都為之振奮的新聞﹐那就是TrueScale將會陸續推出一系列跨越60年代到現今超過20款的Rolls Royce手版。幸虧TrueScale當時沒有一次性推出所有的款式﹐因為如此令人興奮的消息令我曾一度蒙蔽雙眼﹑認定要全部收下才心安﹐衷心感謝你TrueScale﹗

在過去的六個月內TrueScale緩慢地推出7款經典的車型﹐我可能對手版要求比合金來得要高﹐所以不太滿意手工質量有不少的瑕僻﹐如接口有明顯的溢膠﹐蝕刻片貼的不整齊等等。當然TrueScale的手版比其它廠來的還是比較有良心的﹐可以說是價廉物美﹐它並未因為掛著Rolls Royce的大名就大賺黑心錢﹐這點比AA和MC的賓利系列好得多﹗

自己比較喜歡的還是Phantom第五代和第六代﹐因為樣子比較典雅大體。相比之下TrueScale的那台黑色Phantom V Park Ward Black就沒有Phantom VI Mulliner Park Ward Black over Silver來的好看。

另外英女皇也擁有兩台Rolls Royce Phantom VI﹐其中一台更是特別的改裝型號﹐車頂明顯加高了﹐車尾窗戶也明顯加大了﹐最近的皇家婚禮盛典終于目睹到了它的真身。

下次有機會跟車模拍照的時候﹐會把這些古典Bentley﹑Rolls Royce和Daimler一字排開﹐背景襯以英倫古堡﹐對主題就叫“六七十年代的英倫紳士們”﹐那將是多麼的好看和耀目﹗

1/43 TrueScale Rolls Royce Phantom VI Mulliner Park Ward Black over Silver

Comments (0)

By admin, August 16, 2011 12:24 pm

世界上有多家攝影雜志擁有完整科學的鏡頭測試設備,一直堅持刊登各自的鏡頭測試報告,形成各自的風格和系列,其中公認為最權威的有德國的<攝影雜志>和<彩色攝影>,美國的<大眾攝影>,法國的<攝影師>和瑞典的<攝影>(即哈蘇實驗室測試報告).

對于同一款鏡頭的測試結果,它們有時趨于一致,有時卻出現偏差,以致引起不少爭議.引起這些偏差的原因主要有鏡頭的個體差异(特別是業余鏡頭存在質量不穩定現象),測試內容的不同(有的將机械質量和光學質量綜合評判,有的僅測試光學質量；有的僅判斷分辨率和MTF值,有的則綜合考慮色彩傳遞,畸變等) 和判斷標准不同(再客觀的測試,也是由許多個枯燥的數值組成的,要轉化為大家能夠理解的測試報告,最后必須用主觀的方式來表達,如分值,星號等).

這些測試應該相信哪一家好呢?于是有人想出將這些測試結果綜合起來判斷,最大程度地消除偶然性.德國攝影器材評論家克勞斯席洛夫1998年時將上述五家測試報告按一定的標准轉化為統一的5分制分值,并加權平均,用這种方法評測了當時市場上流行的205款135單反相机鏡頭,其中變焦鏡頭98款,定焦鏡頭107款,其中獲得了滿分5分的鏡頭有7款,它們是:

Leica Vario-Elmarit R 70-180mm f/2.8(3)
Leica Elmarit R 100mm f/2.8 APO macro(3)
Leica R 180mm f/2 APO(2)
Leica R 180mm f/3.4 APO(3)
Leica R 280mm f/4 APO-IF(3)
Leica R 400mm f/2.8(2)
Canon EF 400mm f/2.8 USM L ?(3)

用這种方法來評价鏡頭光學質量不失為一种有趣的方法,但也存在缺陷.后面括號內的數值表示測試的雜志數,并非上述五家雜志都做了測試,如果同為三家雜志所測試,但雜志卻有所不同,其可比性相對就差了,如果后來又加入新的測試,分值會有變化,變化有可能還不小,因此對于分值接近的兩款鏡頭,光學素質孰高孰低很難說,但如果兩款鏡頭分值相差較遠,它們的光學素質的差距是不可動搖的事實.

徠卡鏡頭名不虛傳,7個滿分鏡頭中占了6個,另一款為佳能的超級400mm大炮.徠卡果然沒有讓人們失望,各款鏡頭均表現出超俗的水平,不過話說回來,它的价格也是鶴立雞群,讓我們多數人不抱擁有的希望.

在4.5-4.99分之間的鏡頭也有7款,它們是:

Leica R 280mm f/2.8 APO-IF = 4.96(4)
Carl-Zeiss Sonnar T* 135mm f/2.8 = 4.87(2)
Canon TS-E 45mm f/2.8 = 4.61(2)
Canon EF 300mm f/2.8 USM L = 4.6(4)
Leica Vario-Elmarit R 80-200mm f/4 = 4.6(3)
Canon EF 135mm f/2 USM L = 4.54(4)
Minolta AF 200mm f/2.8 APO G = 4.52(3)

由上可以看出,變焦鏡頭与專業的定焦鏡頭相比,在光學質量還是有所遜色.除上述兩款徠卡變焦鏡頭外,超過4分的變焦鏡頭只剩下3款:

Carl-Zeiss Vario-Sonnar T* 28-85mm f/3.3-4 = 4.17(4)
Canon EF 70-200mm f/2.8 USM L = 4.07(4)
Nikkor AF 35-70mm f/2.8 D = 4.01(5)

而超過4分的定焦鏡頭數量可觀,包括上述定焦鏡頭總共達到45款. 相對与長焦鏡頭來說,廣角鏡頭(無論是定焦還是變焦)由于涵蓋率和构造的關系,更難達到高分.超過4分的定焦廣角鏡頭(焦長28mm以下)有5款:

Leica R 19mm f/2.8 = 4.28 (3)
Carl-Zeiss Distagon T* 21mm f/2.8 = 4.17 (2)
Leica R 28mm f/2.8 = 4.16 (3)
Pentax SMC-F 28mm f/2.8 = 4.13 (3)
Leica R 28 f/2.8mm = 4.12 (2)

分數最高的超廣角變焦鏡頭是:

Nikkor AF 2.8 20-35mm D = 3.46 (4)
近兩年中又有許多新鏡頭面世,最近克勞斯席洛夫對上述評价做了調整,去掉了徠卡鏡頭和康太時單反相机鏡頭,加入了康太時G系列鏡頭(旁軸平視取景相机),即所列鏡頭全部為自動對焦鏡頭,共計變焦鏡頭115款,定焦鏡頭103款.對評分標准也做了一些修改,去掉了爭議較大的德國<攝影雜志>測試,換之以英國的<業余愛好 者>雜志測試.綜合評分達到4. 5分以上的鏡頭被定為”杰出鏡頭”,目前共有18款,它們是:

Pentax SMC-FA 85mm f/1.4 = 4.85 (3)
Carl-Zeiss Planar T* 45mm f/2 G = 4.77 (3)
Sigma AF 50mm f/2.8 Macro EX = 4.77 (3)
Canon EF 135mm f/2 USM L = 4.72 (4)
Canon TS-E 90mm f/2.8 = 4.72 (2)
Nikkor AF 60mm f/2.8 Micro = 4.63 (4)
Carl-Zeiss Sonnar T* 90mm f/2.8 G = 4.62 (3)
Canon EF 400mm f/2.8 USM L = 4.60(3)
Minolta AF 100mm f/2 = 4.60(3)
Canon EF 100mm f/2.8 Macro = 4.57(5)
Carl-Zeiss Biogon T* 28mm f/2.8 G = 4.57 (3)
Nikkor AF 105mm f/2.8 Micro = 4.54(5)
Nikkor AF 85mm f/1.8 D = 4.53(3)
Canon EF 85mm f/1.8 USM = 4.52(4)
Minolta AF 200mm f/2.8 APO G = 4.50(4)
Canon EF 300mm f/2.8 USM L = 4.50(3)
Nikkor AF-S 300mm f/2.8 ED D = 4.50(2)
Pentax SMC-FA* 200mm f/2.8 ED IF = 4.50(2)

超過4分的變焦鏡頭有9款:

Nikkor AF-S 28-70mm f/2.8 ED D = 4.24 (3)
Minolta AF 80-200mm f/2.8 APO G = 4.18 (3)
Canon EF 70-200mm f/4 USM L = 4.16 (2)
Nikkor AF 35-70mm f/2.8 D = 4.11 (5)
Nikkor AF-S 80-200mm f/2.8 ED D = 4.11 (3)
Canon EF 80-200mm f/2.8 L = 4.07 (5)
Canon EF 70-200mm f/2.8 USM L = 4.06 (3)
Sigma AF 70-200mm f/2.8 EX (HSM) = 4.05 (5)
Canon EF 28-70mm f/2.8 USM L = 4.01 (5)

超過3.75分的28mm以下廣角鏡頭有9款:

Carl-Zeiss Biogon T* 28mm f/2.8 G = 4.57 (3)
Carl-Zeiss Biogon T* 21mm f/2.8 G = 4.28 (2)
Nikkor AF 28mm f/1.4 D = 4.00 (3)
Pentax SMC-FA 28mm f/2.8 = 3.89 (4)
Canon EF 28mm f/2.8 = 3.82 (4)
Canon EF 24mm f/1.4 USM L = 3.82 (3)
Nikkor AF 24mm f/2.8 = 3.77 (3)
Canon EF 20mm f/2.8 = 3.75 (4)
Canon EF 24mm f/2.8 = 3.75 (3)