2018年戴尔科技峰会,迈克尔·戴尔演讲全文翻译

19岁

从学生宿舍起家,成立了自己的公司。

27岁

当同龄人拿着为数不多的薪水挣扎在温饱线上,他已一跃登上《财富》杂志,成为全球最年轻的CEO。

32岁

坐上德克萨斯州首富位置,被《财富》杂志评为美国40岁以下最富有的人。

42岁

临危受命,再度出山,颠覆戴尔公司赖以为基的直销模式,挽救公司于危难,并于9年后以670亿美元收购EMC,一手促成科技史上最大并购。

迈克尔·戴尔的传奇还在继续,在当今工业4.0如火如荼、数字化趋势一浪更比一浪高的时代,企业组织如何在科技洪流中屹立不倒?听一听戴尔科技集团创始人、CEO兼董事长迈克尔·戴尔的见解吧。

戴尔科技集团创始人、CEO兼董事长 迈克尔·戴尔

2018年10月17日戴尔科技峰会迈克尔·戴尔演讲视频:点击结尾处阅读原文观看。



1

Good morning! It’s a wonderful of beauty all today. And I want to thank you all for joining us and spending time with us at this Dell Technologies Forum. You are going to learn a lot about what we are learning from all of you, our customers around the world.

上午好!真是个美好的日子,感谢大家抽出宝贵的时间来参加戴尔科技峰会。今天,我们将与各位分享来自世界各地客户的信息。

It is a super interesting time, you know, the rate of change in our industry is only increasing and the world technology is really expanding.

这是一个十分激动人心的时代,行业步伐不断加快,全球科技也在持续扩张。

You know, when I travel around the world, I hear a lot of our customers, really all of our customers talking in some way about the digital transformation and digitization of our world.

我到世界各地出差的时候,听到很多客户,可以说几乎是我们所有的客户,都在某种程度上谈论着数字化转型和全球数字化。

And it’s affecting every aspect of business and society: education, government, energy, healthcare, manufacturing, retail, ect. No part of our world is unaffected by the digitization and digital transformation.

数字化正在全方位影响着企业与社会,包括教育、政府、能源、医疗、制造业、零售业等各行各业的方方面面。当今时代,数字化浪潮已经席卷世界每个角落。

And for technology, what’s happening is that is becoming more important. It used to be that the IT, the information technology was the in the IT department.

对于科技而言,越来越重要的改变正在发生。过去我们谈IT,说信息技术只是IT部门的事情。

Well, now you have technology in every department, in every part of a company. And I have one customer described it to me, and they said IT used to be a chore, and now it’s core. It’s absolutely essential to our business and our success and our survival.

但是今天,任何一个部门都有IT的身影,IT遍及公司各个地方。曾经有一个客户这么跟我说过:IT过去被当作苦差事,现在已经成了我们整个企业、战略和生存的核心。

And there is at the center of this, a virtuous cycle where we have data, and of course we have much more data now than we had five years ago. And we will have way way more in five years and ten years from now.

在所有这些变化的中心,是一个关于数据的良性循环。当然,我们今天面对的数据量要比五年前多得多,相对地,再过5年、10年,数据量还会更大。

And this data is allowing us to make our products and our services better, to enhance them and provide better outcome and better value and better success for our customers and our users. And then we get more customers, and guess what, we get more data, because where there did the data come from.

基于这些数据,我们可以不断提升产品、服务质量,从而为我们的客户和用户带来更多产出、更高价值和更大的成功。随后我们的客户继续增加,然后你猜会发生什么——我们又得到了更多数据,因为那就是数据的来源。

2

Well, we all know that the number of connected things is increasing very very dramatically. Certainly, in China the progress in the last 40 years has been amazing. We are honored to have been participated in that in the last 20 years, and we’ve seen incredible developments in the last several decades.

大家都知道,互联设备的数量正在迅猛增长。可以看到中国可以在过去的40年里取得了令人瞩目的成就,我们也非常荣幸能够在过去的20年参与到中国的进步当中,并且见证了中国几十年来的突飞猛进。

But now everything is becoming intelligent, from automobiles to aerospace to medical devices to any kind of thing you can imagine is becoming digitized. So this is creating even more and more data.

现在一切都在变得智能,从汽车到航空,到医疗设备到所有你能想象得到的东西,都在经历着数字化,如此一来,越来越多的数据就会产生。

What’s also happening is now, the next generation of computer science, artificial intelligence, machine learning, deep learning, neural networks, allowing these data to be used much more effectively, and this allows us again to enhance the products and services that you all create and we create, and we end up with even more customers and more data.

同时还在发生的,是下一代计算科学、人工智能、机器学习、深度学习、神经网络等,这些技术让数据得到高效利用,再次方便我们改进产品和服务,最后我们就能获得更多的客户和数据。

And the cycle just repeats itself, but going faster and faster. And of course the networks are getting faster as well just around the corner we have the fifth generation cellular network, a thousand times faster, a thousand times lower latency.

这个良性循环周而复始,不断加快。当然网络也会越来越快,5G网络时代即将来临,带来一千倍的速度和千分之一的延迟。

It’s not so we can talk on the phone faster, it’s all about the data. And this explosion data is only going to continue. So there is lots of it, it’s going to grow quite tremendously. And it requires significant tools and capabilities to be able to take energy of this.

这不是指我们打电话会更快捷,而是说数据(传输更迅速),而且这种数据爆炸还会继续。所以,如此多的数据进行无边无际的扩张,就需要有足够的赋能工具把它们充分利用起来。

What we also see in our industry over long pretty time is the pendulous swinging back and forth, from centralized to decentralized, back to centralized and decentralized. And what we believe is that the answer isn’t either or is actually both.

同时也可以看到,在相当长的时间内,我们的行业就像钟摆一样来回摇动,从集中化到分散化,再回到集中化,后又变成分散化。然而,最佳的策略不可能仅是其中之一,而是将两者都覆盖。

And with the incredible intelligence that is being embedded in every device, there will be a boom in what we would refer to as edge computing. So think about anything that you would interact with becoming intelligent because the cost of a sensor is approaching $0.

有了嵌入到每个设备当中的卓越智能,边缘计算将会迎来一大片欣欣向荣的场面。因此,任何一个你所参与的场景都会变得智能,因为传感器的成本已经基本降为0。

And so think of all of the connected devices as not just the ones we were used to, like computers or phones, but now everything gets connected. So it’s going to be a massive build-out in edge computing with very low latency inside the networks.

除了计算机或手机这些我们习以为常的东西,现在所有事物都已经实现了互联。因此,未来边缘计算会大量扩张,在非常低的网络延迟中建设起来。

And we believe that you will have an architecture where you have this huge build-out on the edge, you will have the distributed core of computing and you will have clouds: public clouds, private clouds, hybrid clouds is the multi clouds world, we believe the cloud is not actually a place but rather a way of doing IT.

相信你将会打造这样一种架构:边缘有大规模的扩建,同时还有分布式计算核心。在多云环境下,你也会有云:公有云、私有云以及混合云。当然云并不仅仅是一个场所,而是构建IT的一种方式。

And you can do that yourself you can ask somebody there for you. And in reality, you will most likely have a multi cloud environment where you decide where the right places for any given workload and that will change over time.

你可以自己做,也可以让其他人帮忙。现实中,你最有可能拥有一个多云环境,来决定既定的工作负载应该放在什么地方,而这会随着时间推移做出调整。

3

So a little over two years ago, we completed the combination, bringing together Dell, EMC, VMware, Pivotal many other companies to create the world’s leading infrastructure hardware and software company, in the world essential infrastructure company.

两年多以前,我们完成了合并,把Dell、EMC、VMware、Pivotal等很多其他公司,整合成一家在基础架构、硬件和软件领域都处于世界领先地位的公司,现在的戴尔科技集团正是一家关键的世界级基础架构公司。

I should stop and say a very big thank you to all of our customers in China because our business here has been growing at extraordinary rates. And we were very appreciative for the trust and confidence you place in us. We’ve also seen a similar development around world where our business is growing quite rapidly.

在这里,我要表达一下对所有中国客户的谢意,我们在中国的业务得到了长足发展,非常感谢你们给予我们的信任和信心。与此同时,我们的业务在全球各地都有着类似的迅猛势头。

And as you hear about through this Dell Technologies Forum and in the exposed area will you see many demonstrations and many of our partners. We see our customers really simultaneously dealing with four things all at the same time, probably the most importantly is the digital transformation, the digitization.

在戴尔科技峰会现场和外面的展区,有许多关于合作伙伴的案例展示。可以看到,我们的客户同时在应对四个趋势,可能这其中最重要的就是数字化转型,数字化。

How do we take with these data and how do we use software and AI to express our competitive advantage? If you have a lot of data, you’re not already starting to do this, I would say you probably do get wrong, and your competitors are probably going faster or you have to worry that there will be new company coming along to disrupt your business.

那么我们怎么处理这些数据,怎么利用软件和人工智能取得竞争优势?如果你空有大量数据而没有开始做这些事情,那就失策了。因为你可能会被竞争对手赶超,或者冒出来一家新公司颠覆了你的业务。

And the digital transformation is really about reimagining and organization. When you think about the cost of predicting something going to 0,because of artificial intelligence.

数字化转型的关键就是重塑和组织。由于人工智能的出现,预测成本已经降为0。

When you think about the ability to dramatically improve all the products and services with all this data information, we can’t operate as organizations the way we used to, you know last year or five years ago or ten years ago, so requires all organizations rethink what their business was.

当你想利用这些数据信息极大地提升产品和服务时,就不能按照1年前、5年前甚至10年前的老方式运作了,而需要重新思考业务内容到底是什么。

In many ways, you know, the internet the last thirty years with microprocessors and all those wonderful things was just a foundation layer upon which we are now building the next 30 years, which I think will be even far more exciting, So that is the digital transformation。

从很多方式上讲,互联网过去30年里,出现的类似微处理器这些技术只是基础,从现在开始如何新建下一个30年,才是我们要更加关注的问题,这就是所谓的数字化转型。

And certainly with VMware and Pivotal and Dell EMC, we have great capabilities to help our customers address that.

我们有VMware、Pivotal和戴尔易安信,可以凭借非常卓越的能力帮助客户来实现这一目标。

The second transformation is in the information technology itself , and just like in automobiles, there is a massive effort going on around self-driving, automobiles and electrification. In the data center, there is a massive movement toward the self-driving data center.

第二个转型是信息技术转型,正如汽车行业所发生的那样,大规模的努力围绕着自动驾驶、汽车和电气化展开。数据中心也是一样,各方都投入大量精力到自动化数据中心建设之中。

The autonomous data center, where all the operations can be automated and this providing credible flexibility and credible scalability and also allows for the creation of this multi cloud environment where workloads are moving around and you’re figuring out exactly where the right place for any given workloads.

自动化数据中心,具备无与伦比的灵活性和扩展性,所有的操作都可以自动进行。由此为可移动工作负载的多云环境奠定基础,在这里对于任何既定的工作负载放在什么地方合适,你都可以做出最准确的决策。

And we know from our experience that when we automate and modernize a data center for the predictable workloads, which for most organizations are roughly 85% to 90%, and on premise solution, costs several times less than the public cloud. And this is very surprising to many people.

从经验得知,当我们为可预测的工作负载进行数据中心自动化和现代化时,大部分组织85%-90%的工作负载都是可预测的。而且让许多人意外的是,本地的解决方案的成本只有公有云的几分之一。

But we’ve seen now large number of examples where customers said: I can gonna move everything to the public cloud and they are surprised to find that not everything moves so well to the public cloud and in many cases is more expensive.

现在这样的例子层出不穷,比如有的客户说要把所有业务都迁移到公有云上面,最后却惊讶地发现,并非所有的东西都能顺利迁过去。实际上在很多情况下,迁移到公有云反而花费更高。

Now, all of you will probably find some applications, particularly where there’s high variability in demand where you could make use of a public cloud , that’s why we believe it will be this multi-cloud era. So the transformation of the IT is all about the software defined data center, also known as the self-driving data center. And it’s going to be modern, it’s going to be automated, it’s going to be credibly efficient.

现下环境中,尤其当是需求浮动性较大的时候,所有人都有可能选择到公有云上采购一些应用,这就是多云时代的体现。因此IT转型的根本就在于软件定义的数据中心,即现代化、自动化的高效的自动数据中心。

The third transformation is around the workforce. How do we create the right tools and capabilities to enable and inspire the people inside your organization to be as productive as they can possibly be, not just when they are in the office and not just on one device, but perhaps on many devices. And how do we do that in a secure manner.

第三个是生产力转型。即如何创造绝佳的赋能工具,在最大程度上激发员工生产力,而不是使其局限于在办公室里或者某个设备上。我们究竟怎么做才能安全地实现这一转型呢?

You know, the original business that we started with at Dell, the PC business, I’m proud to say we’ve now had twenty three quarters growth, and gaining share in that business.

众所周知,戴尔从PC业务起家,至今已经连续23个季度保持增长,并且占有越来越多的市场份额。

Well, some companies take it a different choice, we believe that it’s still an incredibly important part of how productivity gets done and is and in fact the most important device is coming into productivity.

尽管其他公司选择了不同的战略,不过我们坚信,PC仍然是实现生产力至关重要的一环,实际上,PC已经开始向生产力转化了。

Finally, if you imagine a world where everything is connected and everything is digitized, from the water supply to the trains and the air traffic control and banking and healthcare, while we have to think carefully about security.

最后一点,在这样一个万物互联的数字化世界,从供水系统到铁路运输、空中交通管制、银行业,再到医疗,方方面面都必须严肃考虑安全问题。

How do we build security into our products so that they are protected and safe and the most valuable data is protected.

那么如何将安全概念内嵌到产品当中,使它们免受侵害,以保护大家最有价值的数字资产呢?

So by bringing together all the family members of Dell Technologies, we have been able to create the world’s leading infrastructure company. And our objective is to be your best partner in your digital future.

为了实践安全性,我们把戴尔科技家族所有成员整合到一起,组建了世界领先的基础架构公司,目的是在你们数字化过程中保驾护航。

So thank you, ladies and gentlemen for joining us. Ah, thank you so much for all the support you provided us over the last twenty years. We continue to invest significantly in China as you heard from Chenhong. I’m looking forward to the rest of my visit here. Thank you.

女士们、先生们,谢谢大家今天来参加戴尔科技峰会,非常感谢你们在过去20年风雨相随,给予我们巨大的支持。正如你们在前面听黄陈宏说的,我们会继续大力加码在中国的投资。非常期待接下来在这里的行程。谢谢大家!

Related:

  • No Related Posts

Re: 使用Bloom Filter提高索引效率(下)

DataDomain中的重复数据检查:

上篇我们介绍了Bloom Filter,以及其以“正确率换空间”的特性。而它的这种特性,非常适合需要对海量数据进行快速检索而内存空间又有限的场景。比如在重复数据删除系统(DataDomain)中可以配合fingerprint Cache,加速重复数据的检查。

我们都知道DataDomain(DD)采用数据块级的去重技术,文件或数据流被切割为较小的segment, 然后使用160-bit的SHA1算法求取每个segment的哈希值,这些哈希值被称为segment fingerprint,用来标志一个segment。具有相同fingerprint的两个segment即为duplicate segment。当有数据写入的时候,DD会首先把数据切成很多小的segment,然后计算出fingerprint,再与系统上现有的fingerprint进行比较,来判断这个数据块是否是重复数据。

为了保证DD的写性能,我们希望快速判断新来的数据是否是重复数据。

最直接的方式是在内存中维护所有的fingerprint。fingerprint的比较直接在内存中完成,性能最优。但是采用这种方式会带来巨大的内存开销。假设现有一台DD上有8TB去重后的数据,以平均一个segment大小8KB来计算,那么系统上一共存在10亿个这样的fingerprint。每个fingerprint 20byte, 我们需要20GB的内存只是来存储所有的fingerprint!显然这种方式是不能接受的。

既然内存不够维护所有的fingerprint,那么通常的做法就是缓存一部分fingerprint在内存,当有新的fingerprint需要比较的时候,先检查Cache,如果Cache没有再检查disk上的。DD就是这样做的,其使用一种叫做Locality Preserved Caching(LPC)的缓存技术来加速重复数据判断,该技术利用了备份数据的一个特性:对同一个文件进行多次备份,其segment出现的序列是相同的。假设有一个1MB的文件被切成了100多个小的数据块,在每次对这个文件进行备份的时候,这些segment总是以相同的顺序出现的。即使某次文件的内容中有了一些修改,会有新的segment产生,但是剩下的segment也会以同样的顺序出现。利用这个特性,当DD发现一个写入的segment和DD上的segment x重复的时候,会尽量把与x来自同一个文件的后续fingerprint放到Cache中,以提高后面的Cache命中率。

但是只有Cache还不够,我们并不能保证在Cache中没有命中的fingerprint一定不在DD上。所以当写入的segment在Cache中没有对应的fingerprint时,我们还是要去disk上找, 这就带来性能瓶颈。那么DD是怎么解决这个问题的呢?就是用前面说的Bloom Filter嘛!

Bloom Filter在DD中的应用:

使用Bloom Filter来表示DD上已有fingerprint的集合,从而只需要很少的内存就可以快速判断出待写入的segment哪些肯定是新的segment,哪些可能是重复的segment。把肯定是新的segment过滤掉,DD只需要继续检查那些可能重复的segment。

DD将这个Bloom Filter称为Summary Vector。和典型的Bloom Filter一样,Summary Vector也由一个很长的位向量组成,这个位向量被初始化为全0。当有新的fingerprint需要加入的时候,会将该fingerprint传入不同的HASH函数,计算出几个HASH值,再映射到位向量上将对应的位设为1.

当需要查询该fingerprint是否在DD上时,同样将通过fingerprint计算出的HASH值,映射到位向量上,如果对应的位置有一个不为1,则可以判断出该fingerprint一定不在DD上。

使用Summary Vector在存在一定误判的基础上大大降低了判断重复数据时对内存的要求,以m/n=8为例,对于8TB的非重复数据,Summary Vector只需要1GB的容量就可以把绝大多数(2%的误差率)的新segment过滤掉!

下面这张图描述了DD进行重复数据判断时的流程,我们可以看到DD分别通过Cache和Summary Vector来过滤掉不必要的segment index lookup的操作。Cache会首先过滤掉很大一部分重复的segment, 接下来Summary Vector又会把绝大多数新的segment过滤掉,受益于Bloom filter以及Cache,DataDomain系统可以减少99%的磁盘访问,从而利用少量的内存空间大幅提高了数据块查重性能。

总结:

除了在重复数据删除系统中的应用,Bloom filter还被广泛应用于各种领域,比如拼写检查、字符串匹配算法、网络包分析工具、Web Cache、文件系统。下面列了一些Bloom Filter在其他领域的应用,有兴趣的同学可以看看:

  1. Google BigTable, Apache HBase Apache Cassandra, 和 Postgresql使用Bloom Filter减少对不存在的行列的磁盘查询, 提高数据库查询操作性能。
  2. Google Chrome 使用Bloom Filter来判断恶意URLs。任何URL都会先经过客户端的Bloom Filter作检查,只有当Bloom Filter返回positive的时候,才会将URL发到Google服务器上作完整的检查。
  3. Bitcoin 使用Bloom Filter加速钱包的同步

作者简介:

Walker Huang from Avamar

2016年2月加入Dell EMC Avamar,现在主要从事Avamar与DataDomain集成数据备份系统的开发和维护。资深跑友,有7次全程马拉松完赛经历。

Related:

  • No Related Posts

VxRail124 Fake Replacement with G410(MR)

VxRail Node Fake Replacement

最近做了个HengTongNode Fake Replacementcase,这个case因该是国内首例,没有官方的PG。这个case得到了VxRail Support TeamEE Team的大力支持,现分享给大家。

Case背景:HengTongVxRail 120是广达Gen2 8nodeclusternode1power键的LED灯不亮,更换了node后,故障依旧。由于这个LED是集成在chassis上,并不是一个FRU部件,所以只能通过更换chassis来解决这个故障,但是单个chassis也不是FRU,并且用户的业务不允许停机更换,最终在supportEE的建议下,通过MR流程,重新order一套4个节点的G410,通过整机更换的办法完成node1node4的在线替换换。故障现象如下图:

fake1.png

Fake replacement:所谓的fake replacement就是用新的node one by one的替换原来node,新的node IDhostnameip address会继承原有的配置信息,变相实现chassis的在线更换。Fake replacement=node remove+node expansion,相关PG都可以在Solve Desktop上找到。但是有一点需要注意的,node remove要求现有的vxrail最低版本必须是4.0.301,如果当前版本低于4.0.301,必须先升级到目标版本。

HengTong的这个case为了确保安全,我们是按照node4321的顺序one by one的替换原有node,即是踢一个/加一个,然后再做vsan disk rebalance。更换完所有4node总共花了4天时间,其中大部分时间都是disk rebalance

实施步骤:

  1. Node remove

具体操作步骤,我就不在这里赘述了,大家按照“how to remove……”PG执行就可以了。这一部分中我要再次强调下remove node之前一定要严格按照PG中公式确保当前vxrailVSAN,CPUMember利用率不高于80%,否则remove后可能导致DU/DL,公式如下:

VSAN used Capacity % = Used Total / (Current capacity – Capacity to be Removed)

CPU_used_% = Consumed_Cluster_CPU/(CPU_capacity – Plan_to_Remove_CPU_sum)

Memory_used_% = Consumed_Cluster_Memory/(Memory_capacity – Plan_to_Remove_Memory_sum)

如果利用率超过80%stopthen call 800 for help!

Remove node之后,该nodeVxRail Manager中会显示NA

fake2.png

  1. Node expansion

把原有node上的光纤线和网线换到新的node上,power on新的node,进到BIOS中,重新设置时间和BMCIP。注意,BIOS的时间是GMT-8,尽量保证新加node的时间和原有时间同步,否则添加节点时可能会失败。

如果用户环境的management网络配有vlan,则需要在加节点前配置vlan id

esxcli network vswitch standard portgroup set -p “Management Network” –v <VLAN_ID>

esxcli network vswitch standard portgroup set -p “VM Network” -v <VLAN_ID>

新节点重启完成之后,在VxRail Management GUI里会检测到这个node,然后按照expansion PG完成add node:

fake3.png

在填写新节点的分配ip地址的时候,esxi,vsan和vmotion起始ip和终止ip要写原来的ip地址

fake4.png

接下来要填写主机信息:

fake5.png

注意这个管理账户,是用于esxi主机和vcenter认证的,在3.5版本的vxrail环境中,系统自动产生用户名和密码,4.0版本在安装实施密码需要手动设置。如果用户的vxrail4.0.x是从3.5版本升上去的,用户是不知道密码是什么的,这时候我们需要用BlowFishUtil.jar来解码这个passwor,上图中的密码就是解码后的。具体步骤请参考:

https://emcservice.my.salesforce.com/kA5f10000008YDpCAM

最后单击“验证”。

注意:如果用户的vxrail版本是从3.5升级到4.0的,在添加4.0版本的node的时候,由于4.0采用了新的认证算法,可能导致验证失败。解决方案:

  1. ssh或者kvm登录新节点的esxi
  2. vi /etc/pam.d/passwd
  3. password requisite所在行最后添加enfore=none,修改后的passwd文件如下

fake6.png

  1. 保存退出
  2. 重新“验证”
  3. Node添加完成

3. VSAN disk rebalance

在实施remove的阶段,node进入维护模式后,会自动发起vsan resync,也就是reblance,这个过程所花时间视用户vsan容量大小决定,HengTong每个node大概花了3个小时左右。

由于这个case是连续换4个node,所以没新加一个node,不需要再手动rebalance一次,因为在后面的remove阶段还是会自动rebalance一次;如果只是换一个节点的话,还是建议手动发起一次。

HengTong的case历时将近4天,MR的G410整机替换了原来的chassis,至此fake replacement完成。

Related:

  • No Related Posts

Re: 【分享】EMC产品备件信息数据库网址(https://alliance.emc.com/PCDPartners/Defaultkm.asp)

备件坏了,不知道怎么找part number?

想扩容,不知道备件是否兼容?

想预定备件,不知道上哪里找part number?

EMC产品备件信息数据库(https://alliance.emc.com/PCDPartners/Defaultkm.asp),提供所有EMC硬件产品备件信息查询功能。只能按产品、备件号、微码。。。功能查询。

不多说,上王道:

1.jpg

VNX、VNXe、CLARiiON、Celerra这么多规格DAE,这么多规格磁盘,上哪里找part number号?(一册在手,VNX、VNXe、CLARiiON、Celerra备件信息无后顾之忧)

注意:这个文档需要powerlink访问权限,部分用户可能无法下载。

http://powerlink.emc.com/km/live1/en_US/Offering_Technical/Technical_Documentation/All_CLARiiON_Disk_and_FLARE_OE_Matrices.pdf?mtcs=ZXZlbnRUeXBlPUttQ2xpY2tDb250ZW50RXZlbnQsZG9jdW1lbnRJZD0wOTAxNDA2NjgwNDI4ZmRhLG5hdmVOb2RlPTBiMDE0MDY2ODAzNTE4M2Q_

Related:

  • No Related Posts